Merge branch 'main' into sami/flash

Adds support glm 4.7 and MiniMax M2.1

Manual testing:
Tensor + Pipeline execution of both models.

Closes #1141 and #1142

.github/workflows/build-app.yml

@@ -1,6 +1,7 @@
 name: Build EXO macOS DMG
 
 on:
+  workflow_dispatch:
   push:
     tags:
       - "v*"
@@ -18,6 +19,7 @@ jobs:
       SPARKLE_ED25519_PRIVATE: ${{ secrets.SPARKLE_ED25519_PRIVATE }}
       SPARKLE_S3_BUCKET: ${{ secrets.SPARKLE_S3_BUCKET }}
       SPARKLE_S3_PREFIX: ${{ secrets.SPARKLE_S3_PREFIX }}
+      EXO_BUG_REPORT_PRESIGNED_URL_ENDPOINT: ${{ secrets.EXO_BUG_REPORT_PRESIGNED_URL_ENDPOINT }}
       AWS_REGION: ${{ secrets.AWS_REGION }}
       EXO_BUILD_NUMBER: ${{ github.run_number }}
       EXO_LIBP2P_NAMESPACE: ${{ github.ref_name }}
@@ -34,7 +36,7 @@ jobs:
 
       - name: Derive release version from tag
         run: |
-          if [[ "$GITHUB_REF_NAME" == "test-app" ]]; then
+          if [[ "$GITHUB_REF_NAME" == "test-app" || "${{ github.event_name }}" == "workflow_dispatch" ]]; then
             VERSION="0.0.0-alpha.0"
             echo "IS_ALPHA=true" >> $GITHUB_ENV
           else
@@ -47,6 +49,32 @@ jobs:
           fi
           echo "RELEASE_VERSION=$VERSION" >> $GITHUB_ENV
 
+      - name: Compute build version from semver
+        run: |
+          VERSION="$RELEASE_VERSION"
+          # Extract major.minor.patch (strip prerelease suffix)
+          BASE_VERSION="${VERSION%%-*}"
+          MAJOR=$(echo "$BASE_VERSION" | cut -d. -f1)
+          MINOR=$(echo "$BASE_VERSION" | cut -d. -f2)
+          PATCH=$(echo "$BASE_VERSION" | cut -d. -f3)
+
+          # Extract prerelease number (e.g., "alpha.2" -> 2, or 999 for releases)
+          if [[ "$VERSION" == *-* ]]; then
+            PRERELEASE_PART="${VERSION#*-}"
+            PRERELEASE_NUM="${PRERELEASE_PART##*.}"
+            # Default to 0 if not a number
+            if ! [[ "$PRERELEASE_NUM" =~ ^[0-9]+$ ]]; then
+              PRERELEASE_NUM=0
+            fi
+          else
+            PRERELEASE_NUM=999
+          fi
+
+          # Compute: PRERELEASE + (1000 * PATCH) + (1_000_000 * MINOR) + (1_000_000_000 * MAJOR)
+          BUILD_VERSION=$((PRERELEASE_NUM + 1000 * PATCH + 1000000 * MINOR + 1000000000 * MAJOR))
+          echo "EXO_BUILD_VERSION=$BUILD_VERSION" >> $GITHUB_ENV
+          echo "Computed build version: $BUILD_VERSION from $VERSION"
+
       - name: Ensure tag commit is on main
         if: github.ref_type == 'tag'
         run: |
@@ -85,11 +113,22 @@ jobs:
           uv python install
           uv sync --locked
 
+      - name: Install Nix
+        uses: cachix/install-nix-action@v31
+        with:
+          nix_path: nixpkgs=channel:nixos-unstable
+
+      - name: Configure Cachix
+        uses: cachix/cachix-action@v14
+        with:
+          name: exo
+          authToken: "${{ secrets.CACHIX_AUTH_TOKEN }}"
+
       - name: Build dashboard
         run: |
-          cd dashboard
-          npm ci
-          npm run build
+          DASHBOARD_OUT=$(nix build .#dashboard --print-build-logs --no-link --print-out-paths)
+          mkdir -p dashboard/build
+          cp -r "$DASHBOARD_OUT"/* dashboard/build/
 
       - name: Install Sparkle CLI
         run: |
@@ -162,11 +201,12 @@ jobs:
             -configuration Release \
             -derivedDataPath build \
             MARKETING_VERSION="$RELEASE_VERSION" \
-            CURRENT_PROJECT_VERSION="$EXO_BUILD_NUMBER" \
+            CURRENT_PROJECT_VERSION="$EXO_BUILD_VERSION" \
             EXO_BUILD_TAG="$RELEASE_VERSION" \
             EXO_BUILD_COMMIT="$GITHUB_SHA" \
             SPARKLE_FEED_URL="$SPARKLE_FEED_URL" \
             SPARKLE_ED25519_PUBLIC="$SPARKLE_ED25519_PUBLIC" \
+            EXO_BUG_REPORT_PRESIGNED_URL_ENDPOINT="$EXO_BUG_REPORT_PRESIGNED_URL_ENDPOINT" \
             CODE_SIGNING_IDENTITY="$SIGNING_IDENTITY" \
             CODE_SIGN_INJECT_BASE_ENTITLEMENTS=YES
           mkdir -p ../../output
@@ -294,5 +334,5 @@ jobs:
           aws s3 cp "$DMG_NAME" "s3://${SPARKLE_S3_BUCKET}/${PREFIX}${DMG_NAME}"
           if [[ "$IS_ALPHA" != "true" ]]; then
             aws s3 cp "$DMG_NAME" "s3://${SPARKLE_S3_BUCKET}/${PREFIX}EXO-latest.dmg"
+            aws s3 cp appcast.xml "s3://${SPARKLE_S3_BUCKET}/${PREFIX}appcast.xml" --content-type application/xml --cache-control no-cache
           fi
-          aws s3 cp appcast.xml "s3://${SPARKLE_S3_BUCKET}/${PREFIX}appcast.xml" --content-type application/xml --cache-control no-cache

.github/workflows/pipeline.yml

@@ -20,6 +20,12 @@ jobs:
         with:
           nix_path: nixpkgs=channel:nixos-unstable
 
+      - uses: cachix/cachix-action@v14
+        name: Configure Cachix
+        with:
+          name: exo
+          authToken: "${{ secrets.CACHIX_AUTH_TOKEN }}"
+
       - name: Configure git user
         run: |
           git config --local user.email "github-actions@users.noreply.github.com"
@@ -88,9 +94,19 @@ jobs:
 
       - uses: ./.github/actions/typecheck
 
-  nix-flake-check:
-    name: Check Nix flake
-    runs-on: ubuntu-latest
+  nix:
+    name: Build and check (${{ matrix.system }})
+    runs-on: ${{ matrix.runner }}
+    strategy:
+      fail-fast: false
+      matrix:
+        include:
+          - runner: macos-26
+            system: aarch64-darwin
+          - runner: ubuntu-latest
+            system: x86_64-linux
+          - runner: ubuntu-24.04-arm
+            system: aarch64-linux
     steps:
       - name: Checkout repository
         uses: actions/checkout@v4
@@ -101,83 +117,20 @@ jobs:
         with:
           nix_path: nixpkgs=channel:nixos-unstable
 
-      - name: Run nix flake check
-        run: |
-          nix flake check
-        shell: bash
+      - uses: cachix/cachix-action@v14
+        name: Configure Cachix
+        with:
+          name: exo
+          authToken: "${{ secrets.CACHIX_AUTH_TOKEN }}"
 
-#  ci:
-#    needs: typecheck
-#    runs-on: ubuntu-latest
-#    permissions:
-#      contents: read
-#    env:
-#      GITHUB_TOKEN: ${{ secrets.GITHUB_TOKEN }}
-#    steps:
-#      - name: Checkout repository
-#        uses: actions/checkout@v4
-#        with:
-#          fetch-depth: 0
-#          token: ${{ secrets.GITHUB_TOKEN }}
-#          lfs: true
-#
-#      - name: Configure git user
-#        run: |
-#          git config --local user.email "github-actions@users.noreply.github.com"
-#          git config --local user.name  "github-actions bot"
-#        shell: bash
-#
-#      - name: Pull LFS files
-#        run: |
-#          echo "Pulling Git LFS files..."
-#          git lfs pull
-#        shell: bash
-#
-#      - name: Setup EXO_HOME and API_PORT
-#        run: |
-#          EXO_HOME=$(mktemp -d -t exo-ci-XXXXXXXX)
-#          # Generate random port (macOS compatible method)
-#          API_PORT=$((49152 + RANDOM % (65535 - 49152 + 1)))
-#          echo "EXO_HOME=$EXO_HOME" >> $GITHUB_ENV
-#          echo "API_PORT=$API_PORT" >> $GITHUB_ENV
-#          echo "Created EXO_HOME: $EXO_HOME"
-#          echo "Generated API_PORT: $API_PORT"
-#        shell: bash
-#
-#      - name: Setup Nix Environment
-#        run: |
-#          echo "Checking for nix installation..."
-#          
-#          # Check if nix binary exists directly
-#          if [ -f /nix/var/nix/profiles/default/bin/nix ]; then
-#            echo "Found nix binary at /nix/var/nix/profiles/default/bin/nix"
-#            export PATH="/nix/var/nix/profiles/default/bin:$PATH"
-#            echo "PATH=$PATH" >> $GITHUB_ENV
-#            nix --version
-#          elif [ -f /nix/var/nix/profiles/default/etc/profile.d/nix-daemon.sh ]; then
-#            echo "Found nix profile script, sourcing..."
-#            source /nix/var/nix/profiles/default/etc/profile.d/nix-daemon.sh
-#            nix --version
-#          elif command -v nix >/dev/null 2>&1; then
-#            echo "Nix already in PATH"
-#            nix --version
-#          else
-#            echo "Nix not found. Debugging info:"
-#            echo "Contents of /nix/var/nix/profiles/default/:"
-#            ls -la /nix/var/nix/profiles/default/ 2>/dev/null || echo "Directory not found"
-#            echo "Contents of /nix/var/nix/profiles/default/bin/:"
-#            ls -la /nix/var/nix/profiles/default/bin/ 2>/dev/null || echo "Directory not found"
-#            exit 1
-#          fi
-#        shell: bash
-#
-#      - uses: ./.github/actions/lint-check
-#
-#      - uses: ./.github/actions/unit-test
-#
-#      - name: Cleanup EXO_HOME
-#        run: |
-#          echo "Cleaning up EXO_HOME: $EXO_HOME"
-#          rm -rf "$EXO_HOME"
-#        shell: bash
-#        if: always()
+      - name: Build all Nix outputs
+        run: |
+          nix flake show --json | jq -r '
+            [
+              (.packages."${{ matrix.system }}" // {} | keys[] | ".#packages.${{ matrix.system }}.\(.)"),
+              (.devShells."${{ matrix.system }}" // {} | keys[] | ".#devShells.${{ matrix.system }}.\(.)")
+            ] | .[]
+          ' | xargs nix build
+
+      - name: Run nix flake check
+        run: nix flake check

.gitignore

@@ -16,6 +16,7 @@ digest.txt
 *.xcuserdatad/
 **/.DS_Store
 app/EXO/build/
+dist/
 
 
 # rust

.mlx_typings/mlx_lm/models/deepseek_v3.pyi

@@ -0,0 +1,156 @@
+"""Type stubs for mlx_lm.models.deepseek_v3"""
+
+from dataclasses import dataclass
+from typing import Any, Dict, Optional
+
+import mlx.core as mx
+import mlx.nn as nn
+
+from .base import BaseModelArgs
+from .switch_layers import SwitchGLU
+
+@dataclass
+class ModelArgs(BaseModelArgs):
+    model_type: str
+    vocab_size: int
+    hidden_size: int
+    intermediate_size: int
+    moe_intermediate_size: int
+    num_hidden_layers: int
+    num_attention_heads: int
+    num_key_value_heads: int
+    n_shared_experts: Optional[int]
+    n_routed_experts: Optional[int]
+    routed_scaling_factor: float
+    kv_lora_rank: int
+    q_lora_rank: Optional[int]
+    qk_rope_head_dim: int
+    v_head_dim: int
+    qk_nope_head_dim: int
+    topk_method: str
+    scoring_func: str
+    norm_topk_prob: bool
+    n_group: int
+    topk_group: int
+    num_experts_per_tok: int
+    moe_layer_freq: int
+    first_k_dense_replace: int
+    max_position_embeddings: int
+    rms_norm_eps: float
+    rope_theta: float
+    rope_scaling: Optional[Dict[str, Any]]
+    attention_bias: bool
+
+class DeepseekV3Attention(nn.Module):
+    config: ModelArgs
+    hidden_size: int
+    num_heads: int
+    max_position_embeddings: int
+    rope_theta: float
+    q_lora_rank: Optional[int]
+    qk_rope_head_dim: int
+    kv_lora_rank: int
+    v_head_dim: int
+    qk_nope_head_dim: int
+    q_head_dim: int
+    scale: float
+    q_proj: nn.Linear
+    q_a_proj: nn.Linear
+    q_a_layernorm: nn.RMSNorm
+    q_b_proj: nn.Linear
+    kv_a_proj_with_mqa: nn.Linear
+    kv_a_layernorm: nn.RMSNorm
+    kv_b_proj: nn.Linear
+    o_proj: nn.Linear
+    rope: Any
+
+    def __init__(self, config: ModelArgs) -> None: ...
+    def __call__(
+        self,
+        x: mx.array,
+        mask: Optional[mx.array] = None,
+        cache: Optional[Any] = None,
+    ) -> mx.array: ...
+
+class DeepseekV3MLP(nn.Module):
+    config: ModelArgs
+    hidden_size: int
+    intermediate_size: int
+    gate_proj: nn.Linear
+    up_proj: nn.Linear
+    down_proj: nn.Linear
+
+    def __init__(
+        self,
+        config: ModelArgs,
+        hidden_size: Optional[int] = None,
+        intermediate_size: Optional[int] = None,
+    ) -> None: ...
+    def __call__(self, x: mx.array) -> mx.array: ...
+
+class MoEGate(nn.Module):
+    config: ModelArgs
+    top_k: int
+    norm_topk_prob: bool
+    n_routed_experts: Optional[int]
+    routed_scaling_factor: float
+    n_group: int
+    topk_group: int
+    weight: mx.array
+    e_score_correction_bias: mx.array
+
+    def __init__(self, config: ModelArgs) -> None: ...
+    def __call__(self, x: mx.array) -> tuple[mx.array, mx.array]: ...
+
+class DeepseekV3MoE(nn.Module):
+    config: ModelArgs
+    num_experts_per_tok: int
+    switch_mlp: SwitchGLU
+    gate: MoEGate
+    shared_experts: DeepseekV3MLP
+    sharding_group: Optional[mx.distributed.Group]
+
+    def __init__(self, config: ModelArgs) -> None: ...
+    def __call__(self, x: mx.array) -> mx.array: ...
+
+class DeepseekV3DecoderLayer(nn.Module):
+    self_attn: DeepseekV3Attention
+    mlp: DeepseekV3MLP | DeepseekV3MoE
+    input_layernorm: nn.RMSNorm
+    post_attention_layernorm: nn.RMSNorm
+
+    def __init__(self, config: ModelArgs, layer_idx: int) -> None: ...
+    def __call__(
+        self,
+        x: mx.array,
+        mask: Optional[mx.array] = None,
+        cache: Optional[Any] = None,
+    ) -> mx.array: ...
+
+class DeepseekV3Model(nn.Module):
+    vocab_size: int
+    embed_tokens: nn.Embedding
+    layers: list[DeepseekV3DecoderLayer]
+    norm: nn.RMSNorm
+
+    def __init__(self, config: ModelArgs) -> None: ...
+    def __call__(
+        self,
+        x: mx.array,
+        cache: Optional[Any] = None,
+    ) -> mx.array: ...
+
+class Model(nn.Module):
+    model_type: str
+    model: DeepseekV3Model
+    lm_head: nn.Linear
+
+    def __init__(self, config: ModelArgs) -> None: ...
+    def __call__(
+        self,
+        inputs: mx.array,
+        cache: Optional[Any] = None,
+    ) -> mx.array: ...
+    def sanitize(self, weights: dict[str, Any]) -> dict[str, Any]: ...
+    @property
+    def layers(self) -> list[DeepseekV3DecoderLayer]: ...

.mlx_typings/mlx_lm/models/switch_layers.pyi

@@ -57,6 +57,11 @@ class SwiGLU(nn.Module):
     def __call__(self, x, gate): ...
 
 class SwitchGLU(nn.Module):
+    gate_proj: SwitchLinear
+    up_proj: SwitchLinear
+    down_proj: SwitchLinear
+    activation: SwiGLU
+
     def __init__(
         self,
         input_dims: int,

.mlx_typings/mlx_lm/tokenizer_utils.pyi

@@ -4,6 +4,7 @@ This type stub file was generated by pyright.
 
 from functools import partial
 from pathlib import Path
+from typing import Any
 
 from transformers import PreTrainedTokenizerFast
 
@@ -103,37 +104,55 @@ class TokenizerWrapper:
     Accessing any attribute other than the ``detokenizer`` is forwarded to the
     huggingface tokenizer.
     """
-    def __init__(self, tokenizer, detokenizer_class=..., eos_token_ids=...) -> None: ...
-    def add_eos_token(self, token: str):  # -> None:
-        ...
-    @property
-    def has_thinking(self):  # -> bool:
-        ...
-    @property
-    def think_start(self):  # -> str | None:
-        ...
-    @property
-    def think_end(self):  # -> str | None:
-        ...
-    @property
-    def has_tool_calling(self):  # -> bool:
-        ...
-    @property
-    def tool_call_start(self):  # -> str | None:
-        ...
-    @property
-    def tool_call_end(self):  # -> str | None:
-        ...
-    @property
-    def detokenizer(self):  # -> NaiveStreamingDetokenizer:
-        """
-        Get a stateful streaming detokenizer.
-        """
 
-    def __getattr__(self, attr):  # -> set[Any] | Any:
-        ...
-    def __setattr__(self, attr, value):  # -> None:
-        ...
+    _tokenizer: PreTrainedTokenizerFast
+    eos_token_id: int | None
+    eos_token: str | None
+    bos_token_id: int | None
+    bos_token: str | None
+    vocab_size: int
+    all_special_tokens: list[str]
+
+    def __init__(
+        self,
+        tokenizer: Any,
+        detokenizer_class: Any = ...,
+        eos_token_ids: list[int] | None = ...,
+        chat_template: Any = ...,
+        tool_parser: Any = ...,
+        tool_call_start: str | None = ...,
+        tool_call_end: str | None = ...,
+    ) -> None: ...
+    def encode(self, text: str, **kwargs: Any) -> list[int]: ...
+    def decode(self, token_ids: list[int], **kwargs: Any) -> str: ...
+    def apply_chat_template(
+        self,
+        messages: list[dict[str, Any]],
+        tokenize: bool = False,
+        add_generation_prompt: bool = False,
+        tools: Any = None,
+        **kwargs: Any,
+    ) -> str: ...
+    def get_vocab(self) -> dict[str, int]: ...
+    def add_eos_token(self, token: str) -> None: ...
+    @property
+    def has_thinking(self) -> bool: ...
+    @property
+    def think_start(self) -> str | None: ...
+    @property
+    def think_end(self) -> str | None: ...
+    @property
+    def has_tool_calling(self) -> bool: ...
+    @property
+    def tool_call_start(self) -> str | None: ...
+    @property
+    def tool_call_end(self) -> str | None: ...
+    @property
+    def detokenizer(self) -> NaiveStreamingDetokenizer:
+        """Get a stateful streaming detokenizer."""
+
+    def __getattr__(self, attr: str) -> Any: ...
+    def __setattr__(self, attr: str, value: Any) -> None: ...
 
 class NewlineTokenizer(PreTrainedTokenizerFast):
     """A tokenizer that replaces newlines with <n> and <n> with new line."""
@@ -146,18 +165,11 @@ class NewlineTokenizer(PreTrainedTokenizerFast):
     def batch_decode(self, *args, **kwargs):  # -> list[str]:
         ...
 
-def load_tokenizer(
+def load(
     model_path: Path,
-    tokenizer_config_extra=...,
-    return_tokenizer=...,
-    eos_token_ids=...,
-) -> (
-    TokenizerWrapper
-    | type[SPMStreamingDetokenizer]
-    | partial[SPMStreamingDetokenizer]
-    | type[BPEStreamingDetokenizer]
-    | type[NaiveStreamingDetokenizer]
-):
+    tokenizer_config_extra: dict[str, Any] | None = None,
+    eos_token_ids: list[int] | int | None = None,
+) -> TokenizerWrapper:
     """Load a huggingface tokenizer and try to infer the type of streaming
     detokenizer to use.
 
@@ -165,4 +177,7 @@ def load_tokenizer(
     a Hugging Face repo ID.
     """
 
-def no_bos_or_eos(sequence: list, bos: int, eos: int) -> list: ...
+# Alias for backward compatibility
+load_tokenizer = load
+
+def no_bos_or_eos(sequence: list[int], bos: int, eos: int) -> list[int]: ...

.prettierrc

@@ -0,0 +1,3 @@
+{
+  "useTabs": true
+}

.swift-format

@@ -0,0 +1,6 @@
+{
+  "version": 1,
+  "indentation": {
+    "spaces": 4
+  }
+}

AGENTS.md

@@ -0,0 +1,96 @@
+# AGENTS.md
+
+This file provides guidance to AI coding agents when working with code in this repository.
+
+## Project Overview
+
+exo is a distributed AI inference system that connects multiple devices into a cluster. It enables running large language models across multiple machines using MLX as the inference backend and libp2p for peer-to-peer networking.
+
+## Build & Run Commands
+
+```bash
+# Build the dashboard (required before running exo)
+cd dashboard && npm install && npm run build && cd ..
+
+# Run exo (starts both master and worker with API at http://localhost:52415)
+uv run exo
+
+# Run with verbose logging
+uv run exo -v   # or -vv for more verbose
+
+# Run tests (excludes slow tests by default)
+uv run pytest
+
+# Run all tests including slow tests
+uv run pytest -m ""
+
+# Run a specific test file
+uv run pytest src/exo/shared/tests/test_election.py
+
+# Run a specific test function
+uv run pytest src/exo/shared/tests/test_election.py::test_function_name
+
+# Type checking (strict mode)
+uv run basedpyright
+
+# Linting
+uv run ruff check
+
+# Format code (using nix)
+nix fmt
+```
+
+## Architecture
+
+### Node Composition
+A single exo `Node` (src/exo/main.py) runs multiple components:
+- **Router**: libp2p-based pub/sub messaging via Rust bindings (exo_pyo3_bindings)
+- **Worker**: Handles inference tasks, downloads models, manages runner processes
+- **Master**: Coordinates cluster state, places model instances across nodes
+- **Election**: Bully algorithm for master election
+- **API**: FastAPI server for OpenAI-compatible chat completions
+
+### Message Flow
+Components communicate via typed pub/sub topics (src/exo/routing/topics.py):
+- `GLOBAL_EVENTS`: Master broadcasts indexed events to all workers
+- `LOCAL_EVENTS`: Workers send events to master for indexing
+- `COMMANDS`: Workers/API send commands to master
+- `ELECTION_MESSAGES`: Election protocol messages
+- `CONNECTION_MESSAGES`: libp2p connection updates
+
+### Event Sourcing
+The system uses event sourcing for state management:
+- `State` (src/exo/shared/types/state.py): Immutable state object
+- `apply()` (src/exo/shared/apply.py): Pure function that applies events to state
+- Master indexes events and broadcasts; workers apply indexed events
+
+### Key Type Hierarchy
+- `src/exo/shared/types/`: Pydantic models for all shared types
+  - `events.py`: Event types (discriminated union)
+  - `commands.py`: Command types
+  - `tasks.py`: Task types for worker execution
+  - `state.py`: Cluster state model
+
+### Rust Components
+Rust code in `rust/` provides:
+- `networking`: libp2p networking (gossipsub, peer discovery)
+- `exo_pyo3_bindings`: PyO3 bindings exposing Rust to Python
+- `system_custodian`: System-level operations
+
+### Dashboard
+Svelte 5 + TypeScript frontend in `dashboard/`. Build output goes to `dashboard/build/` and is served by the API.
+
+## Code Style Requirements
+
+From .cursorrules:
+- Strict, exhaustive typing - never bypass the type-checker
+- Use `Literal[...]` for enum-like sets, `typing.NewType` for primitives
+- Pydantic models with `frozen=True` and `strict=True`
+- Pure functions with injectable effect handlers for side-effects
+- Descriptive names - no abbreviations or 3-letter acronyms
+- Catch exceptions only where you can handle them meaningfully
+- Use `@final` and immutability wherever applicable
+
+## Testing
+
+Tests use pytest-asyncio with `asyncio_mode = "auto"`. Tests are in `tests/` subdirectories alongside the code they test. The `EXO_TESTS=1` env var is set during tests.

CLAUDE.md

@@ -0,0 +1 @@
+AGENTS.md

MISSED_THINGS.md

@@ -0,0 +1,41 @@
+# Missed things
+[X] Log EXO_LIBP2P_NAMESPACE on start in exo/main.py
+[X] Ordering of warmup was changed, which is wrong. It was changed to rank < n-1, then rank=n-1. It should be rank!=0 then rank=0 (this matches the auto_parallel implementation. NOTE: we use a different convention to mlx-lm, our terminal rank is rank=n-1 whereas mlx-lm is rank=0 hence i can see why this was changed wrongly).
+[X] Downloads keying by model_id not shard_metadata (worker/plan.py, worker/main.py).
+[X] Fetching download status of all models on start
+[X] Deduplication of tasks in plan_step.
+[X] resolve_allow_patterns should just be wildcard now.
+[] no mx_barrier in genreate.py mlx_generate at the end.
+[] cache assertion not needed in auto_parallel.py PipelineLastLayer.
+[] GPTOSS support dropped in auto_parallel.py.
+[] sharding changed "all-to-sharded" became _all_to_sharded in auto_parallel.py.
+[] same as above with "sharded-to-all" became _sharded_to_all in auto_parallel.py.
+[] Dropped support for Ministral3Model, DeepseekV32Model, Glm4MoeModel, Qwen3NextModel, GptOssMode in auto_parallel.py.
+[] Dropped prefill/decode code in auto_parallel.py and utils_mlx.py.
+[X] KV_CACHE_BITS should be None to disable quantized KV cache.
+[] Dropped _set_nofile_limit in utils_mlx.py.
+[] We have group optional in load_mlx_items in utils_mlx.py.
+[] Dropped add_missing_chat_templates for GptOss in load_mlx_items in utils_mlx.py.
+[] Dropped model.make_cache in make_kv_cache in utils_mlx.py.
+[X] We put cache limit back in utils_mlx.py.
+[] topology.py remove_node removes the connections after checking if node is is in self._node_id_to_rx_id_map. on beta_1 it checks after, so would remove stale connections I guess?
+[] Missing Glm 4.7 model cards (this isn't ready yet but should be picked up, probably create an issue... the blocker is transforemrs version doesn't support the tokenizer for Glm 4.7. rc-1 does but we can't upgrade as it breaks other things.)
+[] try-except in _command_processor only excepts ValueError. This was silently failing leading to un-debuggable errors (we had a KeyError that was happening ). Changed this to catch Exception instead of ValueError. See exo-v2 89ae38405e0052e3c22405daf094b065878aa873 and fb99fea69b5a39017efc90c5dad0072e677455f0.
+[X] In placement.py, place_instance no longer looks at model_meta.supports_tensor and check if this tensor parallel number of nodes is supported by the model's tensor dimensions.
+[X] In placement.py, place_instanec, we no longer have the special case to exclude DeepSeek v3.1 pipeline parallel (it doesn't work).
+[] logger.warning("You have likely selected ibv for a single node instance; falling back to MlxRing") was changed to debug. That will spam this warning since it happens every time we query instance previews.
+[X] In placement_utils.py, get_mlx_jaccl_coordinators, We no longer prioritise Jaccl Coordinator IP. Now it picks the first one, which is unstable (Jaccl coordinator over TB5 is unstable).
+
+
+
+[X] Downloads keying by model_id not shard_metadata (worker/plan.py, worker/main.py).
+[X] Fetching download status of all models on start
+[X] Deduplication of tasks in plan_step.
+[X] resolve_allow_patterns should just be wildcard now.
+[X] KV_CACHE_BITS should be None to disable quantized KV cache.
+[X] We put cache limit back in utils_mlx.py.
+[X] In placement.py, place_instance no longer looks at model_meta.supports_tensor and check if this tensor parallel number of nodes is supported by the model's tensor dimensions.
+[X] In placement.py, place_instanec, we no longer have the special case to exclude DeepSeek v3.1 pipeline parallel (it doesn't work).
+[X] In placement_utils.py, get_mlx_jaccl_coordinators, We no longer prioritise Jaccl Coordinator IP. Now it picks the first one, which is unstable (Jaccl coordinator over TB5 is unstable).
+
+

README.md

@@ -166,6 +166,24 @@ Download the latest build here: [EXO-latest.dmg](https://assets.exolabs.net/EXO-
 
 The app will ask for permission to modify system settings and install a new Network profile. Improvements to this are being worked on.
 
+#### Uninstalling the macOS App
+
+The recommended way to uninstall is through the app itself: click the menu bar icon → Advanced → Uninstall. This cleanly removes all system components.
+
+If you've already deleted the app, you can run the standalone uninstaller script:
+
+```bash
+sudo ./app/EXO/uninstall-exo.sh
+```
+
+This removes:
+- Network setup LaunchDaemon
+- Network configuration script
+- Log files
+- The "exo" network location
+
+**Note:** You'll need to manually remove EXO from Login Items in System Settings → General → Login Items.
+
 ---
 
 ### Enabling RDMA on macOS
@@ -287,7 +305,10 @@ curl -X DELETE http://localhost:52415/instance/YOUR_INSTANCE_ID
 - List all models: `curl http://localhost:52415/models`
 - Inspect instance IDs and deployment state: `curl http://localhost:52415/state`
 
-For further details, see API types and endpoints in [src/exo/master/api.py](src/exo/master/api.py).
+For further details, see:
+
+- API basic documentation in [docs/api.md](docs/api.md).
+- API types and endpoints in [src/exo/master/api.py](src/exo/master/api.py).
 
 ---
 

app/EXO/EXO/ContentView.swift

@@ -18,10 +18,11 @@ struct ContentView: View {
     @State private var deletingInstanceIDs: Set<String> = []
     @State private var showAllNodes = false
     @State private var showAllInstances = false
+    @State private var showAdvanced = false
     @State private var showDebugInfo = false
     @State private var bugReportInFlight = false
     @State private var bugReportMessage: String?
-    @State private var showAdvancedOptions = false
+    @State private var uninstallInProgress = false
     @State private var pendingNamespace: String = ""
 
     var body: some View {
@@ -70,10 +71,12 @@ struct ContentView: View {
                     .font(.caption)
                     .fontWeight(.semibold)
             }
-            Text("Device discovery won't work. To fix:\n1. Quit EXO\n2. Open System Settings → Privacy & Security → Local Network\n3. Toggle EXO off, then back on\n4. Relaunch EXO")
-                .font(.caption2)
-                .foregroundColor(.secondary)
-                .fixedSize(horizontal: false, vertical: true)
+            Text(
+                "Device discovery won't work. To fix:\n1. Quit EXO\n2. Open System Settings → Privacy & Security → Local Network\n3. Toggle EXO off, then back on\n4. Relaunch EXO"
+            )
+            .font(.caption2)
+            .foregroundColor(.secondary)
+            .fixedSize(horizontal: false, vertical: true)
             Button {
                 openLocalNetworkSettings()
             } label: {
@@ -96,14 +99,18 @@ struct ContentView: View {
 
     private func openLocalNetworkSettings() {
         // Open Privacy & Security settings - Local Network section
-        if let url = URL(string: "x-apple.systempreferences:com.apple.preference.security?Privacy_LocalNetwork") {
+        if let url = URL(
+            string: "x-apple.systempreferences:com.apple.preference.security?Privacy_LocalNetwork")
+        {
             NSWorkspace.shared.open(url)
         }
     }
 
     private var topologySection: some View {
         Group {
-            if let topology = stateService.latestSnapshot?.topologyViewModel(localNodeId: stateService.localNodeId), !topology.nodes.isEmpty {
+            if let topology = stateService.latestSnapshot?.topologyViewModel(
+                localNodeId: stateService.localNodeId), !topology.nodes.isEmpty
+            {
                 TopologyMiniView(topology: topology)
             }
         }
@@ -137,8 +144,10 @@ struct ContentView: View {
                 VStack(alignment: .leading, spacing: 4) {
                     HStack {
                         VStack(alignment: .leading) {
-                            Text("\(overview.usedRam, specifier: "%.0f") / \(overview.totalRam, specifier: "%.0f") GB")
-                                .font(.headline)
+                            Text(
+                                "\(overview.usedRam, specifier: "%.0f") / \(overview.totalRam, specifier: "%.0f") GB"
+                            )
+                            .font(.headline)
                             Text("Memory")
                                 .font(.caption)
                                 .foregroundColor(.secondary)
@@ -247,13 +256,7 @@ struct ContentView: View {
                 Divider()
                     .padding(.vertical, 4)
             }
-            controlButton(title: "Check for Updates") {
-                updater.checkForUpdates()
-            }
-            .padding(.bottom, 8)
-            advancedOptionsSection
-                .padding(.bottom, 8)
-            debugSection
+            advancedSection
                 .padding(.bottom, 8)
             controlButton(title: "Quit", tint: .secondary) {
                 controller.stop()
@@ -262,7 +265,57 @@ struct ContentView: View {
         }
     }
 
-    private func controlButton(title: String, tint: Color = .primary, action: @escaping () -> Void) -> some View {
+    private var advancedSection: some View {
+        VStack(alignment: .leading, spacing: 6) {
+            HStack {
+                Text("Advanced")
+                    .font(.caption)
+                    .foregroundColor(.secondary)
+                Spacer()
+                collapseButton(isExpanded: $showAdvanced)
+            }
+            .animation(nil, value: showAdvanced)
+            if showAdvanced {
+                VStack(alignment: .leading, spacing: 8) {
+                    VStack(alignment: .leading, spacing: 4) {
+                        Text("Cluster Namespace")
+                            .font(.caption2)
+                            .foregroundColor(.secondary)
+                        HStack {
+                            TextField("optional", text: $pendingNamespace)
+                                .textFieldStyle(.roundedBorder)
+                                .font(.caption2)
+                                .onAppear {
+                                    pendingNamespace = controller.customNamespace
+                                }
+                            Button("Save & Restart") {
+                                controller.customNamespace = pendingNamespace
+                                if controller.status == .running || controller.status == .starting {
+                                    controller.restart()
+                                }
+                            }
+                            .font(.caption2)
+                            .disabled(pendingNamespace == controller.customNamespace)
+                        }
+                    }
+                    HoverButton(title: "Check for Updates", small: true) {
+                        updater.checkForUpdates()
+                    }
+                    debugSection
+                    HoverButton(title: "Uninstall", tint: .red, small: true) {
+                        showUninstallConfirmationAlert()
+                    }
+                    .disabled(uninstallInProgress)
+                }
+                .transition(.opacity)
+            }
+        }
+        .animation(.easeInOut(duration: 0.25), value: showAdvanced)
+    }
+
+    private func controlButton(title: String, tint: Color = .primary, action: @escaping () -> Void)
+        -> some View
+    {
         HoverButton(title: title, tint: tint, trailingSystemImage: nil, action: action)
     }
 
@@ -293,9 +346,12 @@ struct ContentView: View {
         Button {
             isExpanded.wrappedValue.toggle()
         } label: {
-            Label(isExpanded.wrappedValue ? "Hide" : "Show All", systemImage: isExpanded.wrappedValue ? "chevron.up" : "chevron.down")
-                .labelStyle(.titleAndIcon)
-                .contentTransition(.symbolEffect(.replace))
+            Label(
+                isExpanded.wrappedValue ? "Hide" : "Show All",
+                systemImage: isExpanded.wrappedValue ? "chevron.up" : "chevron.down"
+            )
+            .labelStyle(.titleAndIcon)
+            .contentTransition(.symbolEffect(.replace))
         }
         .buttonStyle(.plain)
         .font(.caption2)
@@ -383,57 +439,16 @@ struct ContentView: View {
         }
     }
 
-    private var advancedOptionsSection: some View {
-        VStack(alignment: .leading, spacing: 6) {
-            HStack {
-                Text("Advanced Options")
-                    .font(.caption)
-                    .foregroundColor(.secondary)
-                Spacer()
-                collapseButton(isExpanded: $showAdvancedOptions)
-            }
-            .animation(nil, value: showAdvancedOptions)
-            if showAdvancedOptions {
-                VStack(alignment: .leading, spacing: 8) {
-                    VStack(alignment: .leading, spacing: 4) {
-                        Text("Cluster Namespace")
-                            .font(.caption2)
-                            .foregroundColor(.secondary)
-                        HStack {
-                            TextField("optional", text: $pendingNamespace)
-                                .textFieldStyle(.roundedBorder)
-                                .font(.caption2)
-                                .onAppear {
-                                    pendingNamespace = controller.customNamespace
-                                }
-                            Button("Save & Restart") {
-                                controller.customNamespace = pendingNamespace
-                                if controller.status == .running || controller.status == .starting {
-                                    controller.restart()
-                                }
-                            }
-                            .font(.caption2)
-                            .disabled(pendingNamespace == controller.customNamespace)
-                        }
-
-                    }
-                }
-                .transition(.opacity)
-            }
-        }
-        .animation(.easeInOut(duration: 0.25), value: showAdvancedOptions)
-    }
-
     private var debugSection: some View {
-        VStack(alignment: .leading, spacing: 6) {
-            HStack {
-                Text("Debug Info")
-                    .font(.caption)
-                    .foregroundColor(.secondary)
-                Spacer()
-                collapseButton(isExpanded: $showDebugInfo)
+        VStack(alignment: .leading, spacing: 4) {
+            HoverButton(
+                title: "Debug Info",
+                tint: .primary,
+                trailingSystemImage: showDebugInfo ? "chevron.up" : "chevron.down",
+                small: true
+            ) {
+                showDebugInfo.toggle()
             }
-            .animation(nil, value: showDebugInfo)
             if showDebugInfo {
                 VStack(alignment: .leading, spacing: 4) {
                     Text("Version: \(buildTag)")
@@ -446,15 +461,63 @@ struct ContentView: View {
                         .font(.caption2)
                         .foregroundColor(thunderboltStatusColor)
                     interfaceIpList
+                    rdmaStatusView
                     sendBugReportButton
                         .padding(.top, 6)
                 }
+                .padding(.leading, 8)
                 .transition(.opacity)
             }
         }
         .animation(.easeInOut(duration: 0.25), value: showDebugInfo)
     }
 
+    private var rdmaStatusView: some View {
+        let rdma = networkStatusService.status.rdmaStatus
+        return VStack(alignment: .leading, spacing: 1) {
+            Text("RDMA: \(rdmaStatusText(rdma))")
+                .font(.caption2)
+                .foregroundColor(rdmaStatusColor(rdma))
+            if !rdma.devices.isEmpty {
+                Text("  Devices: \(rdma.devices.joined(separator: ", "))")
+                    .font(.caption2)
+                    .foregroundColor(.secondary)
+            }
+            if !rdma.activePorts.isEmpty {
+                Text("  Active Ports:")
+                    .font(.caption2)
+                    .foregroundColor(.secondary)
+                ForEach(rdma.activePorts, id: \.device) { port in
+                    Text("    \(port.device) port \(port.port): \(port.state)")
+                        .font(.caption2)
+                        .foregroundColor(.green)
+                }
+            }
+        }
+    }
+
+    private func rdmaStatusText(_ rdma: RDMAStatus) -> String {
+        switch rdma.rdmaCtlEnabled {
+        case .some(true):
+            return "Enabled"
+        case .some(false):
+            return "Disabled"
+        case nil:
+            return rdma.devices.isEmpty ? "Not Available" : "Available"
+        }
+    }
+
+    private func rdmaStatusColor(_ rdma: RDMAStatus) -> Color {
+        switch rdma.rdmaCtlEnabled {
+        case .some(true):
+            return .green
+        case .some(false):
+            return .orange
+        case nil:
+            return rdma.devices.isEmpty ? .secondary : .green
+        }
+    }
+
     private var sendBugReportButton: some View {
         VStack(alignment: .leading, spacing: 4) {
             Button {
@@ -544,6 +607,88 @@ struct ContentView: View {
         bugReportInFlight = false
     }
 
+    private func showUninstallConfirmationAlert() {
+        let alert = NSAlert()
+        alert.messageText = "Uninstall EXO"
+        alert.informativeText = """
+            This will remove EXO and all its system components:
+
+            • Network configuration daemon
+            • Launch at login registration
+            • EXO network location
+
+            The app will be moved to Trash.
+            """
+        alert.alertStyle = .warning
+        alert.addButton(withTitle: "Uninstall")
+        alert.addButton(withTitle: "Cancel")
+
+        // Style the Uninstall button as destructive
+        if let uninstallButton = alert.buttons.first {
+            uninstallButton.hasDestructiveAction = true
+        }
+
+        let response = alert.runModal()
+        if response == .alertFirstButtonReturn {
+            performUninstall()
+        }
+    }
+
+    private func performUninstall() {
+        uninstallInProgress = true
+
+        // Stop EXO process first
+        controller.cancelPendingLaunch()
+        controller.stop()
+        stateService.stopPolling()
+
+        // Run the privileged uninstall on a background thread
+        // Using .utility QoS to avoid priority inversion with NSAppleScript's subprocess
+        DispatchQueue.global(qos: .utility).async {
+            do {
+                // Remove network setup daemon and components (requires admin privileges)
+                try NetworkSetupHelper.uninstall()
+
+                DispatchQueue.main.async {
+                    // Unregister from launch at login
+                    LaunchAtLoginHelper.disable()
+
+                    // Move app to trash
+                    self.moveAppToTrash()
+
+                    // Quit the app
+                    DispatchQueue.main.asyncAfter(deadline: .now() + 0.5) {
+                        NSApplication.shared.terminate(nil)
+                    }
+                }
+            } catch {
+                DispatchQueue.main.async {
+                    self.showErrorAlert(message: error.localizedDescription)
+                    self.uninstallInProgress = false
+                }
+            }
+        }
+    }
+
+    private func showErrorAlert(message: String) {
+        let alert = NSAlert()
+        alert.messageText = "Uninstall Failed"
+        alert.informativeText = message
+        alert.alertStyle = .critical
+        alert.addButton(withTitle: "OK")
+        alert.runModal()
+    }
+
+    private func moveAppToTrash() {
+        guard let appURL = Bundle.main.bundleURL as URL? else { return }
+        do {
+            try FileManager.default.trashItem(at: appURL, resultingItemURL: nil)
+        } catch {
+            // If we can't trash the app, that's OK - user can do it manually
+            // The important system components have already been cleaned up
+        }
+    }
+
     private var buildTag: String {
         Bundle.main.infoDictionary?["EXOBuildTag"] as? String ?? "unknown"
     }
@@ -557,14 +702,27 @@ private struct HoverButton: View {
     let title: String
     let tint: Color
     let trailingSystemImage: String?
+    let small: Bool
     let action: () -> Void
 
+    init(
+        title: String, tint: Color = .primary, trailingSystemImage: String? = nil,
+        small: Bool = false, action: @escaping () -> Void
+    ) {
+        self.title = title
+        self.tint = tint
+        self.trailingSystemImage = trailingSystemImage
+        self.small = small
+        self.action = action
+    }
+
     @State private var isHovering = false
 
     var body: some View {
         Button(action: action) {
             HStack {
                 Text(title)
+                    .font(small ? .caption : nil)
                 Spacer()
                 if let systemName = trailingSystemImage {
                     Image(systemName: systemName)
@@ -572,8 +730,8 @@ private struct HoverButton: View {
                 }
             }
             .frame(maxWidth: .infinity, alignment: .leading)
-            .padding(.vertical, 6)
-            .padding(.horizontal, 8)
+            .padding(.vertical, small ? 4 : 6)
+            .padding(.horizontal, small ? 6 : 8)
             .background(
                 RoundedRectangle(cornerRadius: 6)
                     .fill(
@@ -588,4 +746,3 @@ private struct HoverButton: View {
         .onHover { isHovering = $0 }
     }
 }
-

app/EXO/EXO/EXOApp.swift

@@ -8,9 +8,9 @@
 import AppKit
 import CoreImage
 import CoreImage.CIFilterBuiltins
+import ServiceManagement
 import Sparkle
 import SwiftUI
-import ServiceManagement
 import UserNotifications
 import os.log
 
@@ -113,7 +113,7 @@ struct EXOApp: App {
         filter.contrast = 0.9
 
         guard let output = filter.outputImage,
-              let rendered = ciContext.createCGImage(output, from: output.extent)
+            let rendered = ciContext.createCGImage(output, from: output.extent)
         else {
             return nil
         }
@@ -126,7 +126,26 @@ struct EXOApp: App {
         do {
             try SMAppService.mainApp.register()
         } catch {
-            Logger().error("Failed to register EXO for launch at login: \(error.localizedDescription)")
+            Logger().error(
+                "Failed to register EXO for launch at login: \(error.localizedDescription)")
+        }
+    }
+}
+
+/// Helper for managing EXO's launch-at-login registration
+enum LaunchAtLoginHelper {
+    private static let logger = Logger(subsystem: "io.exo.EXO", category: "LaunchAtLogin")
+
+    /// Unregisters EXO from launching at login
+    static func disable() {
+        guard SMAppService.mainApp.status == .enabled else { return }
+        do {
+            try SMAppService.mainApp.unregister()
+            logger.info("Unregistered EXO from launch at login")
+        } catch {
+            logger.error(
+                "Failed to unregister EXO from launch at login: \(error.localizedDescription, privacy: .public)"
+            )
         }
     }
 }
@@ -151,7 +170,7 @@ final class SparkleUpdater: NSObject, ObservableObject {
         center.requestAuthorization(options: [.alert, .sound]) { _, _ in }
         controller.updater.automaticallyChecksForUpdates = true
         controller.updater.automaticallyDownloadsUpdates = false
-        controller.updater.updateCheckInterval = 900 // 15 minutes
+        controller.updater.updateCheckInterval = 900  // 15 minutes
         DispatchQueue.main.asyncAfter(deadline: .now() + 5) { [weak controller] in
             controller?.updater.checkForUpdatesInBackground()
         }
@@ -218,7 +237,8 @@ private final class ExoNotificationDelegate: NSObject, UNUserNotificationCenterD
     func userNotificationCenter(
         _ center: UNUserNotificationCenter,
         willPresent notification: UNNotification,
-        withCompletionHandler completionHandler: @escaping (UNNotificationPresentationOptions) -> Void
+        withCompletionHandler completionHandler: @escaping (UNNotificationPresentationOptions) ->
+            Void
     ) {
         completionHandler([.banner, .list, .sound])
     }

app/EXO/EXO/ExoProcessController.swift

@@ -31,7 +31,8 @@ final class ExoProcessController: ObservableObject {
     @Published private(set) var launchCountdownSeconds: Int?
     @Published var customNamespace: String = {
         return UserDefaults.standard.string(forKey: customNamespaceKey) ?? ""
-    }() {
+    }()
+    {
         didSet {
             UserDefaults.standard.set(customNamespace, forKey: customNamespaceKey)
         }
@@ -221,7 +222,9 @@ final class ExoProcessController: ObservableObject {
         if let tag = Bundle.main.infoDictionary?["EXOBuildTag"] as? String, !tag.isEmpty {
             return tag
         }
-        if let short = Bundle.main.infoDictionary?["CFBundleShortVersionString"] as? String, !short.isEmpty {
+        if let short = Bundle.main.infoDictionary?["CFBundleShortVersionString"] as? String,
+            !short.isEmpty
+        {
             return short
         }
         return "dev"

app/EXO/EXO/Info.plist

@@ -8,6 +8,8 @@
 	<string>$(EXO_BUILD_TAG)</string>
 	<key>EXOBuildCommit</key>
 	<string>$(EXO_BUILD_COMMIT)</string>
+	<key>EXOBugReportPresignedUrlEndpoint</key>
+	<string>$(EXO_BUG_REPORT_PRESIGNED_URL_ENDPOINT)</string>
 	<key>NSLocalNetworkUsageDescription</key>
 	<string>EXO needs local network access to discover and connect to other devices in your cluster for distributed AI inference.</string>
 	<key>NSBonjourServices</key>

app/EXO/EXO/Models/ClusterState.swift

@@ -16,10 +16,13 @@ struct ClusterState: Decodable {
         self.instances = rawInstances.mapValues(\.instance)
         self.runners = try container.decode([String: RunnerStatusSummary].self, forKey: .runners)
         self.nodeProfiles = try container.decode([String: NodeProfile].self, forKey: .nodeProfiles)
-        let rawTasks = try container.decodeIfPresent([String: TaggedTask].self, forKey: .tasks) ?? [:]
+        let rawTasks =
+            try container.decodeIfPresent([String: TaggedTask].self, forKey: .tasks) ?? [:]
         self.tasks = rawTasks.compactMapValues(\.task)
         self.topology = try container.decodeIfPresent(Topology.self, forKey: .topology)
-        let rawDownloads = try container.decodeIfPresent([String: [TaggedNodeDownload]].self, forKey: .downloads) ?? [:]
+        let rawDownloads =
+            try container.decodeIfPresent([String: [TaggedNodeDownload]].self, forKey: .downloads)
+            ?? [:]
         self.downloads = rawDownloads.mapValues { $0.compactMap(\.status) }
     }
 
@@ -41,7 +44,8 @@ private struct TaggedInstance: Decodable {
         let payloads = try container.decode([String: ClusterInstancePayload].self)
         guard let entry = payloads.first else {
             throw DecodingError.dataCorrupted(
-                DecodingError.Context(codingPath: decoder.codingPath, debugDescription: "Empty instance payload")
+                DecodingError.Context(
+                    codingPath: decoder.codingPath, debugDescription: "Empty instance payload")
             )
         }
         self.instance = ClusterInstance(
@@ -77,7 +81,8 @@ struct RunnerStatusSummary: Decodable {
         let payloads = try container.decode([String: RunnerStatusDetail].self)
         guard let entry = payloads.first else {
             throw DecodingError.dataCorrupted(
-                DecodingError.Context(codingPath: decoder.codingPath, debugDescription: "Empty runner status payload")
+                DecodingError.Context(
+                    codingPath: decoder.codingPath, debugDescription: "Empty runner status payload")
             )
         }
         self.status = entry.key
@@ -257,7 +262,9 @@ struct ChatCompletionTaskParameters: Decodable, Equatable {
 
     func promptPreview() -> String? {
         guard let messages else { return nil }
-        if let userMessage = messages.last(where: { $0.role?.lowercased() == "user" && ($0.content?.isEmpty == false) }) {
+        if let userMessage = messages.last(where: {
+            $0.role?.lowercased() == "user" && ($0.content?.isEmpty == false)
+        }) {
             return userMessage.content
         }
         return messages.last?.content
@@ -365,5 +372,3 @@ extension ClusterState {
 
     func availableModels() -> [ModelOption] { [] }
 }
-
-

app/EXO/EXO/Services/BugReportService.swift

@@ -1,4 +1,3 @@
-import CryptoKit
 import Foundation
 
 struct BugReportOutcome: Equatable {
@@ -7,17 +6,17 @@ struct BugReportOutcome: Equatable {
 }
 
 enum BugReportError: LocalizedError {
-    case missingCredentials
     case invalidEndpoint
+    case presignedUrlFailed(String)
     case uploadFailed(String)
     case collectFailed(String)
 
     var errorDescription: String? {
         switch self {
-        case .missingCredentials:
-            return "Bug report upload credentials are not set."
         case .invalidEndpoint:
             return "Bug report endpoint is invalid."
+        case .presignedUrlFailed(let message):
+            return "Failed to get presigned URLs: \(message)"
         case .uploadFailed(let message):
             return "Bug report upload failed: \(message)"
         case .collectFailed(let message):
@@ -27,11 +26,13 @@ enum BugReportError: LocalizedError {
 }
 
 struct BugReportService {
-    struct AWSConfig {
-        let accessKey: String
-        let secretKey: String
-        let region: String
-        let bucket: String
+    private struct PresignedUrlsRequest: Codable {
+        let keys: [String]
+    }
+
+    private struct PresignedUrlsResponse: Codable {
+        let urls: [String: String]
+        let expiresIn: Int?
     }
 
     func sendReport(
@@ -39,9 +40,9 @@ struct BugReportService {
         now: Date = Date(),
         isManual: Bool = false
     ) async throws -> BugReportOutcome {
-        let credentials = try loadCredentials()
-        let timestamp = ISO8601DateFormatter().string(from: now)
-        let prefix = "reports/\(timestamp)/"
+        let timestamp = Self.runTimestampString(now)
+        let dayPrefix = Self.dayPrefixString(now)
+        let prefix = "reports/\(dayPrefix)/\(timestamp)/"
 
         let logData = readLog()
         let ifconfigText = try await captureIfconfig()
@@ -66,28 +67,82 @@ struct BugReportService {
             ("\(prefix)exo.log", logData),
             ("\(prefix)state.json", stateData),
             ("\(prefix)events.json", eventsData),
-            ("\(prefix)report.json", reportJSON)
+            ("\(prefix)report.json", reportJSON),
         ]
 
-        let uploader = try S3Uploader(config: credentials)
-        for item in uploads {
-            guard let data = item.data else { continue }
-            try await uploader.upload(
-                objectPath: item.path,
-                body: data
-            )
+        let uploadItems: [(key: String, body: Data)] = uploads.compactMap { item in
+            guard let body = item.data else { return nil }
+            return (key: item.path, body: body)
         }
 
-        return BugReportOutcome(success: true, message: "Bug Report sent. Thank you for helping to improve EXO 1.0.")
+        guard !uploadItems.isEmpty else {
+            return BugReportOutcome(success: false, message: "No data to upload")
+        }
+
+        let presignedUrls = try await fetchPresignedUploadUrls(keys: uploadItems.map(\.key))
+        for item in uploadItems {
+            guard let urlString = presignedUrls[item.key], let url = URL(string: urlString) else {
+                throw BugReportError.uploadFailed("Missing presigned URL for \(item.key)")
+            }
+            try await uploadToPresignedUrl(url: url, body: item.body)
+        }
+
+        return BugReportOutcome(
+            success: true, message: "Bug Report sent. Thank you for helping to improve EXO 1.0.")
     }
 
-    private func loadCredentials() throws -> AWSConfig {
-        return AWSConfig(
-            accessKey: "AKIAYEKP5EMXTOBYDGHX",
-            secretKey: "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",
-            region: "us-east-1",
-            bucket: "exo-bug-reports"
-        )
+    private static func dayPrefixString(_ date: Date) -> String {
+        var calendar = Calendar(identifier: .gregorian)
+        calendar.timeZone = TimeZone(secondsFromGMT: 0) ?? .current
+        let components = calendar.dateComponents([.year, .month, .day], from: date)
+        let year = components.year ?? 0
+        let month = components.month ?? 0
+        let day = components.day ?? 0
+        return String(format: "%04d/%02d/%02d", year, month, day)
+    }
+
+    private static func runTimestampString(_ date: Date) -> String {
+        let formatter = DateFormatter()
+        formatter.locale = Locale(identifier: "en_US_POSIX")
+        formatter.timeZone = TimeZone(secondsFromGMT: 0) ?? .current
+        formatter.dateFormat = "yyyy-MM-dd'T'HHmmss.SSS'Z'"
+        return formatter.string(from: date)
+    }
+
+    private func fetchPresignedUploadUrls(keys: [String], bundle: Bundle = .main) async throws
+        -> [String: String]
+    {
+        guard
+            let endpointString = bundle.infoDictionary?["EXOBugReportPresignedUrlEndpoint"]
+                as? String
+        else {
+            throw BugReportError.invalidEndpoint
+        }
+        let trimmedEndpointString = endpointString.trimmingCharacters(in: .whitespacesAndNewlines)
+        guard !trimmedEndpointString.isEmpty, let endpoint = URL(string: trimmedEndpointString)
+        else {
+            throw BugReportError.invalidEndpoint
+        }
+
+        var request = URLRequest(url: endpoint)
+        request.httpMethod = "POST"
+        request.timeoutInterval = 10
+        request.setValue("application/json", forHTTPHeaderField: "Content-Type")
+
+        let encoder = JSONEncoder()
+        request.httpBody = try encoder.encode(PresignedUrlsRequest(keys: keys))
+
+        let (data, response) = try await URLSession.shared.data(for: request)
+        guard let http = response as? HTTPURLResponse else {
+            throw BugReportError.presignedUrlFailed("Non-HTTP response")
+        }
+        guard (200..<300).contains(http.statusCode) else {
+            throw BugReportError.presignedUrlFailed("HTTP status \(http.statusCode)")
+        }
+
+        let decoder = JSONDecoder()
+        let decoded = try decoder.decode(PresignedUrlsResponse.self, from: data)
+        return decoded.urls
     }
 
     private func readLog() -> Data? {
@@ -100,7 +155,8 @@ struct BugReportService {
     private func captureIfconfig() async throws -> String {
         let result = runCommand(["/sbin/ifconfig"])
         guard result.exitCode == 0 else {
-            throw BugReportError.collectFailed(result.error.isEmpty ? "ifconfig failed" : result.error)
+            throw BugReportError.collectFailed(
+                result.error.isEmpty ? "ifconfig failed" : result.error)
         }
         return result.output
     }
@@ -108,12 +164,23 @@ struct BugReportService {
     private func readDebugInfo() -> DebugInfo {
         DebugInfo(
             thunderboltBridgeDisabled: readThunderboltBridgeDisabled(),
-            interfaces: readInterfaces()
+            interfaces: readInterfaces(),
+            rdma: readRDMADebugInfo()
+        )
+    }
+
+    private func readRDMADebugInfo() -> DebugInfo.RDMADebugInfo {
+        DebugInfo.RDMADebugInfo(
+            rdmaCtlStatus: safeRunCommand(["/usr/bin/rdma_ctl", "status"]),
+            ibvDevices: safeRunCommand(["/usr/bin/ibv_devices"]),
+            ibvDevinfo: safeRunCommand(["/usr/bin/ibv_devinfo"])
         )
     }
 
     private func readThunderboltBridgeDisabled() -> Bool? {
-        let result = runCommand(["/usr/sbin/networksetup", "-getnetworkserviceenabled", "Thunderbolt Bridge"])
+        let result = runCommand([
+            "/usr/sbin/networksetup", "-getnetworkserviceenabled", "Thunderbolt Bridge",
+        ])
         guard result.exitCode == 0 else { return nil }
         let output = result.output.lowercased()
         if output.contains("enabled") {
@@ -156,7 +223,8 @@ struct BugReportService {
         request.timeoutInterval = 5
         do {
             let (data, response) = try await URLSession.shared.data(for: request)
-            guard let http = response as? HTTPURLResponse, (200..<300).contains(http.statusCode) else {
+            guard let http = response as? HTTPURLResponse, (200..<300).contains(http.statusCode)
+            else {
                 return nil
             }
             return data
@@ -165,6 +233,36 @@ struct BugReportService {
         }
     }
 
+    private func uploadToPresignedUrl(url: URL, body: Data) async throws {
+        let maxAttempts = 2
+        var lastError: Error?
+
+        for attempt in 1...maxAttempts {
+            do {
+                var request = URLRequest(url: url)
+                request.httpMethod = "PUT"
+                request.httpBody = body
+                request.timeoutInterval = 30
+
+                let (_, response) = try await URLSession.shared.data(for: request)
+                guard let http = response as? HTTPURLResponse else {
+                    throw BugReportError.uploadFailed("Non-HTTP response")
+                }
+                guard (200..<300).contains(http.statusCode) else {
+                    throw BugReportError.uploadFailed("HTTP status \(http.statusCode)")
+                }
+                return
+            } catch {
+                lastError = error
+                if attempt < maxAttempts {
+                    try await Task.sleep(nanoseconds: 400_000_000)
+                }
+            }
+        }
+
+        throw BugReportError.uploadFailed(lastError?.localizedDescription ?? "Unknown error")
+    }
+
     private func makeReportJson(
         timestamp: String,
         hostName: String,
@@ -182,7 +280,7 @@ struct BugReportService {
             "system": system,
             "exo_version": exo.version as Any,
             "exo_commit": exo.commit as Any,
-            "report_type": isManual ? "manual" : "automated"
+            "report_type": isManual ? "manual" : "automated",
         ]
         return try? JSONSerialization.data(withJSONObject: payload, options: [.prettyPrinted])
     }
@@ -213,10 +311,13 @@ struct BugReportService {
         let user = safeRunCommand(["/usr/bin/whoami"])
         let consoleUser = safeRunCommand(["/usr/bin/stat", "-f%Su", "/dev/console"])
         let uptime = safeRunCommand(["/usr/bin/uptime"])
-        let diskRoot = safeRunCommand(["/bin/sh", "-c", "/bin/df -h / | awk 'NR==2 {print $1, $2, $3, $4, $5}'"])
+        let diskRoot = safeRunCommand([
+            "/bin/sh", "-c", "/bin/df -h / | awk 'NR==2 {print $1, $2, $3, $4, $5}'",
+        ])
 
         let interfacesList = safeRunCommand(["/usr/sbin/ipconfig", "getiflist"])
-        let interfacesAndIPs = interfacesList?
+        let interfacesAndIPs =
+            interfacesList?
             .split(whereSeparator: { $0 == " " || $0 == "\n" })
             .compactMap { iface -> [String: Any]? in
                 let name = String(iface)
@@ -227,7 +328,8 @@ struct BugReportService {
             } ?? []
 
         let wifiSSID: String?
-        let airportPath = "/System/Library/PrivateFrameworks/Apple80211.framework/Versions/Current/Resources/airport"
+        let airportPath =
+            "/System/Library/PrivateFrameworks/Apple80211.framework/Versions/Current/Resources/airport"
         if FileManager.default.isExecutableFile(atPath: airportPath) {
             wifiSSID = safeRunCommand([airportPath, "-I"]).flatMap(parseWifiSSID)
         } else {
@@ -255,7 +357,7 @@ struct BugReportService {
             "disk_root": diskRoot as Any,
             "interfaces_and_ips": interfacesAndIPs,
             "ipconfig_getiflist": interfacesList as Any,
-            "wifi_ssid": wifiSSID as Any
+            "wifi_ssid": wifiSSID as Any,
         ]
     }
 
@@ -313,7 +415,8 @@ struct BugReportService {
         for line in airportOutput.split(separator: "\n") {
             let trimmed = line.trimmingCharacters(in: .whitespaces)
             if trimmed.hasPrefix("SSID:") {
-                return trimmed.replacingOccurrences(of: "SSID:", with: "").trimmingCharacters(in: .whitespaces)
+                return trimmed.replacingOccurrences(of: "SSID:", with: "").trimmingCharacters(
+                    in: .whitespaces)
             }
         }
         return nil
@@ -350,6 +453,7 @@ struct BugReportService {
 private struct DebugInfo {
     let thunderboltBridgeDisabled: Bool?
     let interfaces: [InterfaceStatus]
+    let rdma: RDMADebugInfo
 
     struct InterfaceStatus {
         let name: String
@@ -358,7 +462,21 @@ private struct DebugInfo {
         func toDictionary() -> [String: Any] {
             [
                 "name": name,
-                "ip": ip as Any
+                "ip": ip as Any,
+            ]
+        }
+    }
+
+    struct RDMADebugInfo {
+        let rdmaCtlStatus: String?
+        let ibvDevices: String?
+        let ibvDevinfo: String?
+
+        func toDictionary() -> [String: Any] {
+            [
+                "rdma_ctl_status": rdmaCtlStatus as Any,
+                "ibv_devices": ibvDevices as Any,
+                "ibv_devinfo": ibvDevinfo as Any,
             ]
         }
     }
@@ -366,7 +484,8 @@ private struct DebugInfo {
     func toDictionary() -> [String: Any] {
         [
             "thunderbolt_bridge_disabled": thunderboltBridgeDisabled as Any,
-            "interfaces": interfaces.map { $0.toDictionary() }
+            "interfaces": interfaces.map { $0.toDictionary() },
+            "rdma": rdma.toDictionary(),
         ]
     }
 }
@@ -376,163 +495,3 @@ private struct CommandResult {
     let output: String
     let error: String
 }
-
-private struct S3Uploader {
-    let config: BugReportService.AWSConfig
-
-    init(config: BugReportService.AWSConfig) throws {
-        self.config = config
-    }
-
-    func upload(objectPath: String, body: Data) async throws {
-        let host = "\(config.bucket).s3.amazonaws.com"
-        guard let url = URL(string: "https://\(host)/\(objectPath)") else {
-            throw BugReportError.invalidEndpoint
-        }
-
-        let now = Date()
-        let amzDate = awsTimestamp(now)
-        let dateStamp = dateStamp(now)
-        let payloadHash = sha256Hex(body)
-
-        let headers = [
-            "host": host,
-            "x-amz-content-sha256": payloadHash,
-            "x-amz-date": amzDate
-        ]
-
-        let canonicalRequest = buildCanonicalRequest(
-            method: "PUT",
-            url: url,
-            headers: headers,
-            payloadHash: payloadHash
-        )
-
-        let stringToSign = buildStringToSign(
-            amzDate: amzDate,
-            dateStamp: dateStamp,
-            canonicalRequestHash: sha256Hex(canonicalRequest.data(using: .utf8) ?? Data())
-        )
-
-        let signingKey = deriveKey(secret: config.secretKey, dateStamp: dateStamp, region: config.region, service: "s3")
-        let signature = hmacHex(key: signingKey, data: Data(stringToSign.utf8))
-
-        let signedHeaders = "host;x-amz-content-sha256;x-amz-date"
-        let authorization = """
-AWS4-HMAC-SHA256 Credential=\(config.accessKey)/\(dateStamp)/\(config.region)/s3/aws4_request, SignedHeaders=\(signedHeaders), Signature=\(signature)
-"""
-
-        var request = URLRequest(url: url)
-        request.httpMethod = "PUT"
-        request.httpBody = body
-        request.setValue(headers["x-amz-content-sha256"], forHTTPHeaderField: "x-amz-content-sha256")
-        request.setValue(headers["x-amz-date"], forHTTPHeaderField: "x-amz-date")
-        request.setValue(host, forHTTPHeaderField: "Host")
-        request.setValue(authorization, forHTTPHeaderField: "Authorization")
-
-        let (data, response) = try await URLSession.shared.data(for: request)
-        guard let http = response as? HTTPURLResponse, (200..<300).contains(http.statusCode) else {
-            let statusText = (response as? HTTPURLResponse)?.statusCode ?? -1
-            _ = data // ignore response body for UX
-            throw BugReportError.uploadFailed("HTTP status \(statusText)")
-        }
-    }
-
-    private func buildCanonicalRequest(
-        method: String,
-        url: URL,
-        headers: [String: String],
-        payloadHash: String
-    ) -> String {
-        let canonicalURI = encodePath(url.path)
-        let canonicalQuery = url.query ?? ""
-        let sortedHeaders = headers.sorted { $0.key < $1.key }
-        let canonicalHeaders = sortedHeaders
-            .map { "\($0.key.lowercased()):\($0.value)\n" }
-            .joined()
-        let signedHeaders = sortedHeaders.map { $0.key.lowercased() }.joined(separator: ";")
-
-        return [
-            method,
-            canonicalURI,
-            canonicalQuery,
-            canonicalHeaders,
-            signedHeaders,
-            payloadHash
-        ].joined(separator: "\n")
-    }
-
-    private func encodePath(_ path: String) -> String {
-        return path
-            .split(separator: "/")
-            .map { segment in
-                segment.addingPercentEncoding(withAllowedCharacters: Self.rfc3986) ?? String(segment)
-            }
-            .joined(separator: "/")
-            .prependSlashIfNeeded()
-    }
-
-    private func buildStringToSign(
-        amzDate: String,
-        dateStamp: String,
-        canonicalRequestHash: String
-    ) -> String {
-        """
-AWS4-HMAC-SHA256
-\(amzDate)
-\(dateStamp)/\(config.region)/s3/aws4_request
-\(canonicalRequestHash)
-"""
-    }
-
-    private func deriveKey(secret: String, dateStamp: String, region: String, service: String) -> Data {
-        let kDate = hmac(key: Data(("AWS4" + secret).utf8), data: Data(dateStamp.utf8))
-        let kRegion = hmac(key: kDate, data: Data(region.utf8))
-        let kService = hmac(key: kRegion, data: Data(service.utf8))
-        return hmac(key: kService, data: Data("aws4_request".utf8))
-    }
-
-    private func hmac(key: Data, data: Data) -> Data {
-        let keySym = SymmetricKey(data: key)
-        let mac = HMAC<SHA256>.authenticationCode(for: data, using: keySym)
-        return Data(mac)
-    }
-
-    private func hmacHex(key: Data, data: Data) -> String {
-        hmac(key: key, data: data).map { String(format: "%02x", $0) }.joined()
-    }
-
-    private func sha256Hex(_ data: Data) -> String {
-        let digest = SHA256.hash(data: data)
-        return digest.compactMap { String(format: "%02x", $0) }.joined()
-    }
-
-    private func awsTimestamp(_ date: Date) -> String {
-        let formatter = DateFormatter()
-        formatter.dateFormat = "yyyyMMdd'T'HHmmss'Z'"
-        formatter.timeZone = TimeZone(abbreviation: "UTC")
-        return formatter.string(from: date)
-    }
-
-    private func dateStamp(_ date: Date) -> String {
-        let formatter = DateFormatter()
-        formatter.dateFormat = "yyyyMMdd"
-        formatter.timeZone = TimeZone(abbreviation: "UTC")
-        return formatter.string(from: date)
-    }
-
-    private static let rfc3986: CharacterSet = {
-        var set = CharacterSet.alphanumerics
-        set.insert(charactersIn: "-._~")
-        return set
-    }()
-}
-
-private extension String {
-    func prependSlashIfNeeded() -> String {
-        if hasPrefix("/") {
-            return self
-        }
-        return "/" + self
-    }
-}

app/EXO/EXO/Services/ClusterStateService.swift

@@ -57,7 +57,9 @@ final class ClusterStateService: ObservableObject {
             var request = URLRequest(url: url)
             request.cachePolicy = .reloadIgnoringLocalCacheData
             let (data, response) = try await session.data(for: request)
-            guard let httpResponse = response as? HTTPURLResponse, (200..<300).contains(httpResponse.statusCode) else {
+            guard let httpResponse = response as? HTTPURLResponse,
+                (200..<300).contains(httpResponse.statusCode)
+            else {
                 return
             }
             if let nodeId = try? decoder.decode(String.self, from: data) {
@@ -113,7 +115,9 @@ final class ClusterStateService: ObservableObject {
         }
     }
 
-    func launchInstance(modelId: String, sharding: String, instanceMeta: String, minNodes: Int) async {
+    func launchInstance(modelId: String, sharding: String, instanceMeta: String, minNodes: Int)
+        async
+    {
         do {
             var request = URLRequest(url: baseURL.appendingPathComponent("instance"))
             request.httpMethod = "POST"
@@ -122,7 +126,7 @@ final class ClusterStateService: ObservableObject {
                 "model_id": modelId,
                 "sharding": sharding,
                 "instance_meta": instanceMeta,
-                "min_nodes": minNodes
+                "min_nodes": minNodes,
             ]
             request.httpBody = try JSONSerialization.data(withJSONObject: payload, options: [])
             let (_, response) = try await session.data(for: request)
@@ -143,7 +147,9 @@ final class ClusterStateService: ObservableObject {
         do {
             let url = baseURL.appendingPathComponent("models")
             let (data, response) = try await session.data(from: url)
-            guard let httpResponse = response as? HTTPURLResponse, (200..<300).contains(httpResponse.statusCode) else {
+            guard let httpResponse = response as? HTTPURLResponse,
+                (200..<300).contains(httpResponse.statusCode)
+            else {
                 throw URLError(.badServerResponse)
             }
             let list = try decoder.decode(ModelListResponse.self, from: data)

app/EXO/EXO/Services/LocalNetworkChecker.swift

@@ -110,8 +110,9 @@ final class LocalNetworkChecker: ObservableObject {
                     }
                 case .failed(let error):
                     let errorStr = "\(error)"
-                    if errorStr.contains("65") || errorStr.contains("EHOSTUNREACH") ||
-                       errorStr.contains("permission") || errorStr.contains("denied") {
+                    if errorStr.contains("65") || errorStr.contains("EHOSTUNREACH")
+                        || errorStr.contains("permission") || errorStr.contains("denied")
+                    {
                         resumeOnce(.notWorking(reason: "Permission denied"))
                     } else {
                         resumeOnce(.notWorking(reason: "Failed: \(error.localizedDescription)"))

app/EXO/EXO/Services/NetworkSetupHelper.swift

@@ -5,64 +5,66 @@ import os.log
 enum NetworkSetupHelper {
     private static let logger = Logger(subsystem: "io.exo.EXO", category: "NetworkSetup")
     private static let daemonLabel = "io.exo.networksetup"
-    private static let scriptDestination = "/Library/Application Support/EXO/disable_bridge_enable_dhcp.sh"
+    private static let scriptDestination =
+        "/Library/Application Support/EXO/disable_bridge_enable_dhcp.sh"
     private static let plistDestination = "/Library/LaunchDaemons/io.exo.networksetup.plist"
     private static let requiredStartInterval: Int = 1791
 
     private static let setupScript = """
-#!/usr/bin/env bash
+        #!/usr/bin/env bash
 
-set -euo pipefail
+        set -euo pipefail
 
-PREFS="/Library/Preferences/SystemConfiguration/preferences.plist"
+        PREFS="/Library/Preferences/SystemConfiguration/preferences.plist"
 
-# Remove bridge0 interface
-ifconfig bridge0 &>/dev/null && {
-  ifconfig bridge0 | grep -q 'member' && {
-    ifconfig bridge0 | awk '/member/ {print $2}' | xargs -n1 ifconfig bridge0 deletem 2>/dev/null || true
-  }
-  ifconfig bridge0 destroy 2>/dev/null || true
-}
+        # Remove bridge0 interface
+        ifconfig bridge0 &>/dev/null && {
+          ifconfig bridge0 | grep -q 'member' && {
+            ifconfig bridge0 | awk '/member/ {print $2}' | xargs -n1 ifconfig bridge0 deletem 2>/dev/null || true
+          }
+          ifconfig bridge0 destroy 2>/dev/null || true
+        }
 
-# Remove Thunderbolt Bridge from VirtualNetworkInterfaces in preferences.plist
-/usr/libexec/PlistBuddy -c "Delete :VirtualNetworkInterfaces:Bridge:bridge0" "$PREFS" 2>/dev/null || true
+        # Remove Thunderbolt Bridge from VirtualNetworkInterfaces in preferences.plist
+        /usr/libexec/PlistBuddy -c "Delete :VirtualNetworkInterfaces:Bridge:bridge0" "$PREFS" 2>/dev/null || true
 
-networksetup -listlocations | grep -q exo || {
-  networksetup -createlocation exo
-}
+        networksetup -listlocations | grep -q exo || {
+          networksetup -createlocation exo
+        }
 
-networksetup -switchtolocation exo
-networksetup -listallhardwareports \\
-  | awk -F': ' '/Hardware Port: / {print $2}' \\
-  | while IFS=":" read -r name; do
-      case "$name" in
-        "Ethernet Adapter"*)
-                ;;
-        "Thunderbolt Bridge")
-                ;;
-        "Thunderbolt "*)
-          networksetup -listallnetworkservices \\
-            | grep -q "EXO $name" \\
-              || networksetup -createnetworkservice "EXO $name" "$name" 2>/dev/null \\
-              || continue
-          networksetup -setdhcp "EXO $name"
-                ;;
-        *)
-          networksetup -listallnetworkservices \\
-            | grep -q "$name" \\
-              || networksetup -createnetworkservice "$name" "$name" 2>/dev/null \\
-              || continue
-                ;;
-      esac
-    done
+        networksetup -switchtolocation exo
+        networksetup -listallhardwareports \\
+          | awk -F': ' '/Hardware Port: / {print $2}' \\
+          | while IFS=":" read -r name; do
+              case "$name" in
+                "Ethernet Adapter"*)
+                        ;;
+                "Thunderbolt Bridge")
+                        ;;
+                "Thunderbolt "*)
+                  networksetup -listallnetworkservices \\
+                    | grep -q "EXO $name" \\
+                      || networksetup -createnetworkservice "EXO $name" "$name" 2>/dev/null \\
+                      || continue
+                  networksetup -setdhcp "EXO $name"
+                        ;;
+                *)
+                  networksetup -listallnetworkservices \\
+                    | grep -q "$name" \\
+                      || networksetup -createnetworkservice "$name" "$name" 2>/dev/null \\
+                      || continue
+                        ;;
+              esac
+            done
 
-networksetup -listnetworkservices | grep -q "Thunderbolt Bridge" && {
-  networksetup -setnetworkserviceenabled "Thunderbolt Bridge" off
-} || true
-"""
+        networksetup -listnetworkservices | grep -q "Thunderbolt Bridge" && {
+          networksetup -setnetworkserviceenabled "Thunderbolt Bridge" off
+        } || true
+        """
 
     static func ensureLaunchDaemonInstalled() {
-        Task.detached {
+        // Use .utility priority to match NSAppleScript's internal QoS and avoid priority inversion
+        Task.detached(priority: .utility) {
             do {
                 if daemonAlreadyInstalled() {
                     return
@@ -70,11 +72,70 @@ networksetup -listnetworkservices | grep -q "Thunderbolt Bridge" && {
                 try await installLaunchDaemon()
                 logger.info("Network setup launch daemon installed and started")
             } catch {
-                logger.error("Network setup launch daemon failed: \(error.localizedDescription, privacy: .public)")
+                logger.error(
+                    "Network setup launch daemon failed: \(error.localizedDescription, privacy: .public)"
+                )
             }
         }
     }
 
+    /// Removes all EXO network setup components from the system.
+    /// This includes the LaunchDaemon, scripts, logs, and network location.
+    /// Requires admin privileges.
+    static func uninstall() throws {
+        let uninstallScript = makeUninstallScript()
+        try runShellAsAdmin(uninstallScript)
+        logger.info("EXO network setup components removed successfully")
+    }
+
+    /// Checks if there are any EXO network components installed that need cleanup
+    static func hasInstalledComponents() -> Bool {
+        let manager = FileManager.default
+        let scriptExists = manager.fileExists(atPath: scriptDestination)
+        let plistExists = manager.fileExists(atPath: plistDestination)
+        return scriptExists || plistExists
+    }
+
+    private static func makeUninstallScript() -> String {
+        """
+        set -euo pipefail
+
+        LABEL="\(daemonLabel)"
+        SCRIPT_DEST="\(scriptDestination)"
+        PLIST_DEST="\(plistDestination)"
+        LOG_OUT="/var/log/\(daemonLabel).log"
+        LOG_ERR="/var/log/\(daemonLabel).err.log"
+
+        # Unload the LaunchDaemon if running
+        launchctl bootout system/"$LABEL" 2>/dev/null || true
+
+        # Remove LaunchDaemon plist
+        rm -f "$PLIST_DEST"
+
+        # Remove the script and parent directory if empty
+        rm -f "$SCRIPT_DEST"
+        rmdir "$(dirname "$SCRIPT_DEST")" 2>/dev/null || true
+
+        # Remove log files
+        rm -f "$LOG_OUT" "$LOG_ERR"
+
+        # Switch back to Automatic network location
+        networksetup -switchtolocation Automatic 2>/dev/null || true
+
+        # Delete the exo network location if it exists
+        networksetup -listlocations | grep -q '^exo$' && {
+          networksetup -deletelocation exo 2>/dev/null || true
+        } || true
+
+        # Re-enable Thunderbolt Bridge if it exists
+        networksetup -listnetworkservices | grep -q "Thunderbolt Bridge" && {
+          networksetup -setnetworkserviceenabled "Thunderbolt Bridge" on 2>/dev/null || true
+        } || true
+
+        echo "EXO network components removed successfully"
+        """
+    }
+
     private static func daemonAlreadyInstalled() -> Bool {
         let manager = FileManager.default
         let scriptExists = manager.fileExists(atPath: scriptDestination)
@@ -82,7 +143,8 @@ networksetup -listnetworkservices | grep -q "Thunderbolt Bridge" && {
         guard scriptExists, plistExists else { return false }
         guard
             let data = try? Data(contentsOf: URL(fileURLWithPath: plistDestination)),
-            let plist = try? PropertyListSerialization.propertyList(from: data, options: [], format: nil) as? [String: Any]
+            let plist = try? PropertyListSerialization.propertyList(
+                from: data, options: [], format: nil) as? [String: Any]
         else {
             return false
         }
@@ -92,7 +154,9 @@ networksetup -listnetworkservices | grep -q "Thunderbolt Bridge" && {
         else {
             return false
         }
-        if let programArgs = plist["ProgramArguments"] as? [String], programArgs.contains(scriptDestination) == false {
+        if let programArgs = plist["ProgramArguments"] as? [String],
+            programArgs.contains(scriptDestination) == false
+        {
             return false
         }
         return true
@@ -105,58 +169,59 @@ networksetup -listnetworkservices | grep -q "Thunderbolt Bridge" && {
 
     private static func makeInstallerScript() -> String {
         """
-set -euo pipefail
+        set -euo pipefail
 
-LABEL="\(daemonLabel)"
-SCRIPT_DEST="\(scriptDestination)"
-PLIST_DEST="\(plistDestination)"
+        LABEL="\(daemonLabel)"
+        SCRIPT_DEST="\(scriptDestination)"
+        PLIST_DEST="\(plistDestination)"
 
-mkdir -p "$(dirname "$SCRIPT_DEST")"
+        mkdir -p "$(dirname "$SCRIPT_DEST")"
 
-cat > "$SCRIPT_DEST" <<'EOF_SCRIPT'
-\(setupScript)
-EOF_SCRIPT
-chmod 755 "$SCRIPT_DEST"
+        cat > "$SCRIPT_DEST" <<'EOF_SCRIPT'
+        \(setupScript)
+        EOF_SCRIPT
+        chmod 755 "$SCRIPT_DEST"
 
-cat > "$PLIST_DEST" <<'EOF_PLIST'
-<?xml version="1.0" encoding="UTF-8"?>
-<!DOCTYPE plist PUBLIC "-//Apple//DTD PLIST 1.0//EN" "http://www.apple.com/DTDs/PropertyList-1.0.dtd">
-<plist version="1.0">
-<dict>
-  <key>Label</key>
-  <string>\(daemonLabel)</string>
-  <key>ProgramArguments</key>
-  <array>
-    <string>/bin/bash</string>
-    <string>\(scriptDestination)</string>
-  </array>
-  <key>StartInterval</key>
-  <integer>\(requiredStartInterval)</integer>
-  <key>RunAtLoad</key>
-  <true/>
-  <key>StandardOutPath</key>
-  <string>/var/log/\(daemonLabel).log</string>
-  <key>StandardErrorPath</key>
-  <string>/var/log/\(daemonLabel).err.log</string>
-</dict>
-</plist>
-EOF_PLIST
+        cat > "$PLIST_DEST" <<'EOF_PLIST'
+        <?xml version="1.0" encoding="UTF-8"?>
+        <!DOCTYPE plist PUBLIC "-//Apple//DTD PLIST 1.0//EN" "http://www.apple.com/DTDs/PropertyList-1.0.dtd">
+        <plist version="1.0">
+        <dict>
+          <key>Label</key>
+          <string>\(daemonLabel)</string>
+          <key>ProgramArguments</key>
+          <array>
+            <string>/bin/bash</string>
+            <string>\(scriptDestination)</string>
+          </array>
+          <key>StartInterval</key>
+          <integer>\(requiredStartInterval)</integer>
+          <key>RunAtLoad</key>
+          <true/>
+          <key>StandardOutPath</key>
+          <string>/var/log/\(daemonLabel).log</string>
+          <key>StandardErrorPath</key>
+          <string>/var/log/\(daemonLabel).err.log</string>
+        </dict>
+        </plist>
+        EOF_PLIST
 
-launchctl bootout system/"$LABEL" >/dev/null 2>&1 || true
-launchctl bootstrap system "$PLIST_DEST"
-launchctl enable system/"$LABEL"
-launchctl kickstart -k system/"$LABEL"
-"""
+        launchctl bootout system/"$LABEL" >/dev/null 2>&1 || true
+        launchctl bootstrap system "$PLIST_DEST"
+        launchctl enable system/"$LABEL"
+        launchctl kickstart -k system/"$LABEL"
+        """
     }
 
     private static func runShellAsAdmin(_ script: String) throws {
-        let escapedScript = script
+        let escapedScript =
+            script
             .replacingOccurrences(of: "\\", with: "\\\\")
             .replacingOccurrences(of: "\"", with: "\\\"")
 
         let appleScriptSource = """
-do shell script "\(escapedScript)" with administrator privileges
-"""
+            do shell script "\(escapedScript)" with administrator privileges
+            """
 
         guard let appleScript = NSAppleScript(source: appleScriptSource) else {
             throw NetworkSetupError.scriptCreationFailed

app/EXO/EXO/Services/NetworkStatusService.swift

@@ -35,14 +35,34 @@ struct NetworkStatus: Equatable {
     let thunderboltBridgeState: ThunderboltState?
     let bridgeInactive: Bool?
     let interfaceStatuses: [InterfaceIpStatus]
+    let rdmaStatus: RDMAStatus
 
     static let empty = NetworkStatus(
         thunderboltBridgeState: nil,
         bridgeInactive: nil,
-        interfaceStatuses: []
+        interfaceStatuses: [],
+        rdmaStatus: .empty
     )
 }
 
+struct RDMAStatus: Equatable {
+    let rdmaCtlEnabled: Bool?
+    let devices: [String]
+    let activePorts: [RDMAPort]
+
+    var isAvailable: Bool {
+        rdmaCtlEnabled == true || !devices.isEmpty
+    }
+
+    static let empty = RDMAStatus(rdmaCtlEnabled: nil, devices: [], activePorts: [])
+}
+
+struct RDMAPort: Equatable {
+    let device: String
+    let port: String
+    let state: String
+}
+
 struct InterfaceIpStatus: Equatable {
     let interfaceName: String
     let ipAddress: String?
@@ -59,10 +79,79 @@ private struct NetworkStatusFetcher {
         NetworkStatus(
             thunderboltBridgeState: readThunderboltBridgeState(),
             bridgeInactive: readBridgeInactive(),
-            interfaceStatuses: readInterfaceStatuses()
+            interfaceStatuses: readInterfaceStatuses(),
+            rdmaStatus: readRDMAStatus()
         )
     }
 
+    private func readRDMAStatus() -> RDMAStatus {
+        let rdmaCtlEnabled = readRDMACtlEnabled()
+        let devices = readRDMADevices()
+        let activePorts = readRDMAActivePorts()
+        return RDMAStatus(
+            rdmaCtlEnabled: rdmaCtlEnabled, devices: devices, activePorts: activePorts)
+    }
+
+    private func readRDMACtlEnabled() -> Bool? {
+        let result = runCommand(["rdma_ctl", "status"])
+        guard result.exitCode == 0 else { return nil }
+        let output = result.output.lowercased().trimmingCharacters(in: .whitespacesAndNewlines)
+        if output.contains("enabled") {
+            return true
+        }
+        if output.contains("disabled") {
+            return false
+        }
+        return nil
+    }
+
+    private func readRDMADevices() -> [String] {
+        let result = runCommand(["ibv_devices"])
+        guard result.exitCode == 0 else { return [] }
+        var devices: [String] = []
+        for line in result.output.split(separator: "\n") {
+            let trimmed = line.trimmingCharacters(in: .whitespaces)
+            if trimmed.hasPrefix("---") || trimmed.lowercased().hasPrefix("device")
+                || trimmed.isEmpty
+            {
+                continue
+            }
+            let parts = trimmed.split(separator: " ", maxSplits: 1)
+            if let deviceName = parts.first {
+                devices.append(String(deviceName))
+            }
+        }
+        return devices
+    }
+
+    private func readRDMAActivePorts() -> [RDMAPort] {
+        let result = runCommand(["ibv_devinfo"])
+        guard result.exitCode == 0 else { return [] }
+        var ports: [RDMAPort] = []
+        var currentDevice: String?
+        var currentPort: String?
+
+        for line in result.output.split(separator: "\n") {
+            let trimmed = line.trimmingCharacters(in: .whitespaces)
+            if trimmed.hasPrefix("hca_id:") {
+                currentDevice = trimmed.replacingOccurrences(of: "hca_id:", with: "")
+                    .trimmingCharacters(in: .whitespaces)
+            } else if trimmed.hasPrefix("port:") {
+                currentPort = trimmed.replacingOccurrences(of: "port:", with: "")
+                    .trimmingCharacters(in: .whitespaces)
+            } else if trimmed.hasPrefix("state:") {
+                let state = trimmed.replacingOccurrences(of: "state:", with: "").trimmingCharacters(
+                    in: .whitespaces)
+                if let device = currentDevice, let port = currentPort {
+                    if state.lowercased().contains("active") {
+                        ports.append(RDMAPort(device: device, port: port, state: state))
+                    }
+                }
+            }
+        }
+        return ports
+    }
+
     private func readThunderboltBridgeState() -> ThunderboltState? {
         let result = runCommand(["networksetup", "-getnetworkserviceenabled", "Thunderbolt Bridge"])
         guard result.exitCode == 0 else {
@@ -85,10 +174,11 @@ private struct NetworkStatusFetcher {
     private func readBridgeInactive() -> Bool? {
         let result = runCommand(["ifconfig", "bridge0"])
         guard result.exitCode == 0 else { return nil }
-        guard let statusLine = result.output
-            .components(separatedBy: .newlines)
-            .first(where: { $0.contains("status:") })?
-            .lowercased()
+        guard
+            let statusLine = result.output
+                .components(separatedBy: .newlines)
+                .first(where: { $0.contains("status:") })?
+                .lowercased()
         else {
             return nil
         }
@@ -171,4 +261,3 @@ private struct NetworkStatusFetcher {
         )
     }
 }
-

app/EXO/EXO/ViewModels/InstanceViewModel.swift

@@ -57,7 +57,7 @@ struct InstanceViewModel: Identifiable, Equatable {
         case waiting
         case failed
         case idle
-        case unknown
+        case preparing
 
         var label: String {
             switch self {
@@ -68,7 +68,7 @@ struct InstanceViewModel: Identifiable, Equatable {
             case .waiting: return "Waiting"
             case .failed: return "Failed"
             case .idle: return "Idle"
-            case .unknown: return "Unknown"
+            case .preparing: return "Preparing"
             }
         }
     }
@@ -107,10 +107,13 @@ extension ClusterState {
             let nodeToRunner = instance.shardAssignments.nodeToRunner
             let nodeIds = Array(nodeToRunner.keys)
             let runnerIds = Array(nodeToRunner.values)
-            let nodeNames = nodeIds.compactMap { nodeProfiles[$0]?.friendlyName ?? nodeProfiles[$0]?.modelId ?? $0 }
+            let nodeNames = nodeIds.compactMap {
+                nodeProfiles[$0]?.friendlyName ?? nodeProfiles[$0]?.modelId ?? $0
+            }
             let statuses = runnerIds.compactMap { runners[$0]?.status.lowercased() }
             let downloadProgress = aggregateDownloadProgress(for: nodeIds)
-            let state = InstanceViewModel.State(statuses: statuses, hasActiveDownload: downloadProgress != nil)
+            let state = InstanceViewModel.State(
+                statuses: statuses, hasActiveDownload: downloadProgress != nil)
             let chatTasks = (chatTasksByInstance[entry.key] ?? [])
                 .sorted(by: { $0.sortPriority < $1.sortPriority })
                 .map { InstanceTaskViewModel(task: $0) }
@@ -165,8 +168,8 @@ extension ClusterState {
     }
 }
 
-private extension InstanceViewModel.State {
-    init(statuses: [String], hasActiveDownload: Bool = false) {
+extension InstanceViewModel.State {
+    fileprivate init(statuses: [String], hasActiveDownload: Bool = false) {
         if statuses.contains(where: { $0.contains("failed") }) {
             self = .failed
         } else if hasActiveDownload || statuses.contains(where: { $0.contains("downloading") }) {
@@ -182,7 +185,7 @@ private extension InstanceViewModel.State {
         } else if statuses.isEmpty {
             self = .idle
         } else {
-            self = .unknown
+            self = .preparing
         }
     }
 }
@@ -243,4 +246,3 @@ extension InstanceTaskViewModel {
         self.parameters = task.parameters
     }
 }
-

app/EXO/EXO/ViewModels/NodeViewModel.swift

@@ -87,7 +87,9 @@ struct TopologyViewModel {
 extension ClusterState {
     func topologyViewModel(localNodeId: String?) -> TopologyViewModel? {
         let topologyNodeIds = Set(topology?.nodes.map(\.nodeId) ?? [])
-        let allNodes = nodeViewModels().filter { topologyNodeIds.isEmpty || topologyNodeIds.contains($0.id) }
+        let allNodes = nodeViewModels().filter {
+            topologyNodeIds.isEmpty || topologyNodeIds.contains($0.id)
+        }
         guard !allNodes.isEmpty else { return nil }
 
         let nodesById = Dictionary(uniqueKeysWithValues: allNodes.map { ($0.id, $0) })
@@ -106,18 +108,24 @@ extension ClusterState {
         }
 
         // Rotate so the local node (from /node_id API) is first
-        if let localId = localNodeId, let index = orderedNodes.firstIndex(where: { $0.id == localId }) {
+        if let localId = localNodeId,
+            let index = orderedNodes.firstIndex(where: { $0.id == localId })
+        {
             orderedNodes = Array(orderedNodes[index...]) + Array(orderedNodes[..<index])
         }
 
         let nodeIds = Set(orderedNodes.map(\.id))
-        let edgesArray: [TopologyEdgeViewModel] = topology?.connections?.compactMap { connection in
-            guard nodeIds.contains(connection.localNodeId), nodeIds.contains(connection.sendBackNodeId) else { return nil }
-            return TopologyEdgeViewModel(sourceId: connection.localNodeId, targetId: connection.sendBackNodeId)
-        } ?? []
+        let edgesArray: [TopologyEdgeViewModel] =
+            topology?.connections?.compactMap { connection in
+                guard nodeIds.contains(connection.localNodeId),
+                    nodeIds.contains(connection.sendBackNodeId)
+                else { return nil }
+                return TopologyEdgeViewModel(
+                    sourceId: connection.localNodeId, targetId: connection.sendBackNodeId)
+            } ?? []
         let edges = Set(edgesArray)
 
-        return TopologyViewModel(nodes: orderedNodes, edges: Array(edges), currentNodeId: localNodeId)
+        return TopologyViewModel(
+            nodes: orderedNodes, edges: Array(edges), currentNodeId: localNodeId)
     }
 }
-

app/EXO/EXO/Views/InstanceRowView.swift

@@ -20,8 +20,8 @@ struct InstanceRowView: View {
                 if let progress = instance.downloadProgress {
                     downloadStatusView(progress: progress)
                 } else {
-                statusChip(label: instance.state.label.uppercased(), color: statusColor)
-            }
+                    statusChip(label: instance.state.label.uppercased(), color: statusColor)
+                }
             }
             if let progress = instance.downloadProgress {
                 GeometryReader { geometry in
@@ -83,7 +83,7 @@ struct InstanceRowView: View {
         case .ready: return .teal
         case .waiting, .idle: return .gray
         case .failed: return .red
-        case .unknown: return .secondary
+        case .preparing: return .secondary
         }
     }
 
@@ -97,7 +97,8 @@ struct InstanceRowView: View {
                         .font(.caption)
                         .fontWeight(.semibold)
                     if let subtitle = task.subtitle,
-                       subtitle.caseInsensitiveCompare(parentModelName) != .orderedSame {
+                        subtitle.caseInsensitiveCompare(parentModelName) != .orderedSame
+                    {
                         Text(subtitle)
                             .font(.caption2)
                             .foregroundColor(.secondary)
@@ -234,9 +235,12 @@ struct InstanceRowView: View {
         Button {
             isExpanded.wrappedValue.toggle()
         } label: {
-            Label(isExpanded.wrappedValue ? "Hide" : "Show", systemImage: isExpanded.wrappedValue ? "chevron.up" : "chevron.down")
-                .labelStyle(.titleAndIcon)
-                .contentTransition(.symbolEffect(.replace))
+            Label(
+                isExpanded.wrappedValue ? "Hide" : "Show",
+                systemImage: isExpanded.wrappedValue ? "chevron.up" : "chevron.down"
+            )
+            .labelStyle(.titleAndIcon)
+            .contentTransition(.symbolEffect(.replace))
         }
         .buttonStyle(.plain)
         .font(.caption2)
@@ -311,7 +315,9 @@ struct InstanceRowView: View {
         }
 
         @ViewBuilder
-        private func detailRow(icon: String? = nil, title: String, value: String, tint: Color = .secondary) -> some View {
+        private func detailRow(
+            icon: String? = nil, title: String, value: String, tint: Color = .secondary
+        ) -> some View {
             HStack(alignment: .firstTextBaseline, spacing: 6) {
                 if let icon {
                     Image(systemName: icon)
@@ -329,4 +335,3 @@ struct InstanceRowView: View {
         }
     }
 }
-

app/EXO/EXO/Views/NodeDetailView.swift

@@ -32,4 +32,3 @@ struct NodeDetailView: View {
         }
     }
 }
-

app/EXO/EXO/Views/NodeRowView.swift

@@ -28,4 +28,3 @@ struct NodeRowView: View {
         .padding(.vertical, 4)
     }
 }
-

app/EXO/EXO/Views/TopologyMiniView.swift

@@ -76,30 +76,33 @@ struct TopologyMiniView: View {
 
     private func connectionLines(in size: CGSize) -> some View {
         let positions = positionedNodes(in: size)
-        let positionById = Dictionary(uniqueKeysWithValues: positions.map { ($0.node.id, $0.point) })
+        let positionById = Dictionary(
+            uniqueKeysWithValues: positions.map { ($0.node.id, $0.point) })
         return Canvas { context, _ in
             guard !topology.edges.isEmpty else { return }
             let nodeRadius: CGFloat = 32
             let arrowLength: CGFloat = 10
             let arrowSpread: CGFloat = .pi / 7
             for edge in topology.edges {
-                guard let start = positionById[edge.sourceId], let end = positionById[edge.targetId] else { continue }
+                guard let start = positionById[edge.sourceId], let end = positionById[edge.targetId]
+                else { continue }
                 let dx = end.x - start.x
                 let dy = end.y - start.y
                 let distance = max(CGFloat(hypot(dx, dy)), 1)
                 let ux = dx / distance
                 let uy = dy / distance
-                let adjustedStart = CGPoint(x: start.x + ux * nodeRadius, y: start.y + uy * nodeRadius)
+                let adjustedStart = CGPoint(
+                    x: start.x + ux * nodeRadius, y: start.y + uy * nodeRadius)
                 let adjustedEnd = CGPoint(x: end.x - ux * nodeRadius, y: end.y - uy * nodeRadius)
 
                 var linePath = Path()
                 linePath.move(to: adjustedStart)
                 linePath.addLine(to: adjustedEnd)
-            context.stroke(
+                context.stroke(
                     linePath,
                     with: .color(.secondary.opacity(0.3)),
-                style: StrokeStyle(lineWidth: 1, dash: [4, 4])
-            )
+                    style: StrokeStyle(lineWidth: 1, dash: [4, 4])
+                )
 
                 let angle = atan2(uy, ux)
                 let tip = adjustedEnd
@@ -168,5 +171,3 @@ private struct NodeGlyphView: View {
         .frame(width: 95)
     }
 }
-
-

app/EXO/EXOTests/EXOTests.swift

@@ -6,6 +6,7 @@
 //
 
 import Testing
+
 @testable import EXO
 
 struct EXOTests {

app/EXO/uninstall-exo.sh

@@ -0,0 +1,154 @@
+#!/usr/bin/env bash
+#
+# EXO Uninstaller Script
+#
+# This script removes all EXO system components that persist after deleting the app.
+# Run with: sudo ./uninstall-exo.sh
+#
+# Components removed:
+# - LaunchDaemon: /Library/LaunchDaemons/io.exo.networksetup.plist
+# - Network script: /Library/Application Support/EXO/
+# - Log files: /var/log/io.exo.networksetup.*
+# - Network location: "exo"
+# - Launch at login registration
+#
+
+set -euo pipefail
+
+LABEL="io.exo.networksetup"
+SCRIPT_DEST="/Library/Application Support/EXO/disable_bridge_enable_dhcp.sh"
+PLIST_DEST="/Library/LaunchDaemons/io.exo.networksetup.plist"
+LOG_OUT="/var/log/${LABEL}.log"
+LOG_ERR="/var/log/${LABEL}.err.log"
+APP_BUNDLE_ID="io.exo.EXO"
+
+# Colors for output
+RED='\033[0;31m'
+GREEN='\033[0;32m'
+YELLOW='\033[1;33m'
+NC='\033[0m' # No Color
+
+echo_info() {
+    echo -e "${GREEN}[INFO]${NC} $1"
+}
+
+echo_warn() {
+    echo -e "${YELLOW}[WARN]${NC} $1"
+}
+
+echo_error() {
+    echo -e "${RED}[ERROR]${NC} $1"
+}
+
+# Check if running as root
+if [[ $EUID -ne 0 ]]; then
+    echo_error "This script must be run as root (use sudo)"
+    exit 1
+fi
+
+echo ""
+echo "========================================"
+echo "        EXO Uninstaller"
+echo "========================================"
+echo ""
+
+# Unload the LaunchDaemon if running
+echo_info "Stopping network setup daemon..."
+if launchctl list | grep -q "$LABEL"; then
+    launchctl bootout system/"$LABEL" 2>/dev/null || true
+    echo_info "Daemon stopped"
+else
+    echo_warn "Daemon was not running"
+fi
+
+# Remove LaunchDaemon plist
+if [[ -f "$PLIST_DEST" ]]; then
+    rm -f "$PLIST_DEST"
+    echo_info "Removed LaunchDaemon plist"
+else
+    echo_warn "LaunchDaemon plist not found (already removed?)"
+fi
+
+# Remove the script and parent directory
+if [[ -f "$SCRIPT_DEST" ]]; then
+    rm -f "$SCRIPT_DEST"
+    echo_info "Removed network setup script"
+else
+    echo_warn "Network setup script not found (already removed?)"
+fi
+
+# Remove EXO directory if empty
+if [[ -d "/Library/Application Support/EXO" ]]; then
+    rmdir "/Library/Application Support/EXO" 2>/dev/null && \
+        echo_info "Removed EXO support directory" || \
+        echo_warn "EXO support directory not empty, leaving in place"
+fi
+
+# Remove log files
+if [[ -f "$LOG_OUT" ]] || [[ -f "$LOG_ERR" ]]; then
+    rm -f "$LOG_OUT" "$LOG_ERR"
+    echo_info "Removed log files"
+else
+    echo_warn "Log files not found (already removed?)"
+fi
+
+# Switch back to Automatic network location
+echo_info "Restoring network configuration..."
+if networksetup -listlocations | grep -q "^Automatic$"; then
+    networksetup -switchtolocation Automatic 2>/dev/null || true
+    echo_info "Switched to Automatic network location"
+else
+    echo_warn "Automatic network location not found"
+fi
+
+# Delete the exo network location if it exists
+if networksetup -listlocations | grep -q "^exo$"; then
+    networksetup -deletelocation exo 2>/dev/null || true
+    echo_info "Deleted 'exo' network location"
+else
+    echo_warn "'exo' network location not found (already removed?)"
+fi
+
+# Re-enable Thunderbolt Bridge if it exists
+if networksetup -listnetworkservices 2>/dev/null | grep -q "Thunderbolt Bridge"; then
+    networksetup -setnetworkserviceenabled "Thunderbolt Bridge" on 2>/dev/null || true
+    echo_info "Re-enabled Thunderbolt Bridge"
+fi
+
+# Note about launch at login registration
+# SMAppService-based login items cannot be removed from a shell script.
+# They can only be unregistered from within the app itself or manually via System Settings.
+echo_warn "Launch at login must be removed manually:"
+echo_warn "  System Settings → General → Login Items → Remove EXO"
+
+# Check if EXO.app exists in common locations
+APP_FOUND=false
+for app_path in "/Applications/EXO.app" "$HOME/Applications/EXO.app"; do
+    if [[ -d "$app_path" ]]; then
+        if [[ "$APP_FOUND" == false ]]; then
+            echo ""
+            APP_FOUND=true
+        fi
+        echo_warn "EXO.app found at: $app_path"
+        echo_warn "You may want to move it to Trash manually."
+    fi
+done
+
+echo ""
+echo "========================================"
+echo_info "EXO uninstall complete!"
+echo "========================================"
+echo ""
+echo "The following have been removed:"
+echo "  • Network setup LaunchDaemon"
+echo "  • Network configuration script"
+echo "  • Log files"
+echo "  • 'exo' network location"
+echo ""
+echo "Your network has been restored to use the 'Automatic' location."
+echo "Thunderbolt Bridge has been re-enabled (if present)."
+echo ""
+echo "Manual step required:"
+echo "  Remove EXO from Login Items in System Settings → General → Login Items"
+echo ""
+

dashboard/dashboard.nix

@@ -0,0 +1,60 @@
+{ lib
+, config
+, dream2nix
+, ...
+}:
+let
+  # Read and parse the lock file
+  rawLockFile = builtins.fromJSON (builtins.readFile "${config.deps.dashboardSrc}/package-lock.json");
+
+  # For packages with bundleDependencies, filter out deps that are bundled
+  # (bundled deps are inside the tarball, not separate lockfile entries)
+  fixedPackages = lib.mapAttrs
+    (path: entry:
+      if entry ? bundleDependencies && entry.bundleDependencies != [ ]
+      then entry // {
+        dependencies = lib.filterAttrs
+          (name: _: !(lib.elem name entry.bundleDependencies))
+          (entry.dependencies or { });
+      }
+      else entry
+    )
+    (rawLockFile.packages or { });
+
+  fixedLockFile = rawLockFile // { packages = fixedPackages; };
+in
+{
+  imports = [
+    dream2nix.modules.dream2nix.nodejs-package-lock-v3
+    dream2nix.modules.dream2nix.nodejs-granular-v3
+  ];
+
+  name = "exo-dashboard";
+  version = "1.0.0";
+
+  mkDerivation = {
+    src = config.deps.dashboardSrc;
+
+    buildPhase = ''
+      runHook preBuild
+      npm run build
+      runHook postBuild
+    '';
+
+    installPhase = ''
+      runHook preInstall
+      cp -r build $out/build
+      runHook postInstall
+    '';
+  };
+
+  deps = { nixpkgs, ... }: {
+    inherit (nixpkgs) stdenv;
+    dashboardSrc = null; # Injected by parts.nix
+  };
+
+  nodejs-package-lock-v3 = {
+    # Don't use packageLockFile - provide the fixed lock content directly
+    packageLock = fixedLockFile;
+  };
+}

dashboard/parts.nix

@@ -0,0 +1,44 @@
+{ inputs, ... }:
+{
+  perSystem =
+    { pkgs, lib, ... }:
+    let
+      # Filter source to only include dashboard directory
+      src = lib.cleanSourceWith {
+        src = inputs.self;
+        filter =
+          path: type:
+          let
+            baseName = builtins.baseNameOf path;
+            inDashboardDir =
+              (lib.hasInfix "/dashboard/" path)
+              || (lib.hasSuffix "/dashboard" (builtins.dirOf path))
+              || (baseName == "dashboard" && type == "directory");
+          in
+          inDashboardDir;
+      };
+
+      # Build the dashboard with dream2nix (includes node_modules in output)
+      dashboardFull = inputs.dream2nix.lib.evalModules {
+        packageSets.nixpkgs = pkgs;
+        modules = [
+          ./dashboard.nix
+          {
+            paths.projectRoot = inputs.self;
+            paths.projectRootFile = "flake.nix";
+            paths.package = inputs.self + "/dashboard";
+          }
+          # Inject the filtered source
+          {
+            deps.dashboardSrc = lib.mkForce "${src}/dashboard";
+          }
+        ];
+      };
+    in
+    {
+      # Extract just the static site from the full build
+      packages.dashboard = pkgs.runCommand "exo-dashboard" { } ''
+        cp -r ${dashboardFull}/build $out
+      '';
+    };
+}

dashboard/src/app.d.ts

@@ -11,4 +11,3 @@ declare global {
 }
 
 export {};
-

dashboard/src/lib/components/index.ts

@@ -1,8 +1,7 @@
-export { default as TopologyGraph } from './TopologyGraph.svelte';
-export { default as ChatForm } from './ChatForm.svelte';
-export { default as ChatMessages } from './ChatMessages.svelte';
-export { default as ChatAttachments } from './ChatAttachments.svelte';
-export { default as ChatSidebar } from './ChatSidebar.svelte';
-export { default as ModelCard } from './ModelCard.svelte';
-export { default as MarkdownContent } from './MarkdownContent.svelte';
-
+export { default as TopologyGraph } from "./TopologyGraph.svelte";
+export { default as ChatForm } from "./ChatForm.svelte";
+export { default as ChatMessages } from "./ChatMessages.svelte";
+export { default as ChatAttachments } from "./ChatAttachments.svelte";
+export { default as ChatSidebar } from "./ChatSidebar.svelte";
+export { default as ModelCard } from "./ModelCard.svelte";
+export { default as MarkdownContent } from "./MarkdownContent.svelte";

dashboard/src/lib/types/files.ts

@@ -13,55 +13,124 @@ export interface ChatUploadedFile {
 }
 
 export interface ChatAttachment {
-	type: 'image' | 'text' | 'pdf' | 'audio';
+	type: "image" | "text" | "pdf" | "audio";
 	name: string;
 	content?: string;
 	base64Url?: string;
 	mimeType?: string;
 }
 
-export type FileCategory = 'image' | 'text' | 'pdf' | 'audio' | 'unknown';
+export type FileCategory = "image" | "text" | "pdf" | "audio" | "unknown";
 
-export const IMAGE_EXTENSIONS = ['.jpg', '.jpeg', '.png', '.gif', '.webp', '.svg'];
-export const IMAGE_MIME_TYPES = ['image/jpeg', 'image/png', 'image/gif', 'image/webp', 'image/svg+xml'];
+export const IMAGE_EXTENSIONS = [
+	".jpg",
+	".jpeg",
+	".png",
+	".gif",
+	".webp",
+	".svg",
+];
+export const IMAGE_MIME_TYPES = [
+	"image/jpeg",
+	"image/png",
+	"image/gif",
+	"image/webp",
+	"image/svg+xml",
+];
 
 export const TEXT_EXTENSIONS = [
-	'.txt', '.md', '.json', '.xml', '.yaml', '.yml', '.csv', '.log',
-	'.js', '.ts', '.jsx', '.tsx', '.py', '.java', '.cpp', '.c', '.h',
-	'.css', '.html', '.htm', '.sql', '.sh', '.bat', '.rs', '.go',
-	'.rb', '.php', '.swift', '.kt', '.scala', '.r', '.dart', '.vue', '.svelte'
+	".txt",
+	".md",
+	".json",
+	".xml",
+	".yaml",
+	".yml",
+	".csv",
+	".log",
+	".js",
+	".ts",
+	".jsx",
+	".tsx",
+	".py",
+	".java",
+	".cpp",
+	".c",
+	".h",
+	".css",
+	".html",
+	".htm",
+	".sql",
+	".sh",
+	".bat",
+	".rs",
+	".go",
+	".rb",
+	".php",
+	".swift",
+	".kt",
+	".scala",
+	".r",
+	".dart",
+	".vue",
+	".svelte",
 ];
 export const TEXT_MIME_TYPES = [
-	'text/plain', 'text/markdown', 'text/csv', 'text/html', 'text/css',
-	'application/json', 'application/xml', 'text/xml', 'application/javascript',
-	'text/javascript', 'application/typescript'
+	"text/plain",
+	"text/markdown",
+	"text/csv",
+	"text/html",
+	"text/css",
+	"application/json",
+	"application/xml",
+	"text/xml",
+	"application/javascript",
+	"text/javascript",
+	"application/typescript",
 ];
 
-export const PDF_EXTENSIONS = ['.pdf'];
-export const PDF_MIME_TYPES = ['application/pdf'];
+export const PDF_EXTENSIONS = [".pdf"];
+export const PDF_MIME_TYPES = ["application/pdf"];
 
-export const AUDIO_EXTENSIONS = ['.mp3', '.wav', '.ogg', '.m4a'];
-export const AUDIO_MIME_TYPES = ['audio/mpeg', 'audio/wav', 'audio/ogg', 'audio/mp4'];
+export const AUDIO_EXTENSIONS = [".mp3", ".wav", ".ogg", ".m4a"];
+export const AUDIO_MIME_TYPES = [
+	"audio/mpeg",
+	"audio/wav",
+	"audio/ogg",
+	"audio/mp4",
+];
 
 /**
  * Get file category based on MIME type and extension
  */
-export function getFileCategory(mimeType: string, fileName: string): FileCategory {
-	const extension = fileName.toLowerCase().slice(fileName.lastIndexOf('.'));
-	
-	if (IMAGE_MIME_TYPES.includes(mimeType) || IMAGE_EXTENSIONS.includes(extension)) {
-		return 'image';
+export function getFileCategory(
+	mimeType: string,
+	fileName: string,
+): FileCategory {
+	const extension = fileName.toLowerCase().slice(fileName.lastIndexOf("."));
+
+	if (
+		IMAGE_MIME_TYPES.includes(mimeType) ||
+		IMAGE_EXTENSIONS.includes(extension)
+	) {
+		return "image";
 	}
 	if (PDF_MIME_TYPES.includes(mimeType) || PDF_EXTENSIONS.includes(extension)) {
-		return 'pdf';
+		return "pdf";
 	}
-	if (AUDIO_MIME_TYPES.includes(mimeType) || AUDIO_EXTENSIONS.includes(extension)) {
-		return 'audio';
+	if (
+		AUDIO_MIME_TYPES.includes(mimeType) ||
+		AUDIO_EXTENSIONS.includes(extension)
+	) {
+		return "audio";
 	}
-	if (TEXT_MIME_TYPES.includes(mimeType) || TEXT_EXTENSIONS.includes(extension) || mimeType.startsWith('text/')) {
-		return 'text';
+	if (
+		TEXT_MIME_TYPES.includes(mimeType) ||
+		TEXT_EXTENSIONS.includes(extension) ||
+		mimeType.startsWith("text/")
+	) {
+		return "text";
 	}
-	return 'unknown';
+	return "unknown";
 }
 
 /**
@@ -69,36 +138,36 @@ export function getFileCategory(mimeType: string, fileName: string): FileCategor
  */
 export function getAcceptString(categories: FileCategory[]): string {
 	const accepts: string[] = [];
-	
+
 	for (const category of categories) {
 		switch (category) {
-			case 'image':
+			case "image":
 				accepts.push(...IMAGE_EXTENSIONS, ...IMAGE_MIME_TYPES);
 				break;
-			case 'text':
+			case "text":
 				accepts.push(...TEXT_EXTENSIONS, ...TEXT_MIME_TYPES);
 				break;
-			case 'pdf':
+			case "pdf":
 				accepts.push(...PDF_EXTENSIONS, ...PDF_MIME_TYPES);
 				break;
-			case 'audio':
+			case "audio":
 				accepts.push(...AUDIO_EXTENSIONS, ...AUDIO_MIME_TYPES);
 				break;
 		}
 	}
-	
-	return accepts.join(',');
+
+	return accepts.join(",");
 }
 
 /**
  * Format file size for display
  */
 export function formatFileSize(bytes: number): string {
-	if (bytes === 0) return '0 B';
+	if (bytes === 0) return "0 B";
 	const k = 1024;
-	const sizes = ['B', 'KB', 'MB', 'GB'];
+	const sizes = ["B", "KB", "MB", "GB"];
 	const i = Math.floor(Math.log(bytes) / Math.log(k));
-	return parseFloat((bytes / Math.pow(k, i)).toFixed(1)) + ' ' + sizes[i];
+	return parseFloat((bytes / Math.pow(k, i)).toFixed(1)) + " " + sizes[i];
 }
 
 /**
@@ -128,42 +197,44 @@ export function readFileAsText(file: File): Promise<string> {
 /**
  * Process uploaded files into ChatUploadedFile format
  */
-export async function processUploadedFiles(files: File[]): Promise<ChatUploadedFile[]> {
+export async function processUploadedFiles(
+	files: File[],
+): Promise<ChatUploadedFile[]> {
 	const results: ChatUploadedFile[] = [];
-	
+
 	for (const file of files) {
-		const id = Date.now().toString() + Math.random().toString(36).substring(2, 9);
+		const id =
+			Date.now().toString() + Math.random().toString(36).substring(2, 9);
 		const category = getFileCategory(file.type, file.name);
-		
+
 		const base: ChatUploadedFile = {
 			id,
 			name: file.name,
 			size: file.size,
 			type: file.type,
-			file
+			file,
 		};
-		
+
 		try {
-			if (category === 'image') {
+			if (category === "image") {
 				const preview = await readFileAsDataURL(file);
 				results.push({ ...base, preview });
-			} else if (category === 'text' || category === 'unknown') {
+			} else if (category === "text" || category === "unknown") {
 				const textContent = await readFileAsText(file);
 				results.push({ ...base, textContent });
-			} else if (category === 'pdf') {
+			} else if (category === "pdf") {
 				results.push(base);
-			} else if (category === 'audio') {
+			} else if (category === "audio") {
 				const preview = await readFileAsDataURL(file);
 				results.push({ ...base, preview });
 			} else {
 				results.push(base);
 			}
 		} catch (error) {
-			console.error('Error processing file:', file.name, error);
+			console.error("Error processing file:", file.name, error);
 			results.push(base);
 		}
 	}
-	
+
 	return results;
 }
-

dashboard/src/routes/+page.svelte

@@ -593,7 +593,7 @@ function toggleInstanceDownloadDetails(nodeId: string): void {
 		// Unwrap the instance
 		const [instanceTag, instance] = getTagged(instanceWrapped);
 		if (!instance || typeof instance !== 'object') {
-			return { isDownloading: false, progress: null, statusText: 'UNKNOWN', perNode: [] };
+			return { isDownloading: false, progress: null, statusText: 'PREPARING', perNode: [] };
 		}
 
 		const inst = instance as { shardAssignments?: { nodeToRunner?: Record<string, string>; runnerToShard?: Record<string, unknown>; modelId?: string } };
@@ -706,7 +706,7 @@ function toggleInstanceDownloadDetails(nodeId: string): void {
 	function deriveInstanceStatus(instanceWrapped: unknown): { statusText: string; statusClass: string } {
 		const [, instance] = getTagged(instanceWrapped);
 		if (!instance || typeof instance !== 'object') {
-			return { statusText: 'UNKNOWN', statusClass: 'inactive' };
+			return { statusText: 'PREPARING', statusClass: 'inactive' };
 		}
 		
 		const inst = instance as { shardAssignments?: { runnerToShard?: Record<string, unknown> } };
@@ -735,7 +735,7 @@ function toggleInstanceDownloadDetails(nodeId: string): void {
 
 		const has = (s: string) => statuses.includes(s);
 
-		if (statuses.length === 0) return { statusText: 'UNKNOWN', statusClass: 'inactive' };
+		if (statuses.length === 0) return { statusText: 'PREPARING', statusClass: 'inactive' };
 		if (has('Failed')) return { statusText: 'FAILED', statusClass: 'failed' };
 		if (has('Shutdown')) return { statusText: 'SHUTDOWN', statusClass: 'inactive' };
 		if (has('Loading')) return { statusText: 'LOADING', statusClass: 'starting' };
@@ -1267,9 +1267,9 @@ function toggleInstanceDownloadDetails(nodeId: string): void {
 							<div class="flex-1 h-px bg-gradient-to-r from-exo-yellow/30 to-transparent"></div>
 						</div>
 						
-						<div 
+						<div
 							bind:this={instancesContainerRef}
-							class="max-h-72 space-y-3 overflow-y-auto"
+							class="max-h-72 xl:max-h-96 space-y-3 overflow-y-auto overflow-x-hidden py-px"
 						>
 								{#each Object.entries(instanceData) as [id, instance]}
 									{@const downloadInfo = getInstanceDownloadStatus(id, instance)}
@@ -1773,7 +1773,7 @@ function toggleInstanceDownloadDetails(nodeId: string): void {
 								<h3 class="text-xs text-exo-yellow font-mono tracking-[0.2em] uppercase">Instances</h3>
 								<div class="flex-1 h-px bg-gradient-to-r from-exo-yellow/30 to-transparent"></div>
 							</div>
-								<div class="space-y-3 max-h-72 overflow-y-auto pr-1">
+								<div class="space-y-3 max-h-72 xl:max-h-96 overflow-y-auto overflow-x-hidden py-px pr-1">
 									{#each Object.entries(instanceData) as [id, instance]}
 										{@const downloadInfo = getInstanceDownloadStatus(id, instance)}
 										{@const statusText = downloadInfo.statusText}

dashboard/src/routes/downloads/+page.svelte

@@ -199,7 +199,13 @@
 					const rawProgress = (downloadPayload as Record<string, unknown>).download_progress
 						?? (downloadPayload as Record<string, unknown>).downloadProgress
 						?? {};
-					const totalBytes = getBytes((rawProgress as Record<string, unknown>).total_bytes ?? (rawProgress as Record<string, unknown>).totalBytes);
+					// For DownloadCompleted, total_bytes is at top level; for DownloadOngoing, it's inside download_progress
+					const totalBytes = getBytes(
+						(downloadPayload as Record<string, unknown>).total_bytes
+						?? (downloadPayload as Record<string, unknown>).totalBytes
+						?? (rawProgress as Record<string, unknown>).total_bytes
+						?? (rawProgress as Record<string, unknown>).totalBytes
+					);
 					const downloadedBytes = getBytes((rawProgress as Record<string, unknown>).downloaded_bytes ?? (rawProgress as Record<string, unknown>).downloadedBytes);
 					const speed = (rawProgress as Record<string, unknown>).speed as number ?? 0;
 					const etaMs = (rawProgress as Record<string, unknown>).eta_ms as number ?? (rawProgress as Record<string, unknown>).etaMs as number ?? 0;
@@ -332,8 +338,13 @@
 								<div class="text-lg font-mono text-white truncate">{node.nodeName}</div>
 								<div class="text-xs text-exo-light-gray font-mono truncate">{node.nodeId}</div>
 							</div>
-							<div class="text-xs font-mono uppercase tracking-wider whitespace-nowrap shrink-0">
-								<span class="text-green-400">{node.models.filter(m => m.status === 'completed').length}</span><span class="text-exo-yellow"> /{node.models.length} models</span>
+							<div class="text-xs font-mono uppercase tracking-wider whitespace-nowrap shrink-0 text-right">
+								<div>
+									<span class="text-green-400">{node.models.filter(m => m.status === 'completed').length}</span><span class="text-exo-yellow"> / {node.models.length} models</span>
+								</div>
+								<div class="text-exo-light-gray normal-case tracking-normal">
+									{formatBytes(node.models.filter(m => m.status === 'completed').reduce((sum, m) => sum + m.totalBytes, 0))} on disk
+								</div>
 							</div>
 						</div>
 
@@ -385,7 +396,7 @@
 								</div>
 
 								<div class="flex items-center justify-between text-xs font-mono text-exo-light-gray">
-									<span>{model.status === 'completed' ? 'Completed' : `${formatSpeed(model.speed)} • ETA ${formatEta(model.etaMs)}`}</span>
+									<span>{model.status === 'completed' ? `Completed (${formatBytes(model.totalBytes)})` : `${formatSpeed(model.speed)} • ETA ${formatEta(model.etaMs)}`}</span>
 									{#if model.status !== 'completed'}
 										<span>{model.files.length} file{model.files.length === 1 ? '' : 's'}</span>
 									{/if}

dashboard/vite.config.ts

@@ -1,16 +1,15 @@
-import tailwindcss from '@tailwindcss/vite';
-import { sveltekit } from '@sveltejs/kit/vite';
-import { defineConfig } from 'vite';
+import tailwindcss from "@tailwindcss/vite";
+import { sveltekit } from "@sveltejs/kit/vite";
+import { defineConfig } from "vite";
 
 export default defineConfig({
 	plugins: [tailwindcss(), sveltekit()],
 	server: {
 		proxy: {
-			'/v1': 'http://localhost:52415',
-			'/state': 'http://localhost:52415',
-			'/models': 'http://localhost:52415',
-			'/instance': 'http://localhost:52415'
-		}
-	}
+			"/v1": "http://localhost:52415",
+			"/state": "http://localhost:52415",
+			"/models": "http://localhost:52415",
+			"/instance": "http://localhost:52415",
+		},
+	},
 });
-

docs/api.md

@@ -0,0 +1,212 @@
+# EXO API – Technical Reference
+
+This document describes the REST API exposed by the **EXO ** service, as implemented in:
+
+`src/exo/master/api.py`
+
+The API is used to manage model instances in the cluster, inspect cluster state, and perform inference using an OpenAI-compatible interface.
+
+Base URL example:
+
+```
+http://localhost:52415
+```
+
+## 1. General / Meta Endpoints
+
+### Get Master Node ID
+
+**GET** `/node_id`
+
+Returns the identifier of the current master node.
+
+**Response (example):**
+
+```json
+{
+  "node_id": "node-1234"
+}
+```
+
+### Get Cluster State
+
+**GET** `/state`
+
+Returns the current state of the cluster, including nodes and active instances.
+
+**Response:**
+JSON object describing topology, nodes, and instances.
+
+### Get Events
+
+**GET** `/events`
+
+Returns the list of internal events recorded by the master (mainly for debugging and observability).
+
+**Response:**
+Array of event objects.
+
+## 2. Model Instance Management
+
+### Create Instance
+
+**POST** `/instance`
+
+Creates a new model instance in the cluster.
+
+**Request body (example):**
+
+```json
+{
+  "instance": {
+    "model_id": "llama-3.2-1b",
+    "placement": { }
+  }
+}
+```
+
+**Response:**
+JSON description of the created instance.
+
+### Delete Instance
+
+**DELETE** `/instance/{instance_id}`
+
+Deletes an existing instance by ID.
+
+**Path parameters:**
+
+* `instance_id`: string, ID of the instance to delete
+
+**Response:**
+Status / confirmation JSON.
+
+### Get Instance
+
+**GET** `/instance/{instance_id}`
+
+Returns details of a specific instance.
+
+**Path parameters:**
+
+* `instance_id`: string
+
+**Response:**
+JSON description of the instance.
+
+### Preview Placements
+
+**GET** `/instance/previews?model_id=...`
+
+Returns possible placement previews for a given model.
+
+**Query parameters:**
+
+* `model_id`: string, required
+
+**Response:**
+Array of placement preview objects.
+
+### Compute Placement
+
+**GET** `/instance/placement`
+
+Computes a placement for a potential instance without creating it.
+
+**Query parameters (typical):**
+
+* `model_id`: string
+* `sharding`: string or config
+* `instance_meta`: JSON-encoded metadata
+* `min_nodes`: integer
+
+**Response:**
+JSON object describing the proposed placement / instance configuration.
+
+### Place Instance (Dry Operation)
+
+**POST** `/place_instance`
+
+Performs a placement operation for an instance (planning step), without necessarily creating it.
+
+**Request body:**
+JSON describing the instance to be placed.
+
+**Response:**
+Placement result.
+
+## 3. Models
+
+### List Models
+
+**GET** `/models`
+**GET** `/v1/models` (alias)
+
+Returns the list of available models and their metadata.
+
+**Response:**
+Array of model descriptors.
+
+## 4. Inference / Chat Completions
+
+### OpenAI-Compatible Chat Completions
+
+**POST** `/v1/chat/completions`
+
+Executes a chat completion request using an OpenAI-compatible schema. Supports streaming and non-streaming modes.
+
+**Request body (example):**
+
+```json
+{
+  "model": "llama-3.2-1b",
+  "messages": [
+    { "role": "system", "content": "You are a helpful assistant." },
+    { "role": "user", "content": "Hello" }
+  ],
+  "stream": false
+}
+```
+
+**Response:**
+OpenAI-compatible chat completion response.
+
+### Benchmarked Chat Completions
+
+**POST** `/bench/chat/completions`
+
+Same as `/v1/chat/completions`, but also returns performance and generation statistics.
+
+**Request body:**
+Same schema as `/v1/chat/completions`.
+
+**Response:**
+Chat completion plus benchmarking metrics.
+
+## 5. Complete Endpoint Summary
+
+```
+GET     /node_id
+GET     /state
+GET     /events
+
+POST    /instance
+GET     /instance/{instance_id}
+DELETE  /instance/{instance_id}
+
+GET     /instance/previews
+GET     /instance/placement
+POST    /place_instance
+
+GET     /models
+GET     /v1/models
+
+POST    /v1/chat/completions
+POST    /bench/chat/completions
+```
+
+## 6. Notes
+
+* The `/v1/chat/completions` endpoint is compatible with the OpenAI API format, so existing OpenAI clients can be pointed to EXO by changing the base URL.
+* The instance placement endpoints allow you to plan and preview cluster allocations before actually creating instances.
+* The `/events` and `/state` endpoints are primarily intended for operational visibility and debugging.

flake.lock

@@ -1,5 +1,42 @@
 {
   "nodes": {
+    "crane": {
+      "locked": {
+        "lastModified": 1767744144,
+        "narHash": "sha256-9/9ntI0D+HbN4G0TrK3KmHbTvwgswz7p8IEJsWyef8Q=",
+        "owner": "ipetkov",
+        "repo": "crane",
+        "rev": "2fb033290bf6b23f226d4c8b32f7f7a16b043d7e",
+        "type": "github"
+      },
+      "original": {
+        "owner": "ipetkov",
+        "repo": "crane",
+        "type": "github"
+      }
+    },
+    "dream2nix": {
+      "inputs": {
+        "nixpkgs": [
+          "nixpkgs"
+        ],
+        "purescript-overlay": "purescript-overlay",
+        "pyproject-nix": "pyproject-nix"
+      },
+      "locked": {
+        "lastModified": 1765953015,
+        "narHash": "sha256-5FBZbbWR1Csp3Y2icfRkxMJw/a/5FGg8hCXej2//bbI=",
+        "owner": "nix-community",
+        "repo": "dream2nix",
+        "rev": "69eb01fa0995e1e90add49d8ca5bcba213b0416f",
+        "type": "github"
+      },
+      "original": {
+        "owner": "nix-community",
+        "repo": "dream2nix",
+        "type": "github"
+      }
+    },
     "fenix": {
       "inputs": {
         "nixpkgs": [
@@ -8,11 +45,11 @@
         "rust-analyzer-src": "rust-analyzer-src"
       },
       "locked": {
-        "lastModified": 1761893049,
-        "narHash": "sha256-1TtFDPhC+ZsrOOtBnry1EZC+WipTTvsOVjIEVugqji8=",
+        "lastModified": 1768287139,
+        "narHash": "sha256-nsXFt0OzUi6K7dUzzJD5/v9e0Ic+fvclfIW936/43ZM=",
         "owner": "nix-community",
         "repo": "fenix",
-        "rev": "c2ac9a5c0d6d16630c3b225b874bd14528d1abe6",
+        "rev": "a4a3aa956931f90f35453cb519e4545e9ad7f773",
         "type": "github"
       },
       "original": {
@@ -21,25 +58,59 @@
         "type": "github"
       }
     },
-    "flake-utils": {
-      "inputs": {
-        "systems": "systems"
-      },
+    "flake-compat": {
+      "flake": false,
       "locked": {
-        "lastModified": 1731533236,
-        "narHash": "sha256-l0KFg5HjrsfsO/JpG+r7fRrqm12kzFHyUHqHCVpMMbI=",
-        "owner": "numtide",
-        "repo": "flake-utils",
-        "rev": "11707dc2f618dd54ca8739b309ec4fc024de578b",
+        "lastModified": 1696426674,
+        "narHash": "sha256-kvjfFW7WAETZlt09AgDn1MrtKzP7t90Vf7vypd3OL1U=",
+        "owner": "edolstra",
+        "repo": "flake-compat",
+        "rev": "0f9255e01c2351cc7d116c072cb317785dd33b33",
         "type": "github"
       },
       "original": {
-        "owner": "numtide",
-        "repo": "flake-utils",
+        "owner": "edolstra",
+        "repo": "flake-compat",
+        "type": "github"
+      }
+    },
+    "flake-parts": {
+      "inputs": {
+        "nixpkgs-lib": [
+          "nixpkgs"
+        ]
+      },
+      "locked": {
+        "lastModified": 1768135262,
+        "narHash": "sha256-PVvu7OqHBGWN16zSi6tEmPwwHQ4rLPU9Plvs8/1TUBY=",
+        "owner": "hercules-ci",
+        "repo": "flake-parts",
+        "rev": "80daad04eddbbf5a4d883996a73f3f542fa437ac",
+        "type": "github"
+      },
+      "original": {
+        "owner": "hercules-ci",
+        "repo": "flake-parts",
         "type": "github"
       }
     },
     "nixpkgs": {
+      "locked": {
+        "lastModified": 1768127708,
+        "narHash": "sha256-1Sm77VfZh3mU0F5OqKABNLWxOuDeHIlcFjsXeeiPazs=",
+        "owner": "NixOS",
+        "repo": "nixpkgs",
+        "rev": "ffbc9f8cbaacfb331b6017d5a5abb21a492c9a38",
+        "type": "github"
+      },
+      "original": {
+        "owner": "NixOS",
+        "ref": "nixos-unstable",
+        "repo": "nixpkgs",
+        "type": "github"
+      }
+    },
+    "nixpkgs-swift": {
       "locked": {
         "lastModified": 1761672384,
         "narHash": "sha256-o9KF3DJL7g7iYMZq9SWgfS1BFlNbsm6xplRjVlOCkXI=",
@@ -50,27 +121,74 @@
       },
       "original": {
         "owner": "NixOS",
-        "ref": "nixos-unstable",
         "repo": "nixpkgs",
+        "rev": "08dacfca559e1d7da38f3cf05f1f45ee9bfd213c",
+        "type": "github"
+      }
+    },
+    "purescript-overlay": {
+      "inputs": {
+        "flake-compat": "flake-compat",
+        "nixpkgs": [
+          "dream2nix",
+          "nixpkgs"
+        ],
+        "slimlock": "slimlock"
+      },
+      "locked": {
+        "lastModified": 1728546539,
+        "narHash": "sha256-Sws7w0tlnjD+Bjck1nv29NjC5DbL6nH5auL9Ex9Iz2A=",
+        "owner": "thomashoneyman",
+        "repo": "purescript-overlay",
+        "rev": "4ad4c15d07bd899d7346b331f377606631eb0ee4",
+        "type": "github"
+      },
+      "original": {
+        "owner": "thomashoneyman",
+        "repo": "purescript-overlay",
+        "type": "github"
+      }
+    },
+    "pyproject-nix": {
+      "inputs": {
+        "nixpkgs": [
+          "dream2nix",
+          "nixpkgs"
+        ]
+      },
+      "locked": {
+        "lastModified": 1763017646,
+        "narHash": "sha256-Z+R2lveIp6Skn1VPH3taQIuMhABg1IizJd8oVdmdHsQ=",
+        "owner": "pyproject-nix",
+        "repo": "pyproject.nix",
+        "rev": "47bd6f296502842643078d66128f7b5e5370790c",
+        "type": "github"
+      },
+      "original": {
+        "owner": "pyproject-nix",
+        "repo": "pyproject.nix",
         "type": "github"
       }
     },
     "root": {
       "inputs": {
+        "crane": "crane",
+        "dream2nix": "dream2nix",
         "fenix": "fenix",
-        "flake-utils": "flake-utils",
+        "flake-parts": "flake-parts",
         "nixpkgs": "nixpkgs",
+        "nixpkgs-swift": "nixpkgs-swift",
         "treefmt-nix": "treefmt-nix"
       }
     },
     "rust-analyzer-src": {
       "flake": false,
       "locked": {
-        "lastModified": 1761849405,
-        "narHash": "sha256-igXdvC+WCUN+3gnfk+ptT7rMmxQuY6WbIg1rXMUN1DM=",
+        "lastModified": 1768224240,
+        "narHash": "sha256-Pp1dDrXKPBUJReZnnDElFyHYn67XTd48zRhToheLjtk=",
         "owner": "rust-lang",
         "repo": "rust-analyzer",
-        "rev": "f7de8ae045a5fe80f1203c5a1c3015b05f7c3550",
+        "rev": "725349602e525df37f377701e001fe8aab807878",
         "type": "github"
       },
       "original": {
@@ -80,18 +198,25 @@
         "type": "github"
       }
     },
-    "systems": {
+    "slimlock": {
+      "inputs": {
+        "nixpkgs": [
+          "dream2nix",
+          "purescript-overlay",
+          "nixpkgs"
+        ]
+      },
       "locked": {
-        "lastModified": 1681028828,
-        "narHash": "sha256-Vy1rq5AaRuLzOxct8nz4T6wlgyUR7zLU309k9mBC768=",
-        "owner": "nix-systems",
-        "repo": "default",
-        "rev": "da67096a3b9bf56a91d16901293e51ba5b49a27e",
+        "lastModified": 1688756706,
+        "narHash": "sha256-xzkkMv3neJJJ89zo3o2ojp7nFeaZc2G0fYwNXNJRFlo=",
+        "owner": "thomashoneyman",
+        "repo": "slimlock",
+        "rev": "cf72723f59e2340d24881fd7bf61cb113b4c407c",
         "type": "github"
       },
       "original": {
-        "owner": "nix-systems",
-        "repo": "default",
+        "owner": "thomashoneyman",
+        "repo": "slimlock",
         "type": "github"
       }
     },
@@ -102,11 +227,11 @@
         ]
       },
       "locked": {
-        "lastModified": 1762938485,
-        "narHash": "sha256-AlEObg0syDl+Spi4LsZIBrjw+snSVU4T8MOeuZJUJjM=",
+        "lastModified": 1768158989,
+        "narHash": "sha256-67vyT1+xClLldnumAzCTBvU0jLZ1YBcf4vANRWP3+Ak=",
         "owner": "numtide",
         "repo": "treefmt-nix",
-        "rev": "5b4ee75aeefd1e2d5a1cc43cf6ba65eba75e83e4",
+        "rev": "e96d59dff5c0d7fddb9d113ba108f03c3ef99eca",
         "type": "github"
       },
       "original": {

flake.nix

@@ -3,118 +3,134 @@
 
   inputs = {
     nixpkgs.url = "github:NixOS/nixpkgs/nixos-unstable";
-    flake-utils.url = "github:numtide/flake-utils";
-    # Provides Rust dev-env integration:
+
+    flake-parts = {
+      url = "github:hercules-ci/flake-parts";
+      inputs.nixpkgs-lib.follows = "nixpkgs";
+    };
+
+    crane.url = "github:ipetkov/crane";
+
     fenix = {
       url = "github:nix-community/fenix";
       inputs.nixpkgs.follows = "nixpkgs";
     };
-    # Provides formatting infrastructure:
+
     treefmt-nix = {
       url = "github:numtide/treefmt-nix";
       inputs.nixpkgs.follows = "nixpkgs";
     };
+
+    dream2nix = {
+      url = "github:nix-community/dream2nix";
+      inputs.nixpkgs.follows = "nixpkgs";
+    };
+
+    # Pinned nixpkgs for swift-format (swift is broken on x86_64-linux in newer nixpkgs)
+    nixpkgs-swift.url = "github:NixOS/nixpkgs/08dacfca559e1d7da38f3cf05f1f45ee9bfd213c";
   };
 
-  # TODO: figure out caching story
-  # nixConfig = {
-  #   # nix community cachix
-  #   extra-trusted-public-keys = "nix-community.cachix.org-1:mB9FSh9qf2dCimDSUo8Zy7bkq5CX+/rkCWyvRCYg3Fs=";
-  #   extra-substituters = "https://nix-community.cachix.org";
-  # };
+  nixConfig = {
+    extra-trusted-public-keys = "exo.cachix.org-1:okq7hl624TBeAR3kV+g39dUFSiaZgLRkLsFBCuJ2NZI=";
+    extra-substituters = "https://exo.cachix.org";
+  };
 
   outputs =
     inputs:
-    let
+    inputs.flake-parts.lib.mkFlake { inherit inputs; } {
       systems = [
         "x86_64-linux"
         "aarch64-darwin"
         "aarch64-linux"
       ];
-      fenixToolchain = system: inputs.fenix.packages.${system}.complete;
-    in
-    inputs.flake-utils.lib.eachSystem systems (
-      system:
-      let
-        pkgs = import inputs.nixpkgs {
-          inherit system;
-          overlays = [ inputs.fenix.overlays.default ];
-        };
-        treefmtEval = inputs.treefmt-nix.lib.evalModule pkgs {
-          projectRootFile = "flake.nix";
-          programs.ruff-format.enable = true;
-          programs.ruff-format.excludes = [ "rust/exo_pyo3_bindings/exo_pyo3_bindings.pyi" ];
-          programs.rustfmt.enable = true;
-          programs.rustfmt.package = (fenixToolchain system).rustfmt;
-          programs.nixpkgs-fmt.enable = true;
-        };
-      in
-      {
-        formatter = treefmtEval.config.build.wrapper;
-        checks.formatting = treefmtEval.config.build.check inputs.self;
-        checks.lint = pkgs.runCommand "lint-check" { } ''
-          export RUFF_CACHE_DIR="$TMPDIR/ruff-cache"
-          ${pkgs.ruff}/bin/ruff check ${inputs.self}/
-          touch $out
-        '';
 
-        devShells.default = pkgs.mkShell {
-          packages =
-            with pkgs;
-            [
-              # PYTHON
-              python313
-              uv
-              ruff
-              basedpyright
+      imports = [
+        inputs.treefmt-nix.flakeModule
+        ./dashboard/parts.nix
+        ./rust/parts.nix
+      ];
 
-              # RUST
-              ((fenixToolchain system).withComponents [
-                "cargo"
-                "rustc"
-                "clippy"
-                "rustfmt"
-                "rust-src"
-              ])
-              rustup # Just here to make RustRover happy
+      perSystem =
+        { config, self', inputs', pkgs, lib, system, ... }:
+        let
+          fenixToolchain = inputs'.fenix.packages.complete;
+          # Use pinned nixpkgs for swift-format (swift is broken on x86_64-linux in newer nixpkgs)
+          pkgsSwift = import inputs.nixpkgs-swift { inherit system; };
+        in
+        {
+          treefmt = {
+            projectRootFile = "flake.nix";
+            programs = {
+              nixpkgs-fmt.enable = true;
+              ruff-format = {
+                enable = true;
+                excludes = [ "rust/exo_pyo3_bindings/exo_pyo3_bindings.pyi" ];
+              };
+              rustfmt = {
+                enable = true;
+                package = config.rust.toolchain;
+              };
+              prettier = {
+                enable = true;
+                includes = [ "*.ts" ];
+              };
+              swift-format = {
+                enable = true;
+                package = pkgsSwift.swiftPackages.swift-format;
+              };
+            };
+          };
 
-              # NIX
-              nixpkgs-fmt
-
-              # SVELTE
-              nodejs
-
-              # MISC
-              just
-              jq
-            ]
-            ++ (pkgs.lib.optionals pkgs.stdenv.isLinux [
-              # IFCONFIG
-              unixtools.ifconfig
-
-              # Build dependencies for Linux
-              pkg-config
-              openssl
-            ])
-            ++ (pkgs.lib.optionals pkgs.stdenv.isDarwin [
-              # MACMON
-              macmon
-            ]);
-
-          shellHook = ''
-            # PYTHON
-            export LD_LIBRARY_PATH="$LD_LIBRARY_PATH:${pkgs.python313}/lib"
-            ${pkgs.lib.optionalString pkgs.stdenv.isLinux ''
-              # Build environment for Linux
-              export PKG_CONFIG_PATH="${pkgs.openssl.dev}/lib/pkgconfig:$PKG_CONFIG_PATH"
-              export LD_LIBRARY_PATH="${pkgs.openssl.out}/lib:$LD_LIBRARY_PATH"
-            ''}
-            echo
-            echo "🍎🍎 Run 'just <recipe>' to get started"
-            just --list
+          checks.lint = pkgs.runCommand "lint-check" { } ''
+            export RUFF_CACHE_DIR="$TMPDIR/ruff-cache"
+            ${pkgs.ruff}/bin/ruff check ${inputs.self}/
+            touch $out
           '';
 
+          devShells.default = with pkgs; pkgs.mkShell {
+            inputsFrom = [ self'.checks.cargo-build ];
+
+            packages =
+              [
+                # FORMATTING
+                config.treefmt.build.wrapper
+
+                # PYTHON
+                python313
+                uv
+                ruff
+                basedpyright
+
+                # RUST
+                config.rust.toolchain
+                maturin
+
+                # NIX
+                nixpkgs-fmt
+
+                # SVELTE
+                nodejs
+
+                # MISC
+                just
+                jq
+              ]
+              ++ lib.optionals stdenv.isLinux [
+                unixtools.ifconfig
+              ]
+              ++ lib.optionals stdenv.isDarwin [
+                macmon
+              ];
+
+            OPENSSL_NO_VENDOR = "1";
+
+            shellHook = ''
+              export LD_LIBRARY_PATH="$LD_LIBRARY_PATH:${python313}/lib"
+              ${lib.optionalString stdenv.isLinux ''
+                export LD_LIBRARY_PATH="${openssl.out}/lib:$LD_LIBRARY_PATH"
+              ''}
+            '';
+          };
         };
-      }
-    );
+    };
 }

pyproject.toml

@@ -8,30 +8,18 @@ dependencies = [
     "aiofiles>=24.1.0",
     "aiohttp>=3.12.14",
     "types-aiofiles>=24.1.0.20250708",
-    "typeguard>=4.4.4",
     "pydantic>=2.11.7",
-    "base58>=2.1.1",
-    "cryptography>=45.0.5",
     "fastapi>=0.116.1",
     "filelock>=3.18.0",
-    "aiosqlite>=0.21.0",
-    "networkx>=3.5",
-    "protobuf>=6.32.0",
-    "rich>=14.1.0",
     "rustworkx>=0.17.1",
-    "sqlmodel>=0.0.24",
-    "sqlalchemy[asyncio]>=2.0.43",
-    "greenlet>=3.2.4",
     "huggingface-hub>=0.33.4",
     "psutil>=7.0.0",
     "loguru>=0.7.3",
-    "textual>=5.3.0",
     "exo_pyo3_bindings", # rust bindings
     "anyio==4.11.0",
-    "bidict>=0.23.1",
-    "mlx>=0.30.1; sys_platform == 'darwin'",
-    "mlx[cpu]>=0.30.1; sys_platform == 'linux'",
-    "mlx-lm>=0.28.3",
+    "mlx==0.30.1; sys_platform == 'darwin'",
+    "mlx[cpu]==0.30.1; sys_platform == 'linux'",
+    "mlx-lm @ git+https://github.com/AlexCheema/mlx-lm.git@fix-transformers-5.0.0rc2",
     "tiktoken>=0.12.0", # required for kimi k2 tokenizer
     "hypercorn>=0.18.0",
     "openai-harmony>=0.0.8",
@@ -41,10 +29,12 @@ dependencies = [
 exo-master = "exo.master.main:main"
 exo-worker = "exo.worker.main:main"
 exo = "exo.main:main"
+exo-rsh = "exo.rsh.client:main"
 
 # dependencies only required for development
 [dependency-groups]
 dev = [
+    "basedpyright>=1.29.0",
     "pyinstaller>=6.17.0",
     "pytest>=8.4.0",
     "pytest-asyncio>=1.0.0",
@@ -110,6 +100,7 @@ root = "src"
 
 # supported platforms for this project
 [tool.uv]
+prerelease = "allow"
 environments = [
     "sys_platform == 'darwin'",
     "sys_platform == 'linux'",

rust/parts.nix

@@ -0,0 +1,159 @@
+{ inputs, ... }:
+{
+  perSystem =
+    { config, self', inputs', pkgs, lib, ... }:
+    let
+      # Fenix nightly toolchain with all components
+      fenixPkgs = inputs'.fenix.packages;
+      rustToolchain = fenixPkgs.complete.withComponents [
+        "cargo"
+        "rustc"
+        "clippy"
+        "rustfmt"
+        "rust-src"
+        "rust-analyzer"
+      ];
+
+      # Crane with fenix toolchain
+      craneLib = (inputs.crane.mkLib pkgs).overrideToolchain rustToolchain;
+
+      # Source filtering - only include rust/ directory and root Cargo files
+      # This ensures changes to Python/docs/etc don't trigger Rust rebuilds
+      src = lib.cleanSourceWith {
+        src = inputs.self;
+        filter =
+          path: type:
+          let
+            baseName = builtins.baseNameOf path;
+            parentDir = builtins.dirOf path;
+            inRustDir =
+              (lib.hasInfix "/rust/" path)
+              || (lib.hasSuffix "/rust" parentDir)
+              || (baseName == "rust" && type == "directory");
+            isRootCargoFile =
+              (baseName == "Cargo.toml" || baseName == "Cargo.lock")
+              && (builtins.dirOf path == toString inputs.self);
+          in
+          isRootCargoFile
+          || (inRustDir && (craneLib.filterCargoSources path type || lib.hasSuffix ".toml" path || lib.hasSuffix ".md" path));
+      };
+
+      # Common arguments for all Rust builds
+      commonArgs = {
+        inherit src;
+        pname = "exo-rust";
+        version = "0.0.1";
+        strictDeps = true;
+
+        nativeBuildInputs = [
+          pkgs.pkg-config
+          pkgs.python313 # Required for pyo3-build-config
+        ];
+
+        buildInputs = [
+          pkgs.openssl
+          pkgs.python313 # Required for pyo3 tests
+        ];
+
+        OPENSSL_NO_VENDOR = "1";
+
+        # Required for pyo3 tests to find libpython
+        LD_LIBRARY_PATH = lib.makeLibraryPath [ pkgs.python313 ];
+      };
+
+      # Build dependencies once for caching
+      cargoArtifacts = craneLib.buildDepsOnly (
+        commonArgs
+        // {
+          cargoExtraArgs = "--workspace";
+        }
+      );
+    in
+    {
+      # Export toolchain for use in treefmt and devShell
+      options.rust = {
+        toolchain = lib.mkOption {
+          type = lib.types.package;
+          default = rustToolchain;
+          description = "The Rust toolchain to use";
+        };
+      };
+
+      config = {
+        packages = {
+          # The system_custodian binary
+          system_custodian = craneLib.buildPackage (
+            commonArgs
+            // {
+              inherit cargoArtifacts;
+              cargoExtraArgs = "-p system_custodian";
+
+              meta = {
+                description = "System custodian daemon for exo";
+                mainProgram = "system_custodian";
+              };
+            }
+          );
+
+          # Python bindings wheel via maturin
+          exo_pyo3_bindings = craneLib.buildPackage (
+            commonArgs
+            // {
+              inherit cargoArtifacts;
+              pname = "exo_pyo3_bindings";
+
+              nativeBuildInputs = commonArgs.nativeBuildInputs ++ [
+                pkgs.maturin
+              ];
+
+              buildPhaseCargoCommand = ''
+                maturin build \
+                  --release \
+                  --manylinux off \
+                  --manifest-path rust/exo_pyo3_bindings/Cargo.toml \
+                  --features "pyo3/extension-module,pyo3/experimental-async" \
+                  --interpreter ${pkgs.python313}/bin/python \
+                  --out dist
+              '';
+
+              # Don't use crane's default install behavior
+              doNotPostBuildInstallCargoBinaries = true;
+
+              installPhaseCommand = ''
+                mkdir -p $out
+                cp dist/*.whl $out/
+              '';
+            }
+          );
+        };
+
+        checks = {
+          # Full workspace build (all crates)
+          cargo-build = craneLib.buildPackage (
+            commonArgs
+            // {
+              inherit cargoArtifacts;
+              cargoExtraArgs = "--workspace";
+            }
+          );
+          # Run tests with nextest
+          cargo-nextest = craneLib.cargoNextest (
+            commonArgs
+            // {
+              inherit cargoArtifacts;
+              cargoExtraArgs = "--workspace";
+            }
+          );
+
+          # Build documentation
+          cargo-doc = craneLib.cargoDoc (
+            commonArgs
+            // {
+              inherit cargoArtifacts;
+              cargoExtraArgs = "--workspace";
+            }
+          );
+        };
+      };
+    };
+}

src/exo/main.py

@@ -1,6 +1,7 @@
 import argparse
 import multiprocessing as mp
 import os
+import resource
 import signal
 from dataclasses import dataclass, field
 from typing import Self
@@ -195,6 +196,8 @@ class Node:
 
 def main():
     args = Args.parse()
+    soft, hard = resource.getrlimit(resource.RLIMIT_NOFILE)
+    resource.setrlimit(resource.RLIMIT_NOFILE, (max(soft, 65535), hard))
 
     mp.set_start_method("spawn")
     # TODO: Refactor the current verbosity system

src/exo/master/api.py

@@ -1,6 +1,8 @@
+import asyncio
+import os
 import time
 from collections.abc import AsyncGenerator
-from typing import cast
+from typing import Any, Optional, cast
 
 import anyio
 from anyio import create_task_group
@@ -19,6 +21,7 @@ from openai_harmony import (  # pyright: ignore[reportMissingTypeStubs]
     StreamableParser,
     load_harmony_encoding,
 )
+from pydantic import BaseModel
 
 from exo.master.placement import place_instance as get_instance_placements
 from exo.shared.apply import apply
@@ -51,7 +54,9 @@ from exo.shared.types.commands import (
     CreateInstance,
     DeleteInstance,
     ForwarderCommand,
+    LaunchFLASH,
     PlaceInstance,
+    StopFLASH,
     TaskFinished,
 )
 from exo.shared.types.common import CommandId, NodeId, SessionId
@@ -60,7 +65,12 @@ from exo.shared.types.memory import Memory
 from exo.shared.types.models import ModelId, ModelMetadata
 from exo.shared.types.state import State
 from exo.shared.types.tasks import ChatCompletionTaskParams
-from exo.shared.types.worker.instances import Instance, InstanceId, InstanceMeta
+from exo.shared.types.worker.instances import (
+    FLASHInstance,
+    Instance,
+    InstanceId,
+    InstanceMeta,
+)
 from exo.shared.types.worker.shards import Sharding
 from exo.utils.banner import print_startup_banner
 from exo.utils.channels import Receiver, Sender, channel
@@ -70,6 +80,22 @@ from exo.utils.event_buffer import OrderedBuffer
 encoding = load_harmony_encoding(HarmonyEncodingName.HARMONY_GPT_OSS)
 
 
+class ExecuteRequest(BaseModel):
+    """Request to execute a command."""
+
+    command: list[str]
+    cwd: Optional[str] = None
+    env: Optional[dict[str, str]] = None
+
+
+class ExecuteResponse(BaseModel):
+    """Response from command execution."""
+
+    exit_code: int
+    stdout: str
+    stderr: str
+
+
 def chunk_to_response(
     chunk: TokenChunk, command_id: CommandId
 ) -> ChatCompletionResponse:
@@ -178,6 +204,12 @@ class API:
         self.app.post("/bench/chat/completions")(self.bench_chat_completions)
         self.app.get("/state")(lambda: self.state)
         self.app.get("/events")(lambda: self._event_log)
+        # FLASH simulation endpoints
+        self.app.post("/flash/launch")(self.launch_flash)
+        self.app.delete("/flash/{instance_id}")(self.stop_flash)
+        self.app.get("/flash/instances")(self.list_flash_instances)
+        # Remote execution endpoint (used by exo-rsh for MPI)
+        self.app.post("/execute")(self.execute)
 
     async def place_instance(self, payload: PlaceInstanceParams):
         command = PlaceInstance(
@@ -622,6 +654,145 @@ class API:
             ]
         )
 
+    async def launch_flash(
+        self,
+        simulation_name: str,
+        flash_executable_path: str,
+        working_directory: str,
+        parameter_file_path: str = "",
+        ranks_per_node: int = 1,
+        min_nodes: int = 1,
+        hosts: str = "",
+    ) -> dict[str, str]:
+        """Launch a FLASH MPI simulation across the cluster.
+
+        Args:
+            hosts: Optional comma-separated hostnames (e.g., "s14,james21-1").
+                   If not provided, IPs are discovered from topology edges.
+        """
+        command = LaunchFLASH(
+            simulation_name=simulation_name,
+            flash_executable_path=flash_executable_path,
+            parameter_file_path=parameter_file_path,
+            working_directory=working_directory,
+            ranks_per_node=ranks_per_node,
+            min_nodes=min_nodes,
+            hosts=hosts,
+        )
+        await self._send(command)
+
+        return {
+            "message": "FLASH launch command received",
+            "command_id": str(command.command_id),
+            "simulation_name": simulation_name,
+        }
+
+    async def stop_flash(self, instance_id: InstanceId) -> dict[str, str]:
+        """Stop a running FLASH simulation."""
+        if instance_id not in self.state.instances:
+            raise HTTPException(status_code=404, detail="Instance not found")
+
+        instance = self.state.instances[instance_id]
+        if not isinstance(instance, FLASHInstance):
+            raise HTTPException(
+                status_code=400, detail="Instance is not a FLASH simulation"
+            )
+
+        command = StopFLASH(instance_id=instance_id)
+        await self._send(command)
+
+        return {
+            "message": "Stop command received",
+            "command_id": str(command.command_id),
+            "instance_id": str(instance_id),
+        }
+
+    async def list_flash_instances(self) -> list[dict[str, Any]]:
+        """List all FLASH simulation instances."""
+        flash_instances: list[dict[str, Any]] = []
+        for instance_id, instance in self.state.instances.items():
+            if isinstance(instance, FLASHInstance):
+                # Get runner statuses for this instance
+                runner_statuses: dict[str, str | None] = {}
+                for (
+                    node_id,
+                    runner_id,
+                ) in instance.shard_assignments.node_to_runner.items():
+                    runner_status = self.state.runners.get(runner_id)
+                    runner_statuses[str(node_id)] = (
+                        str(runner_status) if runner_status else None
+                    )
+
+                flash_instances.append(
+                    {
+                        "instance_id": str(instance_id),
+                        "simulation_name": instance.simulation_name,
+                        "total_ranks": instance.total_ranks,
+                        "working_directory": instance.working_directory,
+                        "runner_statuses": runner_statuses,
+                    }
+                )
+        return flash_instances
+
+    async def execute(self, request: ExecuteRequest) -> ExecuteResponse:
+        """Execute a command locally. Used by exo-rsh for MPI remote execution."""
+        cmd_str = " ".join(request.command)
+        logger.info(f"Executing: {cmd_str}")
+
+        try:
+            # Build environment
+            env = os.environ.copy()
+            if request.env:
+                env.update(request.env)
+
+            # Check if command contains shell metacharacters
+            # If so, run through shell. mpirun sends complex commands like:
+            # "VAR=value;export VAR;/path/to/prted --args"
+            needs_shell = any(c in cmd_str for c in ";|&$`")
+
+            if needs_shell:
+                process = await asyncio.create_subprocess_shell(
+                    cmd_str,
+                    stdout=asyncio.subprocess.PIPE,
+                    stderr=asyncio.subprocess.PIPE,
+                    cwd=request.cwd,
+                    env=env,
+                )
+            else:
+                process = await asyncio.create_subprocess_exec(
+                    *request.command,
+                    stdout=asyncio.subprocess.PIPE,
+                    stderr=asyncio.subprocess.PIPE,
+                    cwd=request.cwd,
+                    env=env,
+                )
+
+            stdout, stderr = await process.communicate()
+            exit_code = process.returncode or 0
+
+            logger.info(f"Command completed with exit code {exit_code}")
+
+            return ExecuteResponse(
+                exit_code=exit_code,
+                stdout=stdout.decode("utf-8", errors="replace"),
+                stderr=stderr.decode("utf-8", errors="replace"),
+            )
+
+        except FileNotFoundError:
+            logger.error(f"Command not found: {request.command[0]}")
+            return ExecuteResponse(
+                exit_code=127,
+                stdout="",
+                stderr=f"Command not found: {request.command[0]}",
+            )
+        except Exception as e:
+            logger.error(f"Execution error: {e}")
+            return ExecuteResponse(
+                exit_code=1,
+                stdout="",
+                stderr=str(e),
+            )
+
     async def run(self):
         cfg = Config()
         cfg.bind = f"0.0.0.0:{self.port}"

src/exo/master/main.py

@@ -8,6 +8,7 @@ from exo.master.placement import (
     add_instance_to_placements,
     delete_instance,
     get_transition_events,
+    place_flash_instance,
     place_instance,
 )
 from exo.shared.apply import apply
@@ -16,8 +17,10 @@ from exo.shared.types.commands import (
     CreateInstance,
     DeleteInstance,
     ForwarderCommand,
+    LaunchFLASH,
     PlaceInstance,
     RequestEventLog,
+    StopFLASH,
     TaskFinished,
     TestCommand,
 )
@@ -173,6 +176,26 @@ class Master:
                                 self.state.instances, placement
                             )
                             generated_events.extend(transition_events)
+                        case LaunchFLASH():
+                            placement = place_flash_instance(
+                                command,
+                                self.state.topology,
+                                self.state.instances,
+                            )
+                            transition_events = get_transition_events(
+                                self.state.instances, placement
+                            )
+                            generated_events.extend(transition_events)
+                        case StopFLASH():
+                            # Reuse delete_instance logic to stop FLASH simulation
+                            placement = delete_instance(
+                                DeleteInstance(instance_id=command.instance_id),
+                                self.state.instances,
+                            )
+                            transition_events = get_transition_events(
+                                self.state.instances, placement
+                            )
+                            generated_events.extend(transition_events)
                         case TaskFinished():
                             generated_events.append(
                                 TaskDeleted(

src/exo/master/placement.py

@@ -17,20 +17,24 @@ from exo.shared.topology import Topology
 from exo.shared.types.commands import (
     CreateInstance,
     DeleteInstance,
+    LaunchFLASH,
     PlaceInstance,
 )
+from exo.shared.types.common import Host, NodeId
 from exo.shared.types.events import Event, InstanceCreated, InstanceDeleted
 from exo.shared.types.memory import Memory
-from exo.shared.types.models import ModelId
+from exo.shared.types.models import ModelId, ModelMetadata
 from exo.shared.types.topology import NodeInfo
 from exo.shared.types.worker.instances import (
+    FLASHInstance,
     Instance,
     InstanceId,
     InstanceMeta,
     MlxJacclInstance,
     MlxRingInstance,
 )
-from exo.shared.types.worker.shards import Sharding
+from exo.shared.types.worker.runners import RunnerId, ShardAssignments
+from exo.shared.types.worker.shards import PipelineShardMetadata, Sharding
 
 
 def random_ephemeral_port() -> int:
@@ -165,6 +169,9 @@ def place_instance(
                 hosts_by_node=hosts_by_node,
                 ephemeral_port=ephemeral_port,
             )
+        case InstanceMeta.FLASH:
+            # FLASH instances are handled by place_flash_instance()
+            raise ValueError("FLASH instances should use place_flash_instance()")
 
     return target_instances
 
@@ -180,6 +187,148 @@ def delete_instance(
     raise ValueError(f"Instance {command.instance_id} not found")
 
 
+def place_flash_instance(
+    command: LaunchFLASH,
+    topology: Topology,
+    current_instances: Mapping[InstanceId, Instance],
+) -> dict[InstanceId, Instance]:
+    """Place a FLASH simulation instance across available nodes.
+
+    Unlike MLX instances which use ring/JACCL topology for tensor parallelism,
+    FLASH instances use MPI for communication. We just need to provide the
+    node IPs so the runner can generate an MPI hostfile.
+    """
+    instance_id = InstanceId()
+    target_instances = dict(deepcopy(current_instances))
+
+    all_nodes = list(topology.list_nodes())
+
+    if len(all_nodes) < command.min_nodes:
+        raise ValueError(
+            f"Not enough nodes: need {command.min_nodes}, have {len(all_nodes)}"
+        )
+
+    # Select nodes (take the first min_nodes)
+    selected_nodes = all_nodes[: command.min_nodes]
+
+    logger.info(
+        f"Placing FLASH instance '{command.simulation_name}' on {len(selected_nodes)} nodes"
+    )
+
+    # Build shard assignments (one runner per node for FLASH)
+    runner_to_shard: dict[RunnerId, PipelineShardMetadata] = {}
+    node_to_runner: dict[NodeId, RunnerId] = {}
+
+    # Create a dummy ModelMetadata for FLASH (required by ShardMetadata interface)
+    flash_model_meta = ModelMetadata(
+        model_id=ModelId(command.simulation_name),
+        pretty_name=f"FLASH: {command.simulation_name}",
+        storage_size=Memory(in_bytes=0),
+        n_layers=1,
+        hidden_size=1,
+        supports_tensor=False,
+    )
+
+    for i, node_info in enumerate(selected_nodes):
+        runner_id = RunnerId()
+        node_to_runner[node_info.node_id] = runner_id
+        runner_to_shard[runner_id] = PipelineShardMetadata(
+            device_rank=i,
+            world_size=len(selected_nodes),
+            model_meta=flash_model_meta,
+            start_layer=0,
+            end_layer=1,
+            n_layers=1,
+        )
+
+    shard_assignments = ShardAssignments(
+        model_id=ModelId(command.simulation_name),
+        runner_to_shard=runner_to_shard,
+        node_to_runner=node_to_runner,
+    )
+
+    # Build hosts_by_node - get hostnames/IPs for MPI hostfile generation
+    hosts_by_node: dict[NodeId, list[Host]] = {}
+
+    # If explicit hosts are provided, use them directly
+    if command.hosts:
+        explicit_hosts = [h.strip() for h in command.hosts.split(",") if h.strip()]
+        logger.info(f"FLASH placement: explicit hosts provided: {explicit_hosts}")
+        for i, node_info in enumerate(selected_nodes):
+            if i < len(explicit_hosts):
+                hosts_by_node[node_info.node_id] = [Host(ip=explicit_hosts[i], port=0)]
+                logger.info(
+                    f"FLASH placement: node {node_info.node_id} (rank {i}) -> IP {explicit_hosts[i]}"
+                )
+            else:
+                logger.warning(
+                    f"Not enough hosts provided for node {i}, using localhost"
+                )
+                hosts_by_node[node_info.node_id] = [Host(ip="127.0.0.1", port=0)]
+        logger.info(
+            f"FLASH placement: coordinator will be rank 0 at IP {explicit_hosts[0]}"
+        )
+    else:
+        # Try to get IPs from topology edges
+        for node_info in selected_nodes:
+            node_hosts: list[Host] = []
+
+            # Get IP from outgoing edges (connections to other nodes via mDNS discovery)
+            for _, edge_data in topology.out_edges(node_info.node_id):
+                if hasattr(edge_data, "send_back_multiaddr"):
+                    # Extract IP from multiaddr like /ip4/192.168.1.100/tcp/52415
+                    multiaddr = str(edge_data.send_back_multiaddr)
+                    if "/ip4/" in multiaddr:
+                        parts = multiaddr.split("/")
+                        try:
+                            ip_idx = parts.index("ip4") + 1
+                            ip = parts[ip_idx]
+                            # Skip link-local and localhost addresses
+                            if not ip.startswith("169.254.") and not ip.startswith(
+                                "127."
+                            ):
+                                node_hosts.append(Host(ip=ip, port=0))
+                                break
+                        except (ValueError, IndexError):
+                            pass
+
+            # Last resort: use localhost (will only work for single-node)
+            if not node_hosts:
+                logger.warning(
+                    f"Could not determine IP for node {node_info.node_id}, using localhost"
+                )
+                node_hosts.append(Host(ip="127.0.0.1", port=0))
+
+            hosts_by_node[node_info.node_id] = node_hosts
+
+    total_ranks = len(selected_nodes) * command.ranks_per_node
+
+    # Determine coordinator IP - first node's first host IP
+    first_node_id: NodeId = next(iter(hosts_by_node.keys()))
+    coordinator_ip: str = (
+        hosts_by_node[first_node_id][0].ip
+        if hosts_by_node[first_node_id]
+        else "127.0.0.1"
+    )
+
+    target_instances[instance_id] = FLASHInstance(
+        instance_id=instance_id,
+        shard_assignments=shard_assignments,
+        hosts_by_node=hosts_by_node,
+        flash_executable_path=command.flash_executable_path,
+        parameter_file_path=command.parameter_file_path,
+        working_directory=command.working_directory,
+        ranks_per_node=command.ranks_per_node,
+        total_ranks=total_ranks,
+        simulation_name=command.simulation_name,
+        coordinator_ip=coordinator_ip,
+    )
+
+    logger.info(f"Created FLASH instance {instance_id} with {total_ranks} total ranks")
+
+    return target_instances
+
+
 def get_transition_events(
     current_instances: Mapping[InstanceId, Instance],
     target_instances: Mapping[InstanceId, Instance],

src/exo/rsh/__init__.py

@@ -0,0 +1,13 @@
+"""Exo RSH - Remote Shell for MPI without SSH.
+
+This module provides a remote execution mechanism that allows mpirun to spawn
+processes on remote nodes without requiring SSH setup. It works by:
+
+1. Each Exo node runs an API server on port 52415 with an /execute endpoint
+2. The exo-rsh script acts as a drop-in replacement for ssh
+3. When mpirun calls "exo-rsh hostname command", it HTTP POSTs to the target's /execute
+4. The target executes the command and returns output
+
+Usage:
+    mpirun --mca plm_rsh_agent exo-rsh -np 4 --hostfile hosts.txt ./program
+"""

src/exo/rsh/client.py

@@ -0,0 +1,101 @@
+#!/usr/bin/env python3
+"""exo-rsh - Remote shell client for MPI.
+
+This script is called by mpirun as a replacement for ssh.
+Usage: exo-rsh [ssh-options...] hostname command [args...]
+
+It connects to the target node's Exo API (port 52415) and executes the command.
+"""
+
+import json
+import socket
+import sys
+from typing import Any, cast
+from urllib.error import URLError
+from urllib.request import Request, urlopen
+
+# Use the same port as Exo's API server
+EXO_API_PORT = 52415
+
+
+def resolve_hostname(hostname: str) -> str:
+    """Resolve hostname to IP address."""
+    try:
+        return socket.gethostbyname(hostname)
+    except socket.gaierror:
+        # If resolution fails, try using the hostname directly
+        return hostname
+
+
+def main():
+    # Parse arguments - mpirun calls us like: exo-rsh [options] hostname command [args...]
+    # SSH options we might see: -x (disable X11), -o options, etc.
+    args = sys.argv[1:]
+
+    # Skip SSH-style options
+    hostname = None
+    command_start = 0
+
+    i = 0
+    while i < len(args):
+        arg = args[i]
+        if arg.startswith("-"):
+            # Skip option and its value if needed
+            if arg in ("-o", "-i", "-l", "-p", "-F"):
+                i += 2  # Skip option and its argument
+                continue
+            i += 1
+            continue
+        else:
+            # First non-option is the hostname
+            hostname = arg
+            command_start = i + 1
+            break
+        i += 1
+
+    if hostname is None or command_start >= len(args):
+        print("Usage: exo-rsh [options] hostname command [args...]", file=sys.stderr)
+        sys.exit(1)
+
+    command = args[command_start:]
+
+    # Resolve hostname to IP
+    ip = resolve_hostname(hostname)
+
+    # Make request to Exo API
+    url = f"http://{ip}:{EXO_API_PORT}/execute"
+    data = json.dumps({"command": command}).encode("utf-8")
+
+    try:
+        req = Request(url, data=data, headers={"Content-Type": "application/json"})
+        with urlopen(req, timeout=300) as response:  # pyright: ignore[reportAny]
+            response_body: bytes = cast(bytes, response.read())  # pyright: ignore[reportAny]
+            result: dict[str, Any] = json.loads(response_body.decode("utf-8"))  # pyright: ignore[reportAny]
+
+        # Output stdout/stderr
+        stdout: str = cast(str, result.get("stdout", ""))
+        stderr: str = cast(str, result.get("stderr", ""))
+        exit_code: int = cast(int, result.get("exit_code", 0))
+
+        if stdout:
+            sys.stdout.write(stdout)
+            sys.stdout.flush()
+        if stderr:
+            sys.stderr.write(stderr)
+            sys.stderr.flush()
+
+        sys.exit(exit_code)
+
+    except URLError as e:
+        print(
+            f"exo-rsh: Failed to connect to {hostname}:{EXO_API_PORT}: {e}",
+            file=sys.stderr,
+        )
+        sys.exit(255)
+    except Exception as e:
+        print(f"exo-rsh: Error: {e}", file=sys.stderr)
+        sys.exit(1)
+
+
+if __name__ == "__main__":
+    main()

src/exo/shared/models/model_cards.py

@@ -82,6 +82,7 @@ MODEL_CARDS: dict[str, ModelCard] = {
     #         storage_size=Memory.from_kb(754706307),
     #         n_layers=61,
     #         hidden_size=7168,
+    #         supports_tensor=True,
     #     ),
     # ),
     # "deepseek-v3.2-4bit": ModelCard(
@@ -96,6 +97,7 @@ MODEL_CARDS: dict[str, ModelCard] = {
     #         storage_size=Memory.from_kb(754706307 // 2),  # TODO !!!!!
     #         n_layers=61,
     #         hidden_size=7168,
+    #         supports_tensor=True,
     #     ),
     # ),
     # deepseek r1
@@ -554,6 +556,81 @@ MODEL_CARDS: dict[str, ModelCard] = {
             supports_tensor=True,
         ),
     ),
+    "glm-4.7-4bit": ModelCard(
+        short_id="glm-4.7-4bit",
+        model_id=ModelId("mlx-community/GLM-4.7-4bit"),
+        name="GLM 4.7 4bit",
+        description="GLM 4.7 4bit",
+        tags=[],
+        metadata=ModelMetadata(
+            model_id=ModelId("mlx-community/GLM-4.7-4bit"),
+            pretty_name="GLM 4.7 4bit",
+            storage_size=Memory.from_bytes(198556925568),
+            n_layers=91,
+            hidden_size=5120,
+            supports_tensor=True,
+        ),
+    ),
+    "glm-4.7-6bit": ModelCard(
+        short_id="glm-4.7-6bit",
+        model_id=ModelId("mlx-community/GLM-4.7-6bit"),
+        name="GLM 4.7 6bit",
+        description="GLM 4.7 6bit",
+        tags=[],
+        metadata=ModelMetadata(
+            model_id=ModelId("mlx-community/GLM-4.7-6bit"),
+            pretty_name="GLM 4.7 6bit",
+            storage_size=Memory.from_bytes(286737579648),
+            n_layers=91,
+            hidden_size=5120,
+            supports_tensor=True,
+        ),
+    ),
+    "glm-4.7-8bit-gs32": ModelCard(
+        short_id="glm-4.7-8bit-gs32",
+        model_id=ModelId("mlx-community/GLM-4.7-8bit-gs32"),
+        name="GLM 4.7 8bit (gs32)",
+        description="GLM 4.7 8bit (gs32)",
+        tags=[],
+        metadata=ModelMetadata(
+            model_id=ModelId("mlx-community/GLM-4.7-8bit-gs32"),
+            pretty_name="GLM 4.7 8bit (gs32)",
+            storage_size=Memory.from_bytes(396963397248),
+            n_layers=91,
+            hidden_size=5120,
+            supports_tensor=True,
+        ),
+    ),
+    "minimax-m2.1-8bit": ModelCard(
+        short_id="minimax-m2.1-8bit",
+        model_id=ModelId("mlx-community/MiniMax-M2.1-8bit"),
+        name="MiniMax M2.1 8bit",
+        description="MiniMax M2.1 8bit",
+        tags=[],
+        metadata=ModelMetadata(
+            model_id=ModelId("mlx-community/MiniMax-M2.1-8bit"),
+            pretty_name="MiniMax M2.1 8bit",
+            storage_size=Memory.from_bytes(242986745856),
+            n_layers=61,
+            hidden_size=3072,
+            supports_tensor=True,
+        ),
+    ),
+    "minimax-m2.1-3bit": ModelCard(
+        short_id="minimax-m2.1-3bit",
+        model_id=ModelId("mlx-community/MiniMax-M2.1-3bit"),
+        name="MiniMax M2.1 3bit",
+        description="MiniMax M2.1 3bit",
+        tags=[],
+        metadata=ModelMetadata(
+            model_id=ModelId("mlx-community/MiniMax-M2.1-3bit"),
+            pretty_name="MiniMax M2.1 3bit",
+            storage_size=Memory.from_bytes(100086644736),
+            n_layers=61,
+            hidden_size=3072,
+            supports_tensor=True,
+        ),
+    ),
     # "devstral-2-123b-instruct-2512-8bit": ModelCard(
     #     short_id="devstral-2-123b-instruct-2512-8bit",
     #     model_id=ModelId("mlx-community/Devstral-2-123B-Instruct-2512-8bit"),

src/exo/shared/tests/test_apply/test_apply_node_download.py

@@ -2,6 +2,7 @@ from exo.shared.apply import apply_node_download_progress
 from exo.shared.tests.conftest import get_pipeline_shard_metadata
 from exo.shared.types.common import NodeId
 from exo.shared.types.events import NodeDownloadProgress
+from exo.shared.types.memory import Memory
 from exo.shared.types.state import State
 from exo.shared.types.worker.downloads import DownloadCompleted
 from exo.worker.tests.constants import MODEL_A_ID, MODEL_B_ID
@@ -13,6 +14,7 @@ def test_apply_node_download_progress():
     event = DownloadCompleted(
         node_id=NodeId("node-1"),
         shard_metadata=shard1,
+        total_bytes=Memory(),
     )
 
     new_state = apply_node_download_progress(
@@ -28,10 +30,12 @@ def test_apply_two_node_download_progress():
     event1 = DownloadCompleted(
         node_id=NodeId("node-1"),
         shard_metadata=shard1,
+        total_bytes=Memory(),
     )
     event2 = DownloadCompleted(
         node_id=NodeId("node-1"),
         shard_metadata=shard2,
+        total_bytes=Memory(),
     )
     state = State(downloads={NodeId("node-1"): [event1]})
 

src/exo/shared/types/commands.py

@@ -35,6 +35,26 @@ class DeleteInstance(BaseCommand):
     instance_id: InstanceId
 
 
+class LaunchFLASH(BaseCommand):
+    """Command to launch a FLASH MPI simulation."""
+
+    simulation_name: str
+    flash_executable_path: str
+    parameter_file_path: str
+    working_directory: str
+    ranks_per_node: int = 1
+    min_nodes: int = 1
+    # Optional: explicit hostnames for MPI (e.g., "s14,james21-1")
+    # Used when topology edges don't contain IP addresses
+    hosts: str = ""
+
+
+class StopFLASH(BaseCommand):
+    """Command to stop a running FLASH simulation."""
+
+    instance_id: InstanceId
+
+
 class TaskFinished(BaseCommand):
     finished_command_id: CommandId
 
@@ -50,6 +70,8 @@ Command = (
     | PlaceInstance
     | CreateInstance
     | DeleteInstance
+    | LaunchFLASH
+    | StopFLASH
     | TaskFinished
 )
 

src/exo/shared/types/worker/downloads.py

@@ -28,7 +28,7 @@ class DownloadPending(BaseDownloadProgress):
 
 
 class DownloadCompleted(BaseDownloadProgress):
-    pass
+    total_bytes: Memory
 
 
 class DownloadFailed(BaseDownloadProgress):

src/exo/shared/types/worker/instances.py

@@ -14,6 +14,7 @@ class InstanceId(Id):
 class InstanceMeta(str, Enum):
     MlxRing = "MlxRing"
     MlxJaccl = "MlxJaccl"
+    FLASH = "FLASH"
 
 
 class BaseInstance(TaggedModel):
@@ -34,8 +35,27 @@ class MlxJacclInstance(BaseInstance):
     jaccl_coordinators: dict[NodeId, str]
 
 
+class FLASHInstance(BaseInstance):
+    """Instance for FLASH MPI simulation.
+
+    Unlike MLX instances which do tensor parallelism, FLASH instances
+    coordinate MPI processes across nodes. Each node runs one or more
+    MPI ranks of the FLASH simulation.
+    """
+
+    hosts_by_node: dict[NodeId, list[Host]]
+    flash_executable_path: str
+    parameter_file_path: str
+    working_directory: str
+    ranks_per_node: int = 1
+    total_ranks: int
+    simulation_name: str
+    coordinator_ip: str
+    network_interface: str = "en0"  # Network interface for MPI (e.g., en0, eth0)
+
+
 # TODO: Single node instance
-Instance = MlxRingInstance | MlxJacclInstance
+Instance = MlxRingInstance | MlxJacclInstance | FLASHInstance
 
 
 class BoundInstance(CamelCaseModel):

src/exo/worker/engines/mlx/auto_parallel.py

@@ -10,18 +10,24 @@ from mlx.nn.layers.distributed import (
     shard_linear,
     sum_gradients,
 )
-from mlx_lm.models.cache import (
-    _BaseCache,  # pyright: ignore[reportPrivateUsage]
-)
 from mlx_lm.models.deepseek_v3 import DeepseekV3MLP
 from mlx_lm.models.deepseek_v3 import Model as DeepseekV3Model
+from mlx_lm.models.deepseek_v32 import DeepseekV32MLP
+from mlx_lm.models.deepseek_v32 import Model as DeepseekV32Model
+from mlx_lm.models.glm4_moe import Model as Glm4MoeModel
+from mlx_lm.models.glm4_moe import MoE
+from mlx_lm.models.gpt_oss import GptOssMoeModel
+from mlx_lm.models.gpt_oss import Model as GptOssModel
 from mlx_lm.models.llama import Model as LlamaModel
+from mlx_lm.models.minimax import Model as MiniMaxModel
+from mlx_lm.models.ministral3 import Model as Ministral3Model
 from mlx_lm.models.qwen3_moe import Model as Qwen3MoeModel
 from mlx_lm.models.qwen3_moe import Qwen3MoeSparseMoeBlock
+from mlx_lm.models.qwen3_next import Model as Qwen3NextModel
+from mlx_lm.models.qwen3_next import Qwen3NextSparseMoeBlock
 
-from exo.shared.types.worker.shards import (
-    PipelineShardMetadata,
-)
+from exo.shared.logging import logger
+from exo.shared.types.worker.shards import PipelineShardMetadata
 
 
 class _LayerCallable(Protocol):
@@ -91,8 +97,6 @@ class PipelineLastLayer(CustomMlxLayer):
             x, *args, **kwargs
         ).arguments.get("cache", None)
 
-        assert cache is None or issubclass(type(cache), _BaseCache)  # type: ignore
-
         output: mx.array = self.original_layer(x, *args, **kwargs)
 
         if self.r != self.s - 1:
@@ -100,7 +104,6 @@ class PipelineLastLayer(CustomMlxLayer):
                 output, (self.r + 1) % self.s, group=self.group
             )
             if cache is not None:
-                # This change happened upstream - check out mlx github somewhere??
                 cache.keys = mx.depends(cache.keys, output)  # type: ignore[reportUnknownMemberType]
 
         output = mx.distributed.all_gather(output, group=self.group)[-output.shape[0] :]
@@ -132,24 +135,6 @@ def _get_layers(inner_model_instance: nn.Module) -> list[_LayerCallable]:
     return layers
 
 
-def _set_layers(model: nn.Module, layers: list[_LayerCallable]) -> None:
-    inner_model_instance = _inner_model(model)
-    if hasattr(inner_model_instance, "layers"):
-        inner_model_instance.layers = layers
-
-        # Update DeepSeek V3 specific parameters when layers are shrunk
-        if isinstance(model, DeepseekV3Model) and hasattr(
-            inner_model_instance, "num_layers"
-        ):
-            inner_model_instance.start_idx = 0
-            inner_model_instance.end_idx = len(layers)
-            inner_model_instance.num_layers = len(layers)
-    elif hasattr(inner_model_instance, "h"):
-        inner_model_instance.h = layers
-    else:
-        raise ValueError("Model must have either a 'layers' or 'h' attribute")
-
-
 def pipeline_auto_parallel(
     model: nn.Module,
     group: mx.distributed.Group,
@@ -165,8 +150,7 @@ def pipeline_auto_parallel(
     """
     inner_model_instance: nn.Module = _inner_model(model)
 
-    # Handle both model.layers and model.h cases
-    layers: list[_LayerCallable] = _get_layers(inner_model_instance)
+    layers = _get_layers(inner_model_instance)
 
     start_layer, end_layer = model_shard_meta.start_layer, model_shard_meta.end_layer
     device_rank, world_size = model_shard_meta.device_rank, model_shard_meta.world_size
@@ -180,6 +164,17 @@ def pipeline_auto_parallel(
         group=group,
     )
 
+    if isinstance(inner_model_instance, GptOssMoeModel):
+        inner_model_instance.layer_types = inner_model_instance.layer_types[  # type: ignore
+            start_layer:end_layer
+        ]
+        inner_model_instance.swa_idx = inner_model_instance.layer_types.index(  # type: ignore
+            "sliding_attention"
+        )
+        inner_model_instance.ga_idx = inner_model_instance.layer_types.index(  # type: ignore
+            "full_attention"
+        )
+
     _set_layers(model, layers)
 
     assert isinstance(layers, list), (
@@ -204,18 +199,44 @@ def tensor_auto_parallel(
         group=group,
     )
 
+    segments: int = 1
+
+    def _all_to_sharded(path: str, weight: mx.array):
+        if path.endswith("bias"):
+            logger.info(f"Sharding bias for {path} - all to sharded")
+            return weight.ndim - 1, segments
+        return max(weight.ndim - 2, 0), segments
+
     all_to_sharded_linear_in_place = partial(
         shard_inplace,
-        sharding="all-to-sharded",
-        group=group,
-    )
-    sharded_to_all_linear_in_place = partial(
-        shard_inplace,
-        sharding="sharded-to-all",
+        sharding=_all_to_sharded,  # type: ignore
         group=group,
     )
 
-    if isinstance(model, LlamaModel):
+    n = group.size()
+
+    def _sharded_to_all(path: str, weight: mx.array):
+        if path.endswith("bias"):
+            logger.info(f"Sharding bias for {path} - sharded to all")
+            weight /= n
+            return None
+        return -1, segments
+
+    sharded_to_all_linear_in_place = partial(
+        shard_inplace,
+        sharding=_sharded_to_all,  # type: ignore
+        group=group,
+    )
+
+    if hasattr(model, "shard"):
+        try:
+            model.shard(group)  # type: ignore
+            return model
+        except (AttributeError, TypeError, NameError):
+            pass
+
+    if isinstance(model, (LlamaModel, Ministral3Model)):
+        logger.warning("shouldn't be hit - upstream sharding exists")
         tensor_parallel_sharding_strategy = LlamaShardingStrategy(
             group,
             all_to_sharded_linear,
@@ -223,7 +244,8 @@ def tensor_auto_parallel(
             all_to_sharded_linear_in_place,
             sharded_to_all_linear_in_place,
         )
-    elif isinstance(model, DeepseekV3Model):
+    elif isinstance(model, (DeepseekV3Model, DeepseekV32Model)):
+        logger.warning("shouldn't be hit - upstream sharding exists")
         tensor_parallel_sharding_strategy = DeepSeekShardingStrategy(
             group,
             all_to_sharded_linear,
@@ -231,7 +253,15 @@ def tensor_auto_parallel(
             all_to_sharded_linear_in_place,
             sharded_to_all_linear_in_place,
         )
-    elif isinstance(model, Qwen3MoeModel):
+    elif isinstance(model, MiniMaxModel):
+        tensor_parallel_sharding_strategy = MiniMaxShardingStrategy(
+            group,
+            all_to_sharded_linear,
+            sharded_to_all_linear,
+            all_to_sharded_linear_in_place,
+            sharded_to_all_linear_in_place,
+        )
+    elif isinstance(model, (Qwen3MoeModel, Glm4MoeModel, Qwen3NextModel)):
         tensor_parallel_sharding_strategy = QwenShardingStrategy(
             group,
             all_to_sharded_linear,
@@ -239,6 +269,15 @@ def tensor_auto_parallel(
             all_to_sharded_linear_in_place,
             sharded_to_all_linear_in_place,
         )
+    elif isinstance(model, GptOssModel):
+        tensor_parallel_sharding_strategy = GptOssShardingStrategy(
+            group,
+            all_to_sharded_linear,
+            sharded_to_all_linear,
+            all_to_sharded_linear_in_place,
+            sharded_to_all_linear_in_place,
+        )
+
     else:
         raise ValueError(f"Unsupported model type: {type(model)}")
 
@@ -284,13 +323,38 @@ class LlamaShardingStrategy(TensorParallelShardingStrategy):
         return model
 
 
+def _set_layers(model: nn.Module, layers: list[_LayerCallable]) -> None:
+    inner_model_instance = _inner_model(model)
+    if hasattr(inner_model_instance, "layers"):
+        inner_model_instance.layers = layers
+
+        # Update DeepSeek V3 specific parameters when layers are shrunk
+        if isinstance(
+            model, (DeepseekV3Model, DeepseekV32Model, Glm4MoeModel)
+        ) and hasattr(inner_model_instance, "num_layers"):
+            logger.info(
+                f"Setting num_layers to {len(layers)} for model {model.model.__class__.__name__}"
+            )
+            inner_model_instance.start_idx = 0
+            inner_model_instance.end_idx = len(layers)
+            inner_model_instance.num_layers = len(layers)
+        elif isinstance(model, Qwen3MoeModel):
+            logger.info(
+                f"Setting num_hidden_layers to {len(layers)} for model {model.model.__class__.__name__}"
+            )
+            inner_model_instance.num_hidden_layers = len(layers)
+    elif hasattr(inner_model_instance, "h"):
+        inner_model_instance.h = layers
+    else:
+        raise ValueError("Model must have either a 'layers' or 'h' attribute")
+
+
 class DeepSeekShardingStrategy(TensorParallelShardingStrategy):
     def shard_model(self, model: nn.Module) -> nn.Module:
         model = cast(DeepseekV3Model, model)
         for layer in model.layers:
             # Shard the self attention
-            if layer.self_attn.q_lora_rank is None:  # pyright: ignore[reportUnnecessaryComparison]
-                # Unfortunately, q_lora_rank can be None despite typing hints.
+            if layer.self_attn.q_lora_rank is None:
                 layer.self_attn.q_proj = self.all_to_sharded_linear(
                     layer.self_attn.q_proj
                 )
@@ -305,7 +369,7 @@ class DeepSeekShardingStrategy(TensorParallelShardingStrategy):
             layer.self_attn.num_heads //= self.N
 
             # Shard the MLP
-            if isinstance(layer.mlp, DeepseekV3MLP):
+            if isinstance(layer.mlp, (DeepseekV3MLP, DeepseekV32MLP)):
                 layer.mlp.gate_proj = self.all_to_sharded_linear(layer.mlp.gate_proj)
                 layer.mlp.down_proj = self.sharded_to_all_linear(layer.mlp.down_proj)
                 layer.mlp.up_proj = self.all_to_sharded_linear(layer.mlp.up_proj)
@@ -339,6 +403,35 @@ class ShardedDeepseekV3MoE(CustomMlxLayer):
         return y
 
 
+class MiniMaxShardingStrategy(TensorParallelShardingStrategy):
+    def shard_model(self, model: nn.Module) -> nn.Module:
+        model = cast(MiniMaxModel, model)
+        for layer in model.layers:
+            # Shard the self attention
+            layer.self_attn.q_proj = self.all_to_sharded_linear(layer.self_attn.q_proj)
+            layer.self_attn.k_proj = self.all_to_sharded_linear(layer.self_attn.k_proj)
+            layer.self_attn.v_proj = self.all_to_sharded_linear(layer.self_attn.v_proj)
+            layer.self_attn.o_proj = self.sharded_to_all_linear(layer.self_attn.o_proj)
+            layer.self_attn.num_attention_heads //= self.N
+            layer.self_attn.num_key_value_heads //= self.N
+
+            # Shard the MoE. Shard in place since the MoE should be responsible
+            # for aggregating the results.
+            self.all_to_sharded_linear_in_place(
+                layer.block_sparse_moe.switch_mlp.gate_proj
+            )
+            self.sharded_to_all_linear_in_place(
+                layer.block_sparse_moe.switch_mlp.down_proj
+            )
+            self.all_to_sharded_linear_in_place(
+                layer.block_sparse_moe.switch_mlp.up_proj
+            )
+            layer.block_sparse_moe = ShardedQwenMoE(layer.block_sparse_moe)  # pyright: ignore[reportAttributeAccessIssue, reportArgumentType]
+            layer.block_sparse_moe.sharding_group = self.group
+
+        return model
+
+
 class QwenShardingStrategy(TensorParallelShardingStrategy):
     def shard_model(self, model: nn.Module) -> nn.Module:
         model = cast(Qwen3MoeModel, model)
@@ -353,11 +446,13 @@ class QwenShardingStrategy(TensorParallelShardingStrategy):
 
             # Shard the MoE. Shard in place since the MoE should be responsible
             # for aggregating the results.
-            if isinstance(layer.mlp, Qwen3MoeSparseMoeBlock):
+            if isinstance(
+                layer.mlp, (Qwen3MoeSparseMoeBlock, MoE, Qwen3NextSparseMoeBlock)
+            ):
                 self.all_to_sharded_linear_in_place(layer.mlp.switch_mlp.gate_proj)
                 self.sharded_to_all_linear_in_place(layer.mlp.switch_mlp.down_proj)
                 self.all_to_sharded_linear_in_place(layer.mlp.switch_mlp.up_proj)
-                layer.mlp = ShardedQwenMoE(layer.mlp)  # type: ignore
+                layer.mlp = ShardedQwenMoE(layer.mlp)  # pyright: ignore[reportAttributeAccessIssue, reportArgumentType]
                 layer.mlp.sharding_group = self.group
 
             # Shard the MLP
@@ -381,3 +476,50 @@ class ShardedQwenMoE(CustomMlxLayer):
         if self.sharding_group is not None:
             y = mx.distributed.all_sum(y, group=self.sharding_group)
         return y
+
+
+class GptOssShardingStrategy(TensorParallelShardingStrategy):
+    def shard_model(self, model: nn.Module) -> nn.Module:
+        model = cast(GptOssMoeModel, model)
+
+        for layer in model.layers:
+            layer.self_attn.q_proj = self.all_to_sharded_linear(layer.self_attn.q_proj)
+            layer.self_attn.k_proj = self.all_to_sharded_linear(layer.self_attn.k_proj)
+            layer.self_attn.v_proj = self.all_to_sharded_linear(layer.self_attn.v_proj)
+            layer.self_attn.o_proj = self.sharded_to_all_linear(layer.self_attn.o_proj)
+
+            layer.self_attn.num_attention_heads //= self.N
+            layer.self_attn.num_key_value_heads //= self.N
+            layer.self_attn.num_key_value_groups = (
+                layer.self_attn.num_attention_heads
+                // layer.self_attn.num_key_value_heads
+            )
+
+            layer.self_attn.sinks = layer.self_attn.sinks[
+                layer.self_attn.num_attention_heads
+                * self.group.rank() : layer.self_attn.num_attention_heads
+                * (self.group.rank() + 1)
+            ]
+
+            self.all_to_sharded_linear_in_place(layer.mlp.experts.gate_proj)
+            self.sharded_to_all_linear_in_place(layer.mlp.experts.down_proj)
+            self.all_to_sharded_linear_in_place(layer.mlp.experts.up_proj)
+
+            layer.mlp = ShardedGptOssMoE(layer.mlp)  # type: ignore
+            layer.mlp.sharding_group = self.group
+
+        return model
+
+
+class ShardedGptOssMoE(CustomMlxLayer):
+    def __init__(self, layer: nn.Module):
+        super().__init__(layer)
+        self.sharding_group: mx.distributed.Group | None = None
+
+    def __call__(self, x: mx.array) -> mx.array:
+        if self.sharding_group is not None:
+            x = sum_gradients(self.sharding_group)(x)
+        y = self.original_layer(x)
+        if self.sharding_group is not None:
+            y = mx.distributed.all_sum(y, group=self.sharding_group)
+        return y

src/exo/worker/engines/mlx/generator/generate.py

@@ -3,6 +3,7 @@ from typing import Any, Callable, Generator, cast, get_args
 import mlx.core as mx
 from mlx_lm import stream_generate
 from mlx_lm.models.cache import KVCache
+from mlx_lm.sample_utils import make_sampler
 from mlx_lm.tokenizer_utils import TokenizerWrapper
 
 # from exo.engines.mlx.cache import KVPrefixCache
@@ -47,7 +48,6 @@ def maybe_quantize_kv_cache(
 def warmup_inference(
     model: Model,
     tokenizer: TokenizerWrapper,
-    sampler: Callable[[mx.array], mx.array],
 ) -> int:
     content = "Prompt to warm up the inference engine. Repeat this."
 
@@ -70,6 +70,9 @@ def warmup_inference(
         model=model,
     )
 
+    # Use a default sampler for warmup
+    sampler = make_sampler(temp=0.7)
+
     logger.info("Generating warmup tokens")
     for _r in stream_generate(
         model=model,
@@ -115,7 +118,6 @@ def eos_ids_from_tokenizer(tokenizer: TokenizerWrapper) -> list[int]:
 def mlx_generate(
     model: Model,
     tokenizer: TokenizerWrapper,
-    sampler: Callable[[mx.array], mx.array],
     task: ChatCompletionTaskParams,
 ) -> Generator[GenerationResponse]:
     # Ensure that generation stats only contains peak memory for this generation
@@ -125,6 +127,9 @@ def mlx_generate(
     # Currently we support chat-completion tasks only.
     logger.info(f"task_params: {task}")
 
+    if task.seed is not None:
+        mx.random.seed(task.seed)
+
     prompt = apply_chat_template(
         tokenizer=tokenizer,
         chat_task_data=task,
@@ -138,6 +143,11 @@ def mlx_generate(
         eos_ids = eos_ids_from_tokenizer(tokenizer)
         logits_processors = [ban_token_ids(eos_ids)]
 
+    sampler = make_sampler(
+        temp=task.temperature if task.temperature is not None else 0.7,
+        top_p=task.top_p if task.top_p is not None else 1.0,
+    )
+
     max_tokens = task.max_tokens or MAX_TOKENS
     for out in stream_generate(
         model=model,

src/exo/worker/engines/mlx/utils_mlx.py

@@ -1,13 +1,25 @@
 import json
 import os
 import resource
+import sys
 import time
 from pathlib import Path
-from typing import Any, Callable, cast
+from typing import Any, cast
+
+# Monkey-patch for transformers 5.x compatibility
+# Kimi's tokenization_kimi.py imports bytes_to_unicode from the old location
+# which was moved in transformers 5.0.0rc2
+try:
+    import transformers.models.gpt2.tokenization_gpt2 as gpt2_tokenization
+    from transformers.convert_slow_tokenizer import bytes_to_unicode
+
+    if not hasattr(gpt2_tokenization, "bytes_to_unicode"):
+        gpt2_tokenization.bytes_to_unicode = bytes_to_unicode  # type: ignore[attr-defined]
+except ImportError:
+    pass  # transformers < 5.0 or bytes_to_unicode not available
 
 from mlx_lm.models.cache import KVCache, QuantizedKVCache, RotatingKVCache
 from mlx_lm.models.deepseek_v3 import DeepseekV3Model
-from mlx_lm.sample_utils import make_sampler
 from mlx_lm.tokenizer_utils import TokenizerWrapper
 
 from exo.worker.engines.mlx.constants import (
@@ -19,7 +31,7 @@ from exo.worker.engines.mlx.constants import (
 try:
     from mlx_lm.tokenizer_utils import load_tokenizer
 except ImportError:
-    from mlx_lm.tokenizer_utils import load as load_tokenizer  # type: ignore
+    from mlx_lm.tokenizer_utils import load as load_tokenizer
 import contextlib
 
 import mlx.core as mx
@@ -152,6 +164,11 @@ def mlx_distributed_init(
                 os.environ["MLX_JACCL_COORDINATOR"] = jaccl_coordinator
                 group = mx.distributed.init(backend="jaccl", strict=True)
 
+            case _:
+                raise ValueError(
+                    f"Unsupported instance type for MLX distributed: {type(bound_instance.instance)}"
+                )
+
         logger.info(f"Rank {rank} mlx distributed initialization complete")
 
         return group
@@ -176,11 +193,7 @@ def initialize_mlx(
 
 def load_mlx_items(
     bound_instance: BoundInstance, group: Group | None
-) -> tuple[Model, TokenizerWrapper, Callable[[mx.array], mx.array]]:
-    # TODO: pass temperature
-    sampler: Callable[[mx.array], mx.array] = make_sampler(temp=0.7)
-    logger.info("Created a sampler")
-
+) -> tuple[Model, TokenizerWrapper]:
     if group is None:
         logger.info(f"Single device used for {bound_instance.instance}")
         model_path = build_model_path(bound_instance.bound_shard.model_meta.model_id)
@@ -201,7 +214,7 @@ def load_mlx_items(
 
     set_wired_limit_for_model(get_weights_size(bound_instance.bound_shard))
 
-    return cast(Model, model), tokenizer, sampler
+    return cast(Model, model), tokenizer
 
 
 def shard_and_load(
@@ -257,26 +270,70 @@ def shard_and_load(
     return model, tokenizer
 
 
-def get_tokenizer(model_path: Path, shard_metadata: ShardMetadata):
-    # TODO: Let's move away from this custom logic to mlx_lm.load()
-    if "kimi-k2" in shard_metadata.model_meta.model_id.lower():
-        eos_token_ids = [163586]
+def get_tokenizer(model_path: Path, shard_metadata: ShardMetadata) -> TokenizerWrapper:
+    """Load tokenizer for a model shard. Delegates to load_tokenizer_for_model_id."""
+    return load_tokenizer_for_model_id(shard_metadata.model_meta.model_id, model_path)
 
-    elif "glm" in shard_metadata.model_meta.model_id.lower():
-        eos_token_ids = [151336, 151329, 151338]
 
-    else:
-        eos_token_ids = None
+def get_eos_token_ids_for_model(model_id: str) -> list[int] | None:
+    """
+    Get the EOS token IDs for a model based on its ID.
 
-    tokenizer = cast(
-        TokenizerWrapper,
-        load_tokenizer(
-            model_path,
-            tokenizer_config_extra={"trust_remote_code": TRUST_REMOTE_CODE},
-            eos_token_ids=eos_token_ids,
-        ),
+    Some models require explicit EOS token configuration that isn't in their
+    tokenizer config. This function returns the known EOS token IDs for such models.
+
+    Args:
+        model_id: The HuggingFace model ID
+
+    Returns:
+        List of EOS token IDs, or None if the model uses standard tokenizer config
+    """
+    model_id_lower = model_id.lower()
+    if "kimi-k2" in model_id_lower:
+        return [163586]
+    elif "glm" in model_id_lower:
+        return [151336, 151329, 151338]
+    return None
+
+
+def load_tokenizer_for_model_id(model_id: str, model_path: Path) -> TokenizerWrapper:
+    """
+    Load tokenizer for a model given its ID and local path.
+
+    This is the core tokenizer loading logic, handling special cases for different
+    model families (Kimi, GLM, etc.) and transformers 5.x compatibility.
+
+    Args:
+        model_id: The HuggingFace model ID (e.g., "moonshotai/Kimi-K2-Instruct")
+        model_path: Local path where the model/tokenizer files are stored
+
+    Returns:
+        TokenizerWrapper instance configured for the model
+    """
+    model_id_lower = model_id.lower()
+    eos_token_ids = get_eos_token_ids_for_model(model_id)
+
+    # Kimi uses a custom TikTokenTokenizer that transformers 5.x can't load via AutoTokenizer
+    if "kimi-k2" in model_id_lower:
+        sys.path.insert(0, str(model_path))
+        from tokenization_kimi import TikTokenTokenizer  # type: ignore[import-not-found]  # noqa: I001
+
+        hf_tokenizer: Any = TikTokenTokenizer.from_pretrained(model_path)  # pyright: ignore[reportUnknownVariableType,reportUnknownMemberType]
+
+        # Patch encode to use internal tiktoken model directly
+        # transformers 5.x has a bug in the encode->pad path for slow tokenizers
+        def _patched_encode(text: str, **_kwargs: object) -> list[int]:
+            # Pass allowed_special="all" to handle special tokens like <|im_user|>
+            return list(hf_tokenizer.model.encode(text, allowed_special="all"))  # pyright: ignore[reportUnknownMemberType,reportUnknownArgumentType]
+
+        hf_tokenizer.encode = _patched_encode
+        return TokenizerWrapper(hf_tokenizer, eos_token_ids=eos_token_ids)
+
+    tokenizer = load_tokenizer(
+        model_path,
+        tokenizer_config_extra={"trust_remote_code": TRUST_REMOTE_CODE},
+        eos_token_ids=eos_token_ids,
     )
-    assert isinstance(tokenizer, TokenizerWrapper)
 
     return tokenizer
 
@@ -289,15 +346,15 @@ def apply_chat_template(
     messages = chat_task_data.messages
 
     formatted_messages: list[dict[str, Any]] = []
-    for _, message in enumerate(messages):
+    for message in messages:
         if isinstance(message.content, ChatCompletionMessageText):
             message.content = message.content.text
         if isinstance(message.content, list):
-            if len(message.content) != 1:
-                logger.warning("Received malformed prompt")
+            if len(message.content) == 0:
+                logger.warning("Received prompt with no content, skipping")
                 continue
 
-            message.content = message.content[0].text
+            message.content = "\n".join(c.text for c in message.content).strip()
         if message.content is None and message.thinking is None:
             continue
 
@@ -306,13 +363,14 @@ def apply_chat_template(
             {k: v for k, v in message.model_dump().items() if v is not None}  # type: ignore
         )
 
-    prompt: str = tokenizer.apply_chat_template(  # type: ignore
+    prompt: str = tokenizer.apply_chat_template(
         formatted_messages,
         tokenize=False,
         add_generation_prompt=True,
+        tools=chat_task_data.tools,
     )
 
-    return prompt  # type: ignore
+    return prompt
 
 
 class NullKVCache(KVCache):

src/exo/worker/main.py

@@ -217,7 +217,9 @@ class Worker:
                     )
                     if initial_progress.status == "complete":
                         progress = DownloadCompleted(
-                            shard_metadata=shard, node_id=self.node_id
+                            shard_metadata=shard,
+                            node_id=self.node_id,
+                            total_bytes=initial_progress.total_bytes,
                         )
                         self.download_status[shard.model_meta.model_id] = progress
                         await self.event_sender.send(
@@ -364,7 +366,11 @@ class Worker:
             nonlocal self
             nonlocal last_progress_time
             if progress.status == "complete":
-                status = DownloadCompleted(shard_metadata=shard, node_id=self.node_id)
+                status = DownloadCompleted(
+                    shard_metadata=shard,
+                    node_id=self.node_id,
+                    total_bytes=progress.total_bytes,
+                )
                 self.download_status[shard.model_meta.model_id] = status
                 # Footgun!
                 self.event_sender.send_nowait(
@@ -457,7 +463,9 @@ class Worker:
                 ) in self.shard_downloader.get_shard_download_status():
                     if progress.status == "complete":
                         status = DownloadCompleted(
-                            node_id=self.node_id, shard_metadata=progress.shard
+                            node_id=self.node_id,
+                            shard_metadata=progress.shard,
+                            total_bytes=progress.total_bytes,
                         )
                     elif progress.status in ["in_progress", "not_started"]:
                         if progress.downloaded_bytes_this_session.in_bytes == 0:

src/exo/worker/plan.py

@@ -21,7 +21,12 @@ from exo.shared.types.worker.downloads import (
     DownloadOngoing,
     DownloadProgress,
 )
-from exo.shared.types.worker.instances import BoundInstance, Instance, InstanceId
+from exo.shared.types.worker.instances import (
+    BoundInstance,
+    FLASHInstance,
+    Instance,
+    InstanceId,
+)
 from exo.shared.types.worker.runners import (
     RunnerConnected,
     RunnerConnecting,
@@ -50,6 +55,11 @@ def plan(
     all_runners: Mapping[RunnerId, RunnerStatus],  # all global
     tasks: Mapping[TaskId, Task],
 ) -> Task | None:
+    # Check for FLASH instance tasks first
+    flash_task = _plan_flash(runners, instances)
+    if flash_task is not None:
+        return flash_task
+
     # Python short circuiting OR logic should evaluate these sequentially.
     return (
         _kill_runner(runners, all_runners, instances)
@@ -62,6 +72,34 @@ def plan(
     )
 
 
+def _plan_flash(
+    runners: Mapping[RunnerId, RunnerSupervisor],
+    instances: Mapping[InstanceId, Instance],
+) -> Task | None:
+    """Plan tasks specifically for FLASH instances.
+
+    FLASH instances have a simpler lifecycle:
+    - CreateRunner (handled by _create_runner)
+    - LoadModel (starts the simulation immediately)
+    - Shutdown (handled by _kill_runner)
+
+    This function handles the LoadModel step for FLASH instances,
+    skipping the MLX-specific download/init/warmup steps.
+    """
+    for runner in runners.values():
+        instance = runner.bound_instance.instance
+
+        # Only handle FLASH instances
+        if not isinstance(instance, FLASHInstance):
+            continue
+
+        # If runner is idle, emit LoadModel to start the simulation
+        if isinstance(runner.status, RunnerIdle):
+            return LoadModel(instance_id=instance.instance_id)
+
+    return None
+
+
 def _kill_runner(
     runners: Mapping[RunnerId, RunnerSupervisor],
     all_runners: Mapping[RunnerId, RunnerStatus],
@@ -114,6 +152,10 @@ def _model_needs_download(
     download_status: Mapping[ModelId, DownloadProgress],
 ) -> DownloadModel | None:
     for runner in runners.values():
+        # FLASH instances don't need model downloads
+        if isinstance(runner.bound_instance.instance, FLASHInstance):
+            continue
+
         model_id = runner.bound_instance.bound_shard.model_meta.model_id
         if isinstance(runner.status, RunnerIdle) and (
             model_id not in download_status

src/exo/worker/runner/bootstrap.py

@@ -4,7 +4,11 @@ import loguru
 
 from exo.shared.types.events import Event, RunnerStatusUpdated
 from exo.shared.types.tasks import Task
-from exo.shared.types.worker.instances import BoundInstance, MlxJacclInstance
+from exo.shared.types.worker.instances import (
+    BoundInstance,
+    FLASHInstance,
+    MlxJacclInstance,
+)
 from exo.shared.types.worker.runners import RunnerFailed
 from exo.utils.channels import ClosedResourceError, MpReceiver, MpSender
 
@@ -17,20 +21,27 @@ def entrypoint(
     task_receiver: MpReceiver[Task],
     _logger: "loguru.Logger",
 ) -> None:
-    if (
-        isinstance(bound_instance.instance, MlxJacclInstance)
-        and len(bound_instance.instance.ibv_devices) >= 2
-    ):
-        os.environ["MLX_METAL_FAST_SYNCH"] = "1"
-
     global logger
     logger = _logger
 
-    # Import main after setting global logger - this lets us just import logger from this module
+    # Route based on instance type
     try:
-        from exo.worker.runner.runner import main
+        if isinstance(bound_instance.instance, FLASHInstance):
+            # FLASH MPI simulation runner
+            from exo.worker.runner.flash_runner import main
 
-        main(bound_instance, event_sender, task_receiver)
+            main(bound_instance, event_sender, task_receiver)
+        else:
+            # MLX runner (default)
+            if (
+                isinstance(bound_instance.instance, MlxJacclInstance)
+                and len(bound_instance.instance.ibv_devices) >= 2
+            ):
+                os.environ["MLX_METAL_FAST_SYNCH"] = "1"
+
+            from exo.worker.runner.runner import main
+
+            main(bound_instance, event_sender, task_receiver)
     except ClosedResourceError:
         logger.warning("Runner communication closed unexpectedly")
     except Exception as e:

src/exo/worker/runner/flash_runner.py

@@ -0,0 +1,301 @@
+"""FLASH MPI Runner - spawns and monitors FLASH simulations.
+
+Exo-native distributed MPI:
+- Exo handles node discovery and coordination
+- Coordinator generates hostfile from Exo topology
+- mpirun uses exo-rsh (no SSH required) to spawn on remote nodes
+- exo-rsh connects to each node's Exo API (/execute endpoint) for remote execution
+- Workers just report ready and wait
+"""
+
+import os
+import shutil
+import socket
+import subprocess
+import threading
+
+from exo.shared.types.events import (
+    Event,
+    RunnerStatusUpdated,
+    TaskAcknowledged,
+    TaskStatusUpdated,
+)
+from exo.shared.types.tasks import (
+    LoadModel,
+    Shutdown,
+    Task,
+    TaskStatus,
+)
+from exo.shared.types.worker.instances import BoundInstance, FLASHInstance
+from exo.shared.types.worker.runners import (
+    RunnerFailed,
+    RunnerIdle,
+    RunnerLoading,
+    RunnerReady,
+    RunnerRunning,
+    RunnerShutdown,
+    RunnerShuttingDown,
+    RunnerStatus,
+)
+from exo.utils.channels import MpReceiver, MpSender
+from exo.worker.runner.bootstrap import logger
+
+# Find mpirun in PATH, fallback to common locations
+MPIRUN_PATH = shutil.which("mpirun") or "/opt/homebrew/bin/mpirun"
+
+# exo-rsh is installed as console script by exo package
+_exo_rsh_path = shutil.which("exo-rsh")
+if not _exo_rsh_path:
+    raise RuntimeError("exo-rsh not found in PATH - this should be installed with exo")
+EXO_RSH_PATH: str = _exo_rsh_path
+
+
+def get_my_rank(instance: FLASHInstance, my_node_id: str) -> int:
+    """Determine this node's rank based on position in hosts_by_node."""
+    for i, node_id in enumerate(instance.hosts_by_node.keys()):
+        if str(node_id) == str(my_node_id):
+            return i
+    return -1
+
+
+def get_coordinator_host(instance: FLASHInstance) -> str:
+    """Get the IP of the coordinator node."""
+    return instance.coordinator_ip
+
+
+def resolve_host(host: str) -> str:
+    """Resolve host string to a usable hostname for MPI hostfile.
+
+    Accepts either an IP address or hostname. For IPs, attempts to resolve
+    to a hostname via DNS/mDNS. Hostnames are returned as-is after validation.
+    """
+    # Check if input is already a hostname (not an IP)
+    try:
+        socket.inet_aton(host)
+        is_ip = True
+    except socket.error:
+        is_ip = False
+
+    if not is_ip:
+        # Already a hostname, verify it resolves and return as-is
+        try:
+            socket.gethostbyname(host)
+            return host
+        except socket.gaierror:
+            logger.warning(f"Hostname {host} does not resolve, using anyway")
+            return host
+
+    # It's an IP address, try to resolve to hostname
+    try:
+        hostname, _, _ = socket.gethostbyaddr(host)
+        hostname = hostname.split(".")[0]
+        logger.info(f"Resolved {host} to {hostname}")
+        return hostname
+    except socket.herror:
+        pass
+
+    # Fall back to IP
+    logger.warning(f"Could not resolve {host} to hostname, using IP directly")
+    return host
+
+
+def generate_hostfile(instance: FLASHInstance, working_dir: str) -> str:
+    """Generate MPI hostfile from instance topology."""
+    hostfile_path = os.path.join(working_dir, "flash_hosts.txt")
+    with open(hostfile_path, "w") as f:
+        for _node_id, hosts in instance.hosts_by_node.items():
+            if hosts:
+                host = resolve_host(hosts[0].ip)
+                f.write(f"{host} slots={instance.ranks_per_node}\n")
+    logger.info(f"Generated hostfile at {hostfile_path}")
+    with open(hostfile_path, "r") as f:
+        logger.info(f"Hostfile contents:\n{f.read()}")
+    return hostfile_path
+
+
+def main(
+    bound_instance: BoundInstance,
+    event_sender: MpSender[Event],
+    task_receiver: MpReceiver[Task],
+):
+    """Main FLASH runner loop.
+
+    Coordinator: generates hostfile and runs mpirun (uses exo-rsh instead of SSH)
+    Workers: just report ready and wait for mpirun to spawn processes on them
+    """
+    assert isinstance(bound_instance.instance, FLASHInstance)
+    instance = bound_instance.instance
+    runner_id = bound_instance.bound_runner_id
+    my_node_id = str(bound_instance.bound_node_id)
+
+    logger.info(f"FLASH runner starting for simulation: {instance.simulation_name}")
+
+    my_rank = get_my_rank(instance, my_node_id)
+    world_size = len(instance.hosts_by_node)
+    is_coordinator = my_rank == 0
+    coordinator_ip = get_coordinator_host(instance)
+
+    logger.info(
+        f"FLASH node: rank={my_rank}, world_size={world_size}, coordinator={is_coordinator}"
+    )
+    logger.info(f"FLASH coordinator IP: {coordinator_ip}")
+
+    process: subprocess.Popen[bytes] | None = None
+    current_status: RunnerStatus = RunnerIdle()
+    shutdown_requested = False
+
+    event_sender.send(
+        RunnerStatusUpdated(runner_id=runner_id, runner_status=current_status)
+    )
+
+    def monitor_output(proc: subprocess.Popen[bytes]) -> None:
+        """Monitor FLASH stdout for progress updates."""
+        if proc.stdout is None:
+            return
+        for line in iter(proc.stdout.readline, b""):
+            if shutdown_requested:
+                break
+            try:
+                decoded: str = line.decode("utf-8", errors="replace").strip()
+                if decoded:
+                    logger.info(f"[FLASH] {decoded}")
+            except Exception as e:
+                logger.warning(f"Error parsing FLASH output: {e}")
+
+    with task_receiver as tasks:
+        for task in tasks:
+            event_sender.send(
+                TaskStatusUpdated(task_id=task.task_id, task_status=TaskStatus.Running)
+            )
+            event_sender.send(TaskAcknowledged(task_id=task.task_id))
+
+            match task:
+                case LoadModel() if isinstance(current_status, RunnerIdle):
+                    current_status = RunnerLoading()
+                    logger.info("Starting FLASH simulation")
+                    event_sender.send(
+                        RunnerStatusUpdated(
+                            runner_id=runner_id, runner_status=current_status
+                        )
+                    )
+
+                    try:
+                        if is_coordinator:
+                            # Coordinator: generate hostfile and run mpirun
+                            hostfile = generate_hostfile(
+                                instance, instance.working_directory
+                            )
+
+                            iface = instance.network_interface
+                            cmd = [
+                                MPIRUN_PATH,
+                                "-np",
+                                str(instance.total_ranks),
+                                "--hostfile",
+                                hostfile,
+                                "--wdir",
+                                instance.working_directory,
+                                "--oversubscribe",
+                                "--mca",
+                                "btl",
+                                "tcp,self",
+                                "--mca",
+                                "btl_tcp_if_include",
+                                iface,
+                                "--mca",
+                                "oob_tcp_if_include",
+                                iface,
+                                "--mca",
+                                "plm_rsh_no_tree_spawn",
+                                "1",
+                            ]
+
+                            # Use exo-rsh for remote execution (no SSH needed)
+                            cmd.extend(["--mca", "plm_rsh_agent", EXO_RSH_PATH])
+
+                            cmd.append(instance.flash_executable_path)
+
+                            logger.info(f"FLASH distributed launch: {' '.join(cmd)}")
+
+                            process = subprocess.Popen(
+                                cmd,
+                                cwd=instance.working_directory,
+                                stdout=subprocess.PIPE,
+                                stderr=subprocess.STDOUT,
+                            )
+
+                            monitor_thread = threading.Thread(
+                                target=monitor_output, args=(process,), daemon=True
+                            )
+                            monitor_thread.start()
+
+                            current_status = RunnerRunning()
+                            logger.info(
+                                f"FLASH running on {world_size} nodes with {instance.total_ranks} ranks"
+                            )
+
+                        else:
+                            # Worker: mpirun on coordinator will use exo-rsh to spawn processes here
+                            logger.info(
+                                f"Worker {my_rank}: Ready for mpirun to spawn processes via exo-rsh"
+                            )
+                            current_status = RunnerRunning()
+
+                    except Exception as e:
+                        logger.error(f"Failed to start FLASH: {e}")
+                        import traceback
+
+                        logger.error(traceback.format_exc())
+                        current_status = RunnerFailed(error_message=str(e))
+
+                case Shutdown():
+                    shutdown_requested = True
+                    current_status = RunnerShuttingDown()
+                    logger.info("FLASH runner shutting down")
+                    event_sender.send(
+                        RunnerStatusUpdated(
+                            runner_id=runner_id, runner_status=current_status
+                        )
+                    )
+
+                    if process and process.poll() is None:
+                        logger.info("Terminating FLASH simulation")
+                        process.terminate()
+                        try:
+                            process.wait(timeout=10)
+                        except subprocess.TimeoutExpired:
+                            logger.warning("FLASH didn't terminate, killing")
+                            process.kill()
+                            process.wait()
+
+                    current_status = RunnerShutdown()
+
+                case _:
+                    if process and process.poll() is not None:
+                        exit_code = process.returncode
+                        if exit_code == 0:
+                            logger.info("FLASH simulation completed successfully")
+                            current_status = RunnerReady()
+                        else:
+                            logger.error(
+                                f"FLASH simulation failed with code {exit_code}"
+                            )
+                            current_status = RunnerFailed(
+                                error_message=f"Exit code {exit_code}"
+                            )
+
+            event_sender.send(
+                TaskStatusUpdated(task_id=task.task_id, task_status=TaskStatus.Complete)
+            )
+            event_sender.send(
+                RunnerStatusUpdated(runner_id=runner_id, runner_status=current_status)
+            )
+
+            if isinstance(current_status, RunnerShutdown):
+                break
+
+    if process and process.poll() is None:
+        process.terminate()
+        process.wait(timeout=5)
+
+    logger.info("FLASH runner exiting")

src/exo/worker/runner/runner.py

@@ -68,7 +68,6 @@ def main(
 
     model = None
     tokenizer = None
-    sampler = None
     group = None
 
     current_status: RunnerStatus = RunnerIdle()
@@ -110,14 +109,13 @@ def main(
                         )
                     )
 
-                    model, tokenizer, sampler = load_mlx_items(bound_instance, group)
+                    model, tokenizer = load_mlx_items(bound_instance, group)
 
                     current_status = RunnerLoaded()
                     logger.info("runner loaded")
                 case StartWarmup() if isinstance(current_status, RunnerLoaded):
                     assert model
                     assert tokenizer
-                    assert sampler
                     current_status = RunnerWarmingUp()
                     logger.info("runner warming up")
                     event_sender.send(
@@ -130,7 +128,6 @@ def main(
                     toks = warmup_inference(
                         model=model,
                         tokenizer=tokenizer,
-                        sampler=sampler,
                         # kv_prefix_cache=kv_prefix_cache,  # supply for warmup-time prefix caching
                     )
                     logger.info(f"warmed up by generating {toks} tokens")
@@ -144,7 +141,6 @@ def main(
                 ):
                     assert model
                     assert tokenizer
-                    assert sampler
                     logger.info(f"received chat request: {str(task)[:500]}")
                     current_status = RunnerRunning()
                     logger.info("runner running")
@@ -160,7 +156,6 @@ def main(
                     for response in mlx_generate(
                         model=model,
                         tokenizer=tokenizer,
-                        sampler=sampler,
                         task=task_params,
                     ):
                         match response:
@@ -204,7 +199,7 @@ def main(
                 RunnerStatusUpdated(runner_id=runner_id, runner_status=current_status)
             )
             if isinstance(current_status, RunnerShutdown):
-                del model, tokenizer, group, sampler
+                del model, tokenizer, group
                 mx.clear_cache()
                 import gc
 

src/exo/worker/tests/unittests/test_mlx/test_tokenizers.py

@@ -0,0 +1,386 @@
+"""
+Unit tests for tokenizer loading and functionality across all supported models.
+
+This test downloads only tokenizer-related files (not full model weights) to verify
+that tokenizers can be loaded and used correctly for encoding/decoding.
+"""
+
+import asyncio
+import contextlib
+from pathlib import Path
+
+import pytest
+
+from exo.shared.models.model_cards import MODEL_CARDS, ModelCard
+from exo.worker.download.download_utils import (
+    download_file_with_retry,
+    ensure_models_dir,
+    fetch_file_list_with_cache,
+)
+from exo.worker.engines.mlx.utils_mlx import (
+    get_eos_token_ids_for_model,
+    load_tokenizer_for_model_id,
+)
+
+# Files needed for tokenizer functionality
+TOKENIZER_FILE_PATTERNS = [
+    "tokenizer.json",
+    "tokenizer_config.json",
+    "special_tokens_map.json",
+    "vocab.json",
+    "vocab.txt",
+    "merges.txt",
+    "tiktoken.model",
+    "added_tokens.json",
+    "tokenizer.model",
+    "tokenization_*.py",  # Custom tokenizer implementations
+]
+
+
+def is_tokenizer_file(filename: str) -> bool:
+    """Check if a file is needed for tokenizer functionality."""
+    for pattern in TOKENIZER_FILE_PATTERNS:
+        if "*" in pattern:
+            prefix = pattern.split("*")[0]
+            suffix = pattern.split("*")[1]
+            if filename.startswith(prefix) and filename.endswith(suffix):
+                return True
+        elif filename == pattern:
+            return True
+    return False
+
+
+async def download_tokenizer_files(model_id: str) -> Path:
+    """Download only the tokenizer-related files for a model."""
+    target_dir = await ensure_models_dir() / model_id.replace("/", "--")
+    target_dir.mkdir(parents=True, exist_ok=True)
+
+    file_list = await fetch_file_list_with_cache(model_id, "main", recursive=True)
+
+    tokenizer_files = [f for f in file_list if is_tokenizer_file(f.path)]
+
+    if not tokenizer_files:
+        pytest.skip(f"No tokenizer files found for {model_id}")
+
+    for file_entry in tokenizer_files:
+        with contextlib.suppress(FileNotFoundError):
+            await download_file_with_retry(
+                model_id, "main", file_entry.path, target_dir
+            )
+
+    return target_dir
+
+
+# Get a sample of models to test (one per family to keep tests fast)
+def get_test_models() -> list[tuple[str, ModelCard]]:
+    """Get a representative sample of models to test."""
+    # Pick one model from each family to test
+    families: dict[str, tuple[str, ModelCard]] = {}
+    for short_id, card in MODEL_CARDS.items():
+        # Extract family name (e.g., "llama-3.1" from "llama-3.1-8b")
+        parts = short_id.split("-")
+        family = "-".join(parts[:2]) if len(parts) >= 2 else parts[0]
+
+        if family not in families:
+            families[family] = (short_id, card)
+
+    return list(families.values())
+
+
+TEST_MODELS: list[tuple[str, ModelCard]] = get_test_models()
+
+
+@pytest.fixture(scope="module")
+def event_loop():
+    """Create event loop for async tests."""
+    loop = asyncio.new_event_loop()
+    yield loop
+    loop.close()
+
+
+@pytest.mark.parametrize(
+    "short_id,model_card",
+    TEST_MODELS,
+    ids=[m[0] for m in TEST_MODELS],
+)
+@pytest.mark.asyncio
+async def test_tokenizer_encode_decode(short_id: str, model_card: ModelCard) -> None:
+    """Test that tokenizer can encode and decode text correctly."""
+    model_id = str(model_card.model_id)
+
+    # Download tokenizer files
+    model_path = await download_tokenizer_files(model_id)
+
+    # Verify required files exist
+    has_tokenizer = (
+        (model_path / "tokenizer.json").exists()
+        or (model_path / "tokenizer_config.json").exists()
+        or (model_path / "tiktoken.model").exists()
+        or (model_path / "tokenizer.model").exists()
+    )
+    if not has_tokenizer:
+        pytest.skip(f"Required tokenizer files not found for {model_id}")
+
+    # Load tokenizer
+    tokenizer = load_tokenizer_for_model_id(model_id, model_path)
+
+    # Test basic encoding
+    test_text = "Hello, world!"
+    encoded = tokenizer.encode(test_text)
+    assert isinstance(encoded, list), f"encode() should return a list for {model_id}"
+    assert len(encoded) > 0, f"encode() should return non-empty list for {model_id}"
+    assert all(isinstance(t, int) for t in encoded), (
+        f"All tokens should be integers for {model_id}"
+    )
+
+    # Test decoding
+    decoded = tokenizer.decode(encoded)
+    assert isinstance(decoded, str), f"decode() should return a string for {model_id}"
+    assert test_text in decoded or decoded.strip() == test_text.strip(), (
+        f"decode(encode(x)) should preserve text for {model_id}: got {decoded!r}"
+    )
+
+    # Test with longer text
+    long_text = "The quick brown fox jumps over the lazy dog. " * 10
+    long_encoded = tokenizer.encode(long_text)
+    assert len(long_encoded) > len(encoded), (
+        f"Longer text should produce more tokens for {model_id}"
+    )
+
+    # Test empty string
+    empty_encoded = tokenizer.encode("")
+    assert isinstance(empty_encoded, list), (
+        f"encode('') should return a list for {model_id}"
+    )
+
+    # Test special characters
+    special_text = 'Hello!\n\tWorld? <test> & "quotes"'
+    special_encoded = tokenizer.encode(special_text)
+    assert len(special_encoded) > 0, f"Special chars should encode for {model_id}"
+
+    # Test unicode
+    unicode_text = "Hello 世界 🌍"
+    unicode_encoded = tokenizer.encode(unicode_text)
+    assert len(unicode_encoded) > 0, f"Unicode should encode for {model_id}"
+
+
+@pytest.mark.parametrize(
+    "short_id,model_card",
+    TEST_MODELS,
+    ids=[m[0] for m in TEST_MODELS],
+)
+@pytest.mark.asyncio
+async def test_tokenizer_has_required_attributes(
+    short_id: str, model_card: ModelCard
+) -> None:
+    """Test that tokenizer has required attributes for inference."""
+    model_id = str(model_card.model_id)
+
+    model_path = await download_tokenizer_files(model_id)
+
+    has_tokenizer = (
+        (model_path / "tokenizer.json").exists()
+        or (model_path / "tokenizer_config.json").exists()
+        or (model_path / "tiktoken.model").exists()
+        or (model_path / "tokenizer.model").exists()
+    )
+    if not has_tokenizer:
+        pytest.skip(f"Required tokenizer files not found for {model_id}")
+
+    tokenizer = load_tokenizer_for_model_id(model_id, model_path)
+    eos_token_ids = get_eos_token_ids_for_model(model_id)
+
+    # Check for vocabulary size
+    empty_vocab: dict[str, int] = {}
+    vocab_size: int = getattr(tokenizer, "vocab_size", None) or len(
+        getattr(tokenizer, "get_vocab", lambda: empty_vocab)()
+    )
+    assert vocab_size > 0, f"Tokenizer should have vocab_size > 0 for {model_id}"
+
+    # Check for EOS token (either from tokenizer or explicitly provided)
+    has_eos = (
+        eos_token_ids is not None
+        or getattr(tokenizer, "eos_token_id", None) is not None
+        or getattr(tokenizer, "eos_token", None) is not None
+    )
+    assert has_eos, f"Tokenizer should have EOS token for {model_id}"
+
+
+@pytest.mark.parametrize(
+    "short_id,model_card",
+    TEST_MODELS,
+    ids=[m[0] for m in TEST_MODELS],
+)
+@pytest.mark.asyncio
+async def test_tokenizer_special_tokens(short_id: str, model_card: ModelCard) -> None:
+    """Test that tokenizer can encode text containing special tokens.
+
+    This is critical because the actual inference path uses prompts with
+    special tokens from chat templates. If special tokens aren't handled
+    correctly, encoding will fail.
+    """
+    model_id = str(model_card.model_id)
+
+    model_path = await download_tokenizer_files(model_id)
+
+    has_tokenizer = (
+        (model_path / "tokenizer.json").exists()
+        or (model_path / "tokenizer_config.json").exists()
+        or (model_path / "tiktoken.model").exists()
+        or (model_path / "tokenizer.model").exists()
+    )
+    assert has_tokenizer, f"Required tokenizer files not found for {model_id}"
+
+    tokenizer = load_tokenizer_for_model_id(model_id, model_path)
+
+    # Get special tokens from the tokenizer
+    special_tokens: list[str] = []
+
+    # Try to get special tokens from various sources
+    if hasattr(tokenizer, "all_special_tokens"):
+        special_tokens.extend(tokenizer.all_special_tokens)
+    elif hasattr(tokenizer, "_tokenizer") and hasattr(
+        tokenizer._tokenizer,
+        "all_special_tokens",
+    ):
+        special_tokens.extend(tokenizer._tokenizer.all_special_tokens)
+
+    # Also check for common special token attributes
+    for attr in [
+        "bos_token",
+        "eos_token",
+        "pad_token",
+        "unk_token",
+        "sep_token",
+        "cls_token",
+    ]:
+        token = getattr(tokenizer, attr, None)
+        if token is None and hasattr(tokenizer, "_tokenizer"):
+            token = getattr(tokenizer._tokenizer, attr, None)
+        if token and isinstance(token, str) and token not in special_tokens:
+            special_tokens.append(token)
+
+    # If we found special tokens, test encoding text that contains them
+    if special_tokens:
+        # Create text with special tokens interspersed
+        test_with_special = f"{special_tokens[0]}Hello world"
+        if len(special_tokens) > 1:
+            test_with_special += f"{special_tokens[1]}"
+
+        encoded = tokenizer.encode(test_with_special)
+        assert isinstance(encoded, list), (
+            f"encode() with special tokens should return list for {model_id}"
+        )
+        assert len(encoded) > 0, (
+            f"encode() with special tokens should return non-empty list for {model_id}"
+        )
+        assert all(isinstance(t, int) for t in encoded), (
+            f"All tokens should be integers for {model_id}"
+        )
+
+        # Verify we can decode
+        decoded = tokenizer.decode(encoded)
+        assert isinstance(decoded, str), f"decode() should return string for {model_id}"
+
+    # Test with angle-bracket tokens (common format for special tokens)
+    # These should not raise errors even if they're not actual special tokens
+    angle_bracket_text = "<|test|>Hello<|end|>"
+    encoded = tokenizer.encode(angle_bracket_text)
+    assert isinstance(encoded, list), (
+        f"encode() with angle brackets should return list for {model_id}"
+    )
+    assert len(encoded) > 0, (
+        f"encode() with angle brackets should be non-empty for {model_id}"
+    )
+
+
+# Specifically test Kimi tokenizer since it has special handling
+@pytest.mark.asyncio
+async def test_kimi_tokenizer_specifically():
+    """Test Kimi tokenizer with its specific patches and quirks."""
+    kimi_models = [
+        (short_id, card)
+        for short_id, card in MODEL_CARDS.items()
+        if "kimi" in short_id.lower()
+    ]
+
+    if not kimi_models:
+        pytest.skip("No Kimi models found in MODEL_CARDS")
+
+    _, model_card = kimi_models[0]
+    model_id = str(model_card.model_id)
+
+    model_path = await download_tokenizer_files(model_id)
+
+    # Ensure the custom tokenizer file exists
+    if not (model_path / "tokenization_kimi.py").exists():
+        pytest.skip("tokenization_kimi.py not found")
+
+    tokenizer = load_tokenizer_for_model_id(model_id, model_path)
+    eos_token_ids = get_eos_token_ids_for_model(model_id)
+
+    # Test encode/decode cycle
+    test_text = "Hello, world!"
+    encoded = tokenizer.encode(test_text)
+    decoded = tokenizer.decode(encoded)
+
+    assert len(encoded) > 0, "Kimi tokenizer should encode text"
+    assert isinstance(decoded, str), "Kimi tokenizer should decode to string"
+
+    # Test that the patched encode works (returns list of ints)
+    assert all(isinstance(t, int) for t in encoded), "Tokens should be integers"
+
+    # Test encoding text with special tokens (like from chat templates)
+    # This is critical - the warmup inference uses prompts with special tokens
+    special_token_text = "<|im_user|>user<|im_middle|>Hello<|im_end|><|im_assistant|>"
+    special_encoded = tokenizer.encode(special_token_text)
+    assert len(special_encoded) > 0, "Kimi tokenizer should handle special tokens"
+    assert all(isinstance(t, int) for t in special_encoded), (
+        "Special token encoding should return integers"
+    )
+
+    # Verify EOS token is set
+    assert eos_token_ids == [163586], "Kimi EOS token should be [163586]"
+
+
+# Test GLM tokenizer since it also has special handling
+@pytest.mark.asyncio
+async def test_glm_tokenizer_specifically():
+    """Test GLM tokenizer with its specific EOS tokens."""
+    glm_models = [
+        (short_id, card)
+        for short_id, card in MODEL_CARDS.items()
+        if "glm" in short_id.lower()
+    ]
+
+    if not glm_models:
+        pytest.skip("No GLM models found in MODEL_CARDS")
+
+    _, model_card = glm_models[0]
+    model_id = str(model_card.model_id)
+
+    model_path = await download_tokenizer_files(model_id)
+
+    has_tokenizer = (model_path / "tokenizer.json").exists() or (
+        model_path / "tokenizer_config.json"
+    ).exists()
+    if not has_tokenizer:
+        pytest.skip("GLM tokenizer files not found")
+
+    tokenizer = load_tokenizer_for_model_id(model_id, model_path)
+    eos_token_ids = get_eos_token_ids_for_model(model_id)
+
+    # Test encode/decode
+    test_text = "Hello, world!"
+    encoded = tokenizer.encode(test_text)
+    decoded = tokenizer.decode(encoded)
+
+    assert len(encoded) > 0, "GLM tokenizer should encode text"
+    assert isinstance(decoded, str), "GLM tokenizer should decode to string"
+
+    # Verify EOS tokens
+    assert eos_token_ids == [
+        151336,
+        151329,
+        151338,
+    ], "GLM EOS tokens should be correct"

src/exo/worker/tests/unittests/test_plan/test_download_and_loading.py

@@ -1,5 +1,6 @@
 import exo.worker.plan as plan_mod
 from exo.shared.types.common import NodeId
+from exo.shared.types.memory import Memory
 from exo.shared.types.models import ModelId
 from exo.shared.types.tasks import LoadModel
 from exo.shared.types.worker.downloads import DownloadCompleted, DownloadProgress
@@ -94,13 +95,23 @@ def test_plan_loads_model_when_all_shards_downloaded_and_waiting():
 
     # Local node has already marked its shard as downloaded (not actually used by _load_model)
     local_download_status = {
-        MODEL_A_ID: DownloadCompleted(shard_metadata=shard1, node_id=NODE_A)
+        MODEL_A_ID: DownloadCompleted(
+            shard_metadata=shard1, node_id=NODE_A, total_bytes=Memory()
+        )
     }
 
     # Global view has completed downloads for both nodes
     global_download_status = {
-        NODE_A: [DownloadCompleted(shard_metadata=shard1, node_id=NODE_A)],
-        NODE_B: [DownloadCompleted(shard_metadata=shard2, node_id=NODE_B)],
+        NODE_A: [
+            DownloadCompleted(
+                shard_metadata=shard1, node_id=NODE_A, total_bytes=Memory()
+            )
+        ],
+        NODE_B: [
+            DownloadCompleted(
+                shard_metadata=shard2, node_id=NODE_B, total_bytes=Memory()
+            )
+        ],
     }
 
     result = plan_mod.plan(
@@ -140,7 +151,9 @@ def test_plan_does_not_request_download_when_shard_already_downloaded():
 
     # Local status claims the shard is downloaded already
     local_download_status = {
-        MODEL_A_ID: DownloadCompleted(shard_metadata=shard, node_id=NODE_A)
+        MODEL_A_ID: DownloadCompleted(
+            shard_metadata=shard, node_id=NODE_A, total_bytes=Memory()
+        )
     }
 
     # Global view hasn't caught up yet (no completed shards recorded for NODE_A)
@@ -192,10 +205,16 @@ def test_plan_does_not_load_model_until_all_shards_downloaded_globally():
 
     # Only NODE_A's shard is recorded as downloaded globally
     local_download_status = {
-        MODEL_A_ID: DownloadCompleted(shard_metadata=shard1, node_id=NODE_A)
+        MODEL_A_ID: DownloadCompleted(
+            shard_metadata=shard1, node_id=NODE_A, total_bytes=Memory()
+        )
     }
     global_download_status = {
-        NODE_A: [DownloadCompleted(shard_metadata=shard1, node_id=NODE_A)],
+        NODE_A: [
+            DownloadCompleted(
+                shard_metadata=shard1, node_id=NODE_A, total_bytes=Memory()
+            )
+        ],
         NODE_B: [],  # NODE_B has no downloads completed yet
     }
 
@@ -212,9 +231,15 @@ def test_plan_does_not_load_model_until_all_shards_downloaded_globally():
     assert result is None
 
     global_download_status = {
-        NODE_A: [DownloadCompleted(shard_metadata=shard1, node_id=NODE_A)],
+        NODE_A: [
+            DownloadCompleted(
+                shard_metadata=shard1, node_id=NODE_A, total_bytes=Memory()
+            )
+        ],
         NODE_B: [
-            DownloadCompleted(shard_metadata=shard2, node_id=NODE_B)
+            DownloadCompleted(
+                shard_metadata=shard2, node_id=NODE_B, total_bytes=Memory()
+            )
         ],  # NODE_B has no downloads completed yet
     }
 

src/exo/worker/tests/unittests/test_runner/test_event_ordering.py

@@ -111,7 +111,7 @@ def assert_events_equal(test_events: Iterable[Event], true_events: Iterable[Even
 def patch_out_mlx(monkeypatch: pytest.MonkeyPatch):
     # initialize_mlx returns a "group" equal to 1
     monkeypatch.setattr(mlx_runner, "initialize_mlx", make_nothin(1))
-    monkeypatch.setattr(mlx_runner, "load_mlx_items", make_nothin((1, 1, 1)))
+    monkeypatch.setattr(mlx_runner, "load_mlx_items", make_nothin((1, 1)))
     monkeypatch.setattr(mlx_runner, "warmup_inference", make_nothin(1))
     monkeypatch.setattr(mlx_runner, "_check_for_debug_prompts", nothin)
 

tests/headless_runner.py

@@ -49,14 +49,12 @@ class Tests(BaseModel):
     kind: typing.Literal["init", "warmup", "inference"]
 
 
-hn = socket.gethostname()
 mp.set_start_method("spawn", force=True)
 logger_setup(None)
 
 
 async def main():
     logger.info("starting cool server majig")
-    logger.info(hn)
     await assert_downloads()
     cfg = Config()
     cfg.bind = "0.0.0.0:52415"
@@ -81,20 +79,41 @@ async def main():
 async def assert_downloads():
     sd = exo_shard_downloader()
     # await sd.ensure_shard(await build_full_shard(MODEL_CARDS["qwen3-0.6b"].model_id))
-    await sd.ensure_shard(await build_full_shard(MODEL_CARDS["llama-3.2-1b"].model_id))
+    await sd.ensure_shard(
+        await build_full_shard(MODEL_CARDS["llama-3.1-8b-bf16"].model_id)
+    )
+    await sd.ensure_shard(await build_full_shard(MODEL_CARDS["qwen3-30b"].model_id))
+    await sd.ensure_shard(
+        await build_full_shard(MODEL_CARDS["gpt-oss-120b-MXFP4-Q8"].model_id)
+    )
+    await sd.ensure_shard(
+        await build_full_shard(MODEL_CARDS["gpt-oss-20b-4bit"].model_id)
+    )
+    await sd.ensure_shard(
+        await build_full_shard(MODEL_CARDS["glm-4.7-8bit-gs32"].model_id)
+    )
+    await sd.ensure_shard(
+        await build_full_shard(MODEL_CARDS["minimax-m2.1-8bit"].model_id)
+    )
 
 
 async def ring_backend(test: Tests):
     iid = InstanceId(str(hash(str(test.devs))))
-    return await execute_test(test, ring_instance(test, iid))
+    weird_hn = socket.gethostname()
+    for dev in test.devs:
+        if weird_hn.startswith(dev[0]) or dev[0].startswith(weird_hn):
+            hn = dev[0]
+            break
+    else:
+        raise ValueError(f"{weird_hn} not in {test.devs}")
+    return await execute_test(test, ring_instance(test, iid, hn), hn)
 
 
-def ring_instance(test: Tests, iid: InstanceId) -> Instance:
-    global hn
+def ring_instance(test: Tests, iid: InstanceId, hn: str) -> Instance:
     hbn = [Host(ip="i dont care", port=52416) for _ in test.devs]
     world_size = len(test.devs)
     for i in range(world_size):
-        if hn.startswith(test.devs[i][0]):
+        if test.devs[i][0] == hn:
             hn = test.devs[i][0]
             if i - 1 >= 0:
                 hbn[i - 1] = Host(ip=test.devs[i - 1][1], port=52416)
@@ -102,6 +121,8 @@ def ring_instance(test: Tests, iid: InstanceId) -> Instance:
                 hbn[i + 1] = Host(ip=test.devs[i + 1][1], port=52416)
             hbn[i] = Host(ip="0.0.0.0", port=52416)
             break
+    else:
+        raise ValueError(f"{hn} not in {test.devs}")
 
     meta = MODEL_CARDS[test.model_id].metadata
     instance = MlxRingInstance(
@@ -131,10 +152,10 @@ def ring_instance(test: Tests, iid: InstanceId) -> Instance:
     return instance
 
 
-async def execute_test(test: Tests, instance: Instance):
+async def execute_test(test: Tests, instance: Instance, hn: str):
     world_size = len(test.devs)
     iid = InstanceId(str(hash(str(test.devs))))
-    _handle, recv, send = new_runner(instance)
+    _handle, recv, send = new_runner(instance, hn)
     if world_size > 1:
         send.send(ConnectToGroup(instance_id=iid))
     send.send(LoadModel(instance_id=iid))
@@ -181,17 +202,19 @@ async def execute_test(test: Tests, instance: Instance):
 
 async def jaccl_backend(test: Tests):
     iid = InstanceId(str(hash(str(test.devs))))
-    return await execute_test(test, jaccl_instance(test, iid))
+    weird_hn = socket.gethostname()
+    for dev in test.devs:
+        if weird_hn.startswith(dev[0]) or dev[0].startswith(weird_hn):
+            hn = dev[0]
+            break
+    else:
+        raise ValueError(f"{weird_hn} not in {test.devs}")
+    return await execute_test(test, jaccl_instance(test, iid, hn), hn)
 
 
-def jaccl_instance(test: Tests, iid: InstanceId):
-    global hn
+def jaccl_instance(test: Tests, iid: InstanceId, hn: str):
     meta = MODEL_CARDS[test.model_id].metadata
     world_size = len(test.devs)
-    for name, _ in test.devs:
-        if hn.startswith(name):
-            hn = name
-            break
 
     return MlxJacclInstance(
         instance_id=iid,
@@ -220,6 +243,7 @@ def jaccl_instance(test: Tests, iid: InstanceId):
 
 def new_runner(
     instance: Instance,
+    hn: str,
 ) -> tuple[mp.Process, MpReceiver[Event], MpSender[Task]]:
     bound_instance = BoundInstance(
         instance=instance, bound_runner_id=RunnerId(hn), bound_node_id=NodeId(hn)

tests/start_distributed_test.sh

@@ -34,19 +34,23 @@ done
 devs_raw=$(printf "[\"%s\", \"%s\"], " "${weaved[@]}")
 devs="[${devs_raw%, }]"
 
-for i in "${!ips[@]}"; do  
-  { 
-    req="{
-      \"model_id\": \"llama-3.2-1b\",
-      \"devs\": ${devs},
-      \"kind\": \"inference\"
-     }"
-    echo "req $req"
-    curl -sN \
-      -X POST "http://${ips[$i]}:52415/${kind}" \
-      -H "Content-Type: application/json" -d "$req" \
-    2>&1 | sed "s/^/\n${hostnames[$i]}@${ips[$i]}: /" || echo "curl to ${hostnames[$i]} failed"
-  } &
+model_ids=("qwen3-30b" "gpt-oss-120b-MXFP4-Q8" "kimi-k2-thinking")
+
+for model_id in "${model_ids[@]}"; do
+  for i in "${!ips[@]}"; do  
+    { 
+      req="{
+        \"model_id\": \"${model_id}\",
+        \"devs\": ${devs},
+        \"kind\": \"inference\"
+       }"
+      echo "req $req"
+      curl -sN \
+        -X POST "http://${ips[$i]}:52415/${kind}" \
+        -H "Content-Type: application/json" -d "$req" \
+      2>&1 | sed "s/^/\n${hostnames[$i]}@${ips[$i]}: /" || echo "curl to ${hostnames[$i]} failed" && exit 1
+    } &
+  done
+  wait
 done
 
-wait