woahg

better cancellation (#1388 )
a lot of our cleanup logic wasn't running leading to bad shutdown states ## changes - added `try: except` blocks around most task groups - made the runner shutdown code synchronous - abandon the MpReceiver's recv_async thread on cancellation - this only occurs during runner shutdown, the queue closing from the other end should terminate the mp.Queue, cleaning up the thread in its own time. i could try other methods if this is not sufficient. ## outcome ctrl-c just works now! minus the tokio panic of course :) no more hypercorn lifespan errors though!
2026-02-05 11:43:17 -05:00 · 2026-02-05 15:24:57 +00:00 · 2026-02-05 15:22:33 +00:00 · 2026-02-05 15:15:52 +00:00
55 changed files with 974 additions and 3608 deletions
--- a/Cargo.lock
+++ b/Cargo.lock
@@ -141,12 +141,6 @@ version = "0.3.9"
 source = "registry+https://github.com/rust-lang/crates.io-index"
 checksum = "76a2e8124351fda1ef8aaaa3bbd7ebbcb486bbcd4225aca0aa0d84bb2db8fecb"

-[[package]]
-name = "arrayvec"
-version = "0.7.6"
-source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "7c02d123df017efcdfbd739ef81735b36c5ba83ec3c59c80a9d7ecc718f92e50"
-
 [[package]]
 name = "asn1-rs"
 version = "0.7.1"
@@ -304,19 +298,6 @@ version = "1.8.0"
 source = "registry+https://github.com/rust-lang/crates.io-index"
 checksum = "55248b47b0caf0546f7988906588779981c43bb1bc9d0c44087278f80cdb44ba"

-[[package]]
-name = "bigdecimal"
-version = "0.4.9"
-source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "560f42649de9fa436b73517378a147ec21f6c997a546581df4b4b31677828934"
-dependencies = [
- "autocfg",
- "libm",
- "num-bigint",
- "num-integer",
- "num-traits",
-]
-
 [[package]]
 name = "bimap"
 version = "0.6.3"
@@ -353,31 +334,6 @@ dependencies = [
 "generic-array",
 ]

-[[package]]
-name = "bon"
-version = "3.8.1"
-source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "ebeb9aaf9329dff6ceb65c689ca3db33dbf15f324909c60e4e5eef5701ce31b1"
-dependencies = [
- "bon-macros",
- "rustversion",
-]
-
-[[package]]
-name = "bon-macros"
-version = "3.8.1"
-source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "77e9d642a7e3a318e37c2c9427b5a6a48aa1ad55dcd986f3034ab2239045a645"
-dependencies = [
- "darling",
- "ident_case",
- "prettyplease",
- "proc-macro2",
- "quote",
- "rustversion",
- "syn 2.0.111",
-]
-
 [[package]]
 name = "bs58"
 version = "0.5.1"
@@ -541,15 +497,6 @@ version = "0.4.3"
 source = "registry+https://github.com/rust-lang/crates.io-index"
 checksum = "2f421161cb492475f1661ddc9815a745a1c894592070661180fdec3d4872e9c3"

-[[package]]
-name = "convert_case"
-version = "0.10.0"
-source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "633458d4ef8c78b72454de2d54fd6ab2e60f9e02be22f3c6104cdc8a4e0fceb9"
-dependencies = [
- "unicode-segmentation",
-]
-
 [[package]]
 name = "core-foundation"
 version = "0.9.4"
@@ -700,41 +647,6 @@ dependencies = [
 "syn 2.0.111",
 ]

-[[package]]
-name = "darling"
-version = "0.21.3"
-source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "9cdf337090841a411e2a7f3deb9187445851f91b309c0c0a29e05f74a00a48c0"
-dependencies = [
- "darling_core",
- "darling_macro",
-]
-
-[[package]]
-name = "darling_core"
-version = "0.21.3"
-source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "1247195ecd7e3c85f83c8d2a366e4210d588e802133e1e355180a9870b517ea4"
-dependencies = [
- "fnv",
- "ident_case",
- "proc-macro2",
- "quote",
- "strsim",
- "syn 2.0.111",
-]
-
-[[package]]
-name = "darling_macro"
-version = "0.21.3"
-source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "d38308df82d1080de0afee5d069fa14b0326a88c14f15c5ccda35b4a6c414c81"
-dependencies = [
- "darling_core",
- "quote",
- "syn 2.0.111",
-]
-
 [[package]]
 name = "data-encoding"
 version = "2.9.0"
@@ -761,17 +673,6 @@ dependencies = [
 "syn 2.0.111",
 ]

-[[package]]
-name = "delegate"
-version = "0.13.5"
-source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "780eb241654bf097afb00fc5f054a09b687dad862e485fdcf8399bb056565370"
-dependencies = [
- "proc-macro2",
- "quote",
- "syn 2.0.111",
-]
-
 [[package]]
 name = "der"
 version = "0.7.10"
@@ -806,29 +707,6 @@ dependencies = [
 "powerfmt",
 ]

-[[package]]
-name = "derive_more"
-version = "2.1.0"
-source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "10b768e943bed7bf2cab53df09f4bc34bfd217cdb57d971e769874c9a6710618"
-dependencies = [
- "derive_more-impl",
-]
-
-[[package]]
-name = "derive_more-impl"
-version = "2.1.0"
-source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "6d286bfdaf75e988b4a78e013ecd79c581e06399ab53fbacd2d916c2f904f30b"
-dependencies = [
- "convert_case",
- "proc-macro2",
- "quote",
- "rustc_version",
- "syn 2.0.111",
- "unicode-xid",
-]
-
 [[package]]
 name = "digest"
 version = "0.10.7"
@@ -998,37 +876,23 @@ dependencies = [
 name = "exo_pyo3_bindings"
 version = "0.0.1"
 dependencies = [
- "delegate",
- "derive_more",
 "env_logger",
- "extend",
- "futures",
- "impl-trait-for-tuples",
+ "futures-lite",
 "libp2p",
 "log",
 "networking",
- "once_cell",
- "pin-project",
 "pyo3",
 "pyo3-async-runtimes",
 "pyo3-log",
 "pyo3-stub-gen",
- "thiserror 2.0.17",
- "thread_local",
 "tokio",
- "util",
 ]

 [[package]]
-name = "extend"
-version = "1.2.0"
+name = "fastrand"
+version = "2.3.0"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "311a6d2f1f9d60bff73d2c78a0af97ed27f79672f15c238192a5bbb64db56d00"
-dependencies = [
- "proc-macro2",
- "quote",
- "syn 2.0.111",
-]
+checksum = "37909eebbb50d72f9059c3b6d82c0463f2ff062c9e95845c43a6c9c0355411be"

 [[package]]
 name = "ff"
@@ -1138,7 +1002,10 @@ version = "2.6.1"
 source = "registry+https://github.com/rust-lang/crates.io-index"
 checksum = "f78e10609fe0e0b3f4157ffab1876319b5b0db102a2c60dc4626306dc46b44ad"
 dependencies = [
+ "fastrand",
 "futures-core",
+ "futures-io",
+ "parking",
 "pin-project-lite",
 ]

@@ -1625,12 +1492,6 @@ dependencies = [
 "zerovec",
 ]

-[[package]]
-name = "ident_case"
-version = "1.0.1"
-source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "b9e0384b61958566e926dc50660321d12159025e767c18e043daf26b70104c39"
-
 [[package]]
 name = "idna"
 version = "1.1.0"
@@ -1706,17 +1567,6 @@ dependencies = [
 "xmltree",
 ]

-[[package]]
-name = "impl-trait-for-tuples"
-version = "0.2.3"
-source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "a0eb5a3343abf848c0984fe4604b2b105da9539376e24fc0a3b0007411ae4fd9"
-dependencies = [
- "proc-macro2",
- "quote",
- "syn 2.0.111",
-]
-
 [[package]]
 name = "indexmap"
 version = "2.12.1"
@@ -1745,15 +1595,6 @@ dependencies = [
 "generic-array",
 ]

-[[package]]
-name = "internment"
-version = "0.8.6"
-source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "636d4b0f6a39fd684effe2a73f5310df16a3fa7954c26d36833e98f44d1977a2"
-dependencies = [
- "hashbrown 0.15.5",
-]
-
 [[package]]
 name = "inventory"
 version = "0.3.21"
@@ -1880,12 +1721,6 @@ dependencies = [
 "cpufeatures",
 ]

-[[package]]
-name = "keccak-const"
-version = "0.2.0"
-source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "57d8d8ce877200136358e0bbff3a77965875db3af755a11e1fa6b1b3e2df13ea"
-
 [[package]]
 name = "lalrpop-util"
 version = "0.20.2"
@@ -1904,12 +1739,6 @@ version = "0.2.178"
 source = "registry+https://github.com/rust-lang/crates.io-index"
 checksum = "37c93d8daa9d8a012fd8ab92f088405fb202ea0b6ab73ee2482ae66af4f42091"

-[[package]]
-name = "libm"
-version = "0.2.15"
-source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "f9fbbcab51052fe104eb5e5d351cf728d30a5be1fe14d9be8a3b097481fb97de"
-
 [[package]]
 name = "libp2p"
 version = "0.56.0"
@@ -2898,20 +2727,10 @@ dependencies = [
 name = "networking"
 version = "0.0.1"
 dependencies = [
- "delegate",
- "derive_more",
- "either",
- "extend",
- "futures",
- "futures-timer",
- "impl-trait-for-tuples",
- "keccak-const",
 "libp2p",
 "log",
- "thiserror 2.0.17",
 "tokio",
 "tracing-subscriber",
- "util",
 ]

 [[package]]
@@ -2993,17 +2812,6 @@ dependencies = [
 "num-traits",
 ]

-[[package]]
-name = "num-rational"
-version = "0.4.2"
-source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "f83d14da390562dca69fc84082e73e548e1ad308d24accdedd2720017cb37824"
-dependencies = [
- "num-bigint",
- "num-integer",
- "num-traits",
-]
-
 [[package]]
 name = "num-traits"
 version = "0.2.19"
@@ -3307,16 +3115,6 @@ dependencies = [
 "zerocopy",
 ]

-[[package]]
-name = "prettyplease"
-version = "0.2.37"
-source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "479ca8adacdd7ce8f1fb39ce9ecccbfe93a3f1344b3d0d97f20bc0196208f62b"
-dependencies = [
- "proc-macro2",
- "syn 2.0.111",
-]
-
 [[package]]
 name = "primeorder"
 version = "0.13.6"
@@ -3364,28 +3162,14 @@ version = "0.27.2"
 source = "registry+https://github.com/rust-lang/crates.io-index"
 checksum = "ab53c047fcd1a1d2a8820fe84f05d6be69e9526be40cb03b73f86b6b03e6d87d"
 dependencies = [
- "bigdecimal",
- "either",
- "hashbrown 0.16.1",
- "indexmap",
 "indoc",
- "inventory",
 "libc",
- "lock_api",
 "memoffset",
- "num-bigint",
- "num-complex",
- "num-rational",
- "num-traits",
 "once_cell",
- "ordered-float",
- "parking_lot",
 "portable-atomic",
 "pyo3-build-config",
 "pyo3-ffi",
 "pyo3-macros",
- "rust_decimal",
- "smallvec",
 "unindent",
 ]

@@ -3740,12 +3524,6 @@ dependencies = [
 "yasna",
 ]

-[[package]]
-name = "recursion"
-version = "0.5.4"
-source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "9dba2197bf7b1d87b4dd460c195f4edeb45a94e82e8054f8d5f317c1f0e93ca1"
-
 [[package]]
 name = "redox_syscall"
 version = "0.5.18"
@@ -3832,16 +3610,6 @@ dependencies = [
 "tokio",
 ]

-[[package]]
-name = "rust_decimal"
-version = "1.39.0"
-source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "35affe401787a9bd846712274d97654355d21b2a2c092a3139aabe31e9022282"
-dependencies = [
- "arrayvec",
- "num-traits",
-]
-
 [[package]]
 name = "rustc-hash"
 version = "1.1.0"
@@ -4706,24 +4474,12 @@ version = "1.0.22"
 source = "registry+https://github.com/rust-lang/crates.io-index"
 checksum = "9312f7c4f6ff9069b165498234ce8be658059c6728633667c526e27dc2cf1df5"

-[[package]]
-name = "unicode-segmentation"
-version = "1.12.0"
-source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "f6ccf251212114b54433ec949fd6a7841275f9ada20dddd2f29e9ceea4501493"
-
 [[package]]
 name = "unicode-width"
 version = "0.2.2"
 source = "registry+https://github.com/rust-lang/crates.io-index"
 checksum = "b4ac048d71ede7ee76d585517add45da530660ef4390e49b098733c6e897f254"

-[[package]]
-name = "unicode-xid"
-version = "0.2.6"
-source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "ebc1c04c71510c7f702b52b7c350734c9ff1295c464a03335b00bb84fc54f853"
-
 [[package]]
 name = "unicode_names2"
 version = "1.3.0"
@@ -4804,19 +4560,6 @@ version = "0.2.2"
 source = "registry+https://github.com/rust-lang/crates.io-index"
 checksum = "06abde3611657adf66d383f00b093d7faecc7fa57071cce2578660c9f1010821"

-[[package]]
-name = "util"
-version = "0.0.1"
-dependencies = [
- "bon",
- "derive_more",
- "extend",
- "internment",
- "once_cell",
- "recursion",
- "thiserror 2.0.17",
-]
-
 [[package]]
 name = "uuid"
 version = "1.19.0"
--- a/Cargo.toml
+++ b/Cargo.toml
@@ -3,7 +3,6 @@ resolver = "3"
 members = [
    "rust/networking",
    "rust/exo_pyo3_bindings",
-    "rust/util",
 ]

 [workspace.package]
@@ -24,62 +23,18 @@ opt-level = 3
 [workspace.dependencies]
 ## Crate members as common dependencies
 networking = { path = "rust/networking" }
-util = { path = "rust/util" }
-
-# Proc-macro authoring tools
-syn = "2.0"
-quote = "1.0"
-proc-macro2 = "1.0"
-darling = "0.20"
-
-# Macro dependecies
-extend = "1.2"
-delegate = "0.13"
-impl-trait-for-tuples = "0.2"
-clap = "4.5"
-derive_more = { version = "2.0.1", features = ["display"] }
-pin-project = "1"
-
-# Utility dependencies
-itertools = "0.14"
-thiserror = "2"
-internment = "0.8"
-recursion = "0.5"
-regex = "1.11"
-once_cell = "1.21"
-thread_local = "1.1"
-bon = "3.4"
-generativity = "1.1"
-anyhow = "1.0"
-keccak-const = "0.2"
-
-# Functional generics/lenses frameworks
-frunk_core = "0.4"
-frunk = "0.4"
-frunk_utils = "0.2"
-frunk-enum-core = "0.3"

 # Async dependencies
 tokio = "1.46"
-futures = "0.3"
-futures-util = "0.3"
-futures-timer = "3.0"
-
-# Data structures
-either = "1.15"
-ordered-float = "5.0"
-ahash = "0.8"

 # Tracing/logging
 log = "0.4"

 # networking
 libp2p = "0.56"
-libp2p-tcp = "0.44"

 [workspace.lints.rust]
-static_mut_refs = "warn"      # Or use "warn" instead of deny
-incomplete_features = "allow"
+static_mut_refs = "warn"

 # Clippy's lint category level configurations;
 # every member crate needs to inherit these by adding
@@ -100,64 +55,3 @@ perf = { level = "warn", priority = -1 }
 pedantic = { level = "warn", priority = -1 }
 nursery = { level = "warn", priority = -1 }
 cargo = { level = "warn", priority = -1 }
-
-# Individual Clippy lints from the `restriction` category
-arithmetic_side_effects = "warn"
-as_conversions = "warn"
-assertions_on_result_states = "warn"
-clone_on_ref_ptr = "warn"
-decimal_literal_representation = "warn"
-default_union_representation = "warn"
-deref_by_slicing = "warn"
-disallowed_script_idents = "deny"
-else_if_without_else = "warn"
-empty_enum_variants_with_brackets = "warn"
-empty_structs_with_brackets = "warn"
-error_impl_error = "warn"
-exit = "deny"
-expect_used = "warn"
-float_cmp_const = "warn"
-get_unwrap = "warn"
-if_then_some_else_none = "warn"
-impl_trait_in_params = "warn"
-indexing_slicing = "warn"
-infinite_loop = "warn"
-let_underscore_must_use = "warn"
-let_underscore_untyped = "warn"
-lossy_float_literal = "warn"
-mem_forget = "warn"
-missing_inline_in_public_items = "warn"
-multiple_inherent_impl = "warn"
-multiple_unsafe_ops_per_block = "warn"
-mutex_atomic = "warn"
-non_zero_suggestions = "warn"
-panic = "warn"
-partial_pub_fields = "warn"
-pattern_type_mismatch = "warn"
-pub_without_shorthand = "warn"
-rc_buffer = "warn"
-rc_mutex = "warn"
-redundant_type_annotations = "warn"
-renamed_function_params = "warn"
-rest_pat_in_fully_bound_structs = "warn"
-same_name_method = "warn"
-self_named_module_files = "deny"
-semicolon_inside_block = "warn"
-shadow_same = "warn"
-shadow_unrelated = "warn"
-str_to_string = "warn"
-string_add = "warn"
-string_lit_chars_any = "warn"
-string_to_string = "warn"
-tests_outside_test_module = "warn"
-todo = "warn"
-try_err = "warn"
-undocumented_unsafe_blocks = "warn"
-unnecessary_safety_comment = "warn"
-unnecessary_safety_doc = "warn"
-unneeded_field_pattern = "warn"
-unseparated_literal_suffix = "warn"
-unused_result_ok = "warn"
-unused_trait_names = "warn"
-unwrap_used = "warn"
-verbose_file_reads = "warn"
--- a/MISSED_THINGS.md
+++ b/MISSED_THINGS.md
@@ -5,21 +5,21 @@
 [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] no mx_barrier in genreate.py mlx_generate at the end.
+[] no mx_barrier in genreate.py mlx_generate at the end.
 [] cache assertion not needed in auto_parallel.py PipelineLastLayer.
-[X] GPTOSS support dropped in auto_parallel.py.
-[X] sharding changed "all-to-sharded" became _all_to_sharded in auto_parallel.py.
-[X] same as above with "sharded-to-all" became _sharded_to_all in auto_parallel.py.
-[X] Dropped support for Ministral3Model, DeepseekV32Model, Glm4MoeModel, Qwen3NextModel, GptOssMode in auto_parallel.py.
+[] 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.
-[X] Dropped _set_nofile_limit in utils_mlx.py.
-[X] We have group optional in load_mlx_items in utils_mlx.py.
-[X] Dropped add_missing_chat_templates for GptOss in load_mlx_items in utils_mlx.py.
-[X] Dropped model.make_cache in make_kv_cache in utils_mlx.py.
+[] 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.
-[X] 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?
-[X] 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.)
+[] 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).
--- a/dashboard/src/lib/components/ChatForm.svelte
+++ b/dashboard/src/lib/components/ChatForm.svelte
@@ -1,7 +1,6 @@
 <script lang="ts">
  import {
    isLoading,
-    stopGeneration,
    sendMessage,
    generateImage,
    editImage,
@@ -606,92 +605,86 @@
        style="min-height: 28px; max-height: 150px;"
      ></textarea>

-      {#if loading}
-        <button
-          type="button"
-          onclick={() => stopGeneration()}
-          class="px-2.5 sm:px-4 py-1.5 sm:py-2 rounded text-xs sm:text-xs tracking-[0.1em] sm:tracking-[0.15em] uppercase font-medium transition-all duration-200 whitespace-nowrap bg-exo-medium-gray/70 text-exo-light-gray hover:bg-red-900/50 hover:text-red-400 border border-exo-medium-gray/50 hover:border-red-500/50 cursor-pointer"
-          aria-label="Stop generation"
-        >
+      <button
+        type="submit"
+        disabled={!canSend || loading || isEditOnlyWithoutImage}
+        class="px-2.5 sm:px-4 py-1.5 sm:py-2 rounded text-xs sm:text-xs tracking-[0.1em] sm:tracking-[0.15em] uppercase font-medium transition-all duration-200 whitespace-nowrap
+					{!canSend || loading || isEditOnlyWithoutImage
+          ? 'bg-exo-medium-gray/50 text-exo-light-gray cursor-not-allowed'
+          : 'bg-exo-yellow text-exo-black hover:bg-exo-yellow-darker hover:shadow-[0_0_20px_rgba(255,215,0,0.3)]'}"
+        aria-label={shouldShowEditMode
+          ? "Edit image"
+          : isImageModel()
+            ? "Generate image"
+            : "Send message"}
+      >
+        {#if loading}
          <span class="inline-flex items-center gap-1 sm:gap-2">
-            <svg
-              class="w-2.5 h-2.5 sm:w-3 sm:h-3"
-              viewBox="0 0 24 24"
-              fill="currentColor"
+            <span
+              class="w-2.5 h-2.5 sm:w-3 sm:h-3 border-2 border-current border-t-transparent rounded-full animate-spin"
+            ></span>
+            <span class="hidden sm:inline"
+              >{shouldShowEditMode
+                ? "EDITING"
+                : isImageModel()
+                  ? "GENERATING"
+                  : "PROCESSING"}</span
            >
-              <rect x="4" y="4" width="16" height="16" rx="2" />
-            </svg>
-            <span class="hidden sm:inline">STOP</span>
+            <span class="sm:hidden">...</span>
          </span>
-        </button>
-      {:else}
-        <button
-          type="submit"
-          disabled={!canSend || isEditOnlyWithoutImage}
-          class="px-2.5 sm:px-4 py-1.5 sm:py-2 rounded text-xs sm:text-xs tracking-[0.1em] sm:tracking-[0.15em] uppercase font-medium transition-all duration-200 whitespace-nowrap
-            {!canSend || isEditOnlyWithoutImage
-            ? 'bg-exo-medium-gray/50 text-exo-light-gray cursor-not-allowed'
-            : 'bg-exo-yellow text-exo-black hover:bg-exo-yellow-darker hover:shadow-[0_0_20px_rgba(255,215,0,0.3)]'}"
-          aria-label={shouldShowEditMode
-            ? "Edit image"
-            : isImageModel()
-              ? "Generate image"
-              : "Send message"}
-        >
-          {#if shouldShowEditMode}
-            <span class="inline-flex items-center gap-1.5">
-              <svg
-                class="w-3.5 h-3.5"
-                fill="none"
-                viewBox="0 0 24 24"
-                stroke="currentColor"
-                stroke-width="2"
-              >
-                <path
-                  stroke-linecap="round"
-                  stroke-linejoin="round"
-                  d="M11 5H6a2 2 0 00-2 2v11a2 2 0 002 2h11a2 2 0 002-2v-5m-1.414-9.414a2 2 0 112.828 2.828L11.828 15H9v-2.828l8.586-8.586z"
-                />
-              </svg>
-              <span>EDIT</span>
-            </span>
-          {:else if isEditOnlyWithoutImage}
-            <span class="inline-flex items-center gap-1.5">
-              <svg
-                class="w-3.5 h-3.5"
-                fill="none"
-                viewBox="0 0 24 24"
-                stroke="currentColor"
-                stroke-width="2"
-              >
-                <path
-                  stroke-linecap="round"
-                  stroke-linejoin="round"
-                  d="M11 5H6a2 2 0 00-2 2v11a2 2 0 002 2h11a2 2 0 002-2v-5m-1.414-9.414a2 2 0 112.828 2.828L11.828 15H9v-2.828l8.586-8.586z"
-                />
-              </svg>
-              <span>EDIT</span>
-            </span>
-          {:else if isImageModel()}
-            <span class="inline-flex items-center gap-1.5">
-              <svg
-                class="w-3.5 h-3.5"
-                fill="none"
-                viewBox="0 0 24 24"
-                stroke="currentColor"
-                stroke-width="2"
-              >
-                <rect x="3" y="3" width="18" height="18" rx="2" ry="2" />
-                <circle cx="8.5" cy="8.5" r="1.5" />
-                <polyline points="21 15 16 10 5 21" />
-              </svg>
-              <span>GENERATE</span>
-            </span>
-          {:else}
-            SEND
-          {/if}
-        </button>
-      {/if}
+        {:else if shouldShowEditMode}
+          <span class="inline-flex items-center gap-1.5">
+            <svg
+              class="w-3.5 h-3.5"
+              fill="none"
+              viewBox="0 0 24 24"
+              stroke="currentColor"
+              stroke-width="2"
+            >
+              <path
+                stroke-linecap="round"
+                stroke-linejoin="round"
+                d="M11 5H6a2 2 0 00-2 2v11a2 2 0 002 2h11a2 2 0 002-2v-5m-1.414-9.414a2 2 0 112.828 2.828L11.828 15H9v-2.828l8.586-8.586z"
+              />
+            </svg>
+            <span>EDIT</span>
+          </span>
+        {:else if isEditOnlyWithoutImage}
+          <span class="inline-flex items-center gap-1.5">
+            <svg
+              class="w-3.5 h-3.5"
+              fill="none"
+              viewBox="0 0 24 24"
+              stroke="currentColor"
+              stroke-width="2"
+            >
+              <path
+                stroke-linecap="round"
+                stroke-linejoin="round"
+                d="M11 5H6a2 2 0 00-2 2v11a2 2 0 002 2h11a2 2 0 002-2v-5m-1.414-9.414a2 2 0 112.828 2.828L11.828 15H9v-2.828l8.586-8.586z"
+              />
+            </svg>
+            <span>EDIT</span>
+          </span>
+        {:else if isImageModel()}
+          <span class="inline-flex items-center gap-1.5">
+            <svg
+              class="w-3.5 h-3.5"
+              fill="none"
+              viewBox="0 0 24 24"
+              stroke="currentColor"
+              stroke-width="2"
+            >
+              <rect x="3" y="3" width="18" height="18" rx="2" ry="2" />
+              <circle cx="8.5" cy="8.5" r="1.5" />
+              <polyline points="21 15 16 10 5 21" />
+            </svg>
+            <span>GENERATE</span>
+          </span>
+        {:else}
+          SEND
+        {/if}
+      </button>
    </div>

    <!-- Bottom accent line -->
--- a/dashboard/src/lib/stores/app.svelte.ts
+++ b/dashboard/src/lib/stores/app.svelte.ts
@@ -470,7 +470,6 @@ class AppStore {
  messages = $state<Message[]>([]);
  currentResponse = $state("");
  isLoading = $state(false);
-  private currentAbortController: AbortController | null = null;

  // Performance metrics
  ttftMs = $state<number | null>(null); // Time to first token in ms
@@ -1739,11 +1738,9 @@ class AppStore {
        return;
      }

-      this.currentAbortController = new AbortController();
      const response = await fetch("/v1/chat/completions", {
        method: "POST",
        headers: { "Content-Type": "application/json" },
-        signal: this.currentAbortController.signal,
        body: JSON.stringify({
          model: modelToUse,
          messages: apiMessages,
@@ -1857,7 +1854,6 @@ class AppStore {
        "Unknown error",
      );
    } finally {
-      this.currentAbortController = null;
      this.isLoading = false;
      this.currentResponse = "";
      this.saveConversationsToStorage();
@@ -1989,10 +1985,6 @@ class AppStore {
    assistantMessageId: string,
    errorPrefix = "Failed to get response",
  ): void {
-    // Don't show error for user-initiated abort (stop button)
-    if (error instanceof DOMException && error.name === "AbortError") {
-      return;
-    }
    if (this.conversationExists(targetConversationId)) {
      this.updateConversationMessage(
        targetConversationId,
@@ -2034,17 +2026,6 @@ class AppStore {
    return null;
  }

-  /**
-   * Stop the current generation by aborting the HTTP connection.
-   * This triggers backend cancellation via the mechanism in PR #1276.
-   */
-  stopGeneration() {
-    if (this.currentAbortController) {
-      this.currentAbortController.abort();
-      this.currentAbortController = null;
-    }
-  }
-
  /**
   * Send a message to the LLM and stream the response
   */
@@ -2192,13 +2173,11 @@ class AppStore {
      let firstTokenTime: number | null = null;
      let tokenCount = 0;

-      this.currentAbortController = new AbortController();
      const response = await fetch("/v1/chat/completions", {
        method: "POST",
        headers: {
          "Content-Type": "application/json",
        },
-        signal: this.currentAbortController.signal,
        body: JSON.stringify({
          model: modelToUse,
          messages: apiMessages,
@@ -2346,7 +2325,6 @@ class AppStore {
        "Failed to get response",
      );
    } finally {
-      this.currentAbortController = null;
      this.isLoading = false;
      this.currentResponse = "";
      this.saveConversationsToStorage();
@@ -2446,13 +2424,11 @@ class AppStore {
        };
      }

-      this.currentAbortController = new AbortController();
      const response = await fetch("/v1/images/generations", {
        method: "POST",
        headers: {
          "Content-Type": "application/json",
        },
-        signal: this.currentAbortController.signal,
        body: JSON.stringify(requestBody),
      });

@@ -2601,7 +2577,6 @@ class AppStore {
        "Failed to generate image",
      );
    } finally {
-      this.currentAbortController = null;
      this.isLoading = false;
      this.saveConversationsToStorage();
    }
@@ -2730,10 +2705,8 @@ class AppStore {
        );
      }

-      this.currentAbortController = new AbortController();
      const apiResponse = await fetch("/v1/images/edits", {
        method: "POST",
-        signal: this.currentAbortController.signal,
        body: formData,
      });

@@ -2843,7 +2816,6 @@ class AppStore {
        "Failed to edit image",
      );
    } finally {
-      this.currentAbortController = null;
      this.isLoading = false;
      this.saveConversationsToStorage();
    }
@@ -2972,7 +2944,6 @@ export const hasStartedChat = () => appStore.hasStartedChat;
 export const messages = () => appStore.messages;
 export const currentResponse = () => appStore.currentResponse;
 export const isLoading = () => appStore.isLoading;
-export const stopGeneration = () => appStore.stopGeneration();
 export const ttftMs = () => appStore.ttftMs;
 export const tps = () => appStore.tps;
 export const totalTokens = () => appStore.totalTokens;
--- a/flake.nix
+++ b/flake.nix
@@ -118,9 +118,10 @@
            {
              metal-toolchain = pkgs.callPackage ./nix/metal-toolchain.nix { };
              mlx = pkgs.callPackage ./nix/mlx.nix {
-                metal-toolchain = self'.packages.metal-toolchain;
+                inherit (self'.packages) metal-toolchain;
                inherit uvLockMlxVersion;
              };
+              default = self'.packages.exo;
            }
          );

--- a/rust/clippy.toml
+++ b/rust/clippy.toml
@@ -1,2 +0,0 @@
-# we can manually exclude false-positive lint errors for dual packages (if in dependencies)
-#allowed-duplicate-crates = ["hashbrown"]
--- a/rust/exo_pyo3_bindings/Cargo.toml
+++ b/rust/exo_pyo3_bindings/Cargo.toml
@@ -25,44 +25,25 @@ workspace = true
 networking = { workspace = true }

 # interop
-pyo3 = { version = "0.27.1", features = [
-    # "abi3-py311", # tells pyo3 (and maturin) to build using the stable ABI with minimum Python version 3.11
-    "nightly", # enables better-supported GIL integration
-    "experimental-async", # async support in #[pyfunction] & #[pymethods]
-    #"experimental-inspect", # inspection of generated binary => easier to automate type-hint generation
-    #"py-clone", # adding Clone-ing of `Py<T>` without GIL (may cause panics - remove if panics happen)
-    "multiple-pymethods", # allows multiple #[pymethods] sections per class
+pyo3 = { version = "0.27.2", features = [
+    "abi3-py313", # tells pyo3 (and maturin) to build using the stable ABI with minimum Python version 3.11
+    # "nightly", # enables better-supported GIL integration
+    "experimental-async" # async support in #[pyfunction] & #[pymethods]
+    # "experimental-inspect", # inspection of generated binary => easier to automate type-hint generation
+    # "py-clone", # adding Clone-ing of `Py<T>` without GIL (may cause panics - remove if panics happen)
+    # "multiple-pymethods", # allows multiple #[pymethods] sections per class

    # integrations with other libraries
-    "arc_lock", "bigdecimal", "either", "hashbrown", "indexmap", "num-bigint", "num-complex", "num-rational",
-    "ordered-float", "rust_decimal", "smallvec",
+    # "arc_lock", "bigdecimal", "either", "hashbrown", "indexmap", "num-bigint", "num-complex", "num-rational",
+    # "ordered-float", "rust_decimal", "smallvec",
    # "anyhow", "chrono", "chrono-local", "chrono-tz", "eyre", "jiff-02", "lock_api", "parking-lot", "time",  "serde",
 ] }
 pyo3-stub-gen = { version = "0.17.2" }
 pyo3-async-runtimes = { version = "0.27.0", features = ["attributes", "tokio-runtime", "testing"] }
 pyo3-log = "0.13.2"

-# macro dependencies
-extend = { workspace = true }
-delegate = { workspace = true }
-impl-trait-for-tuples = { workspace = true }
-derive_more = { workspace = true }
-pin-project = { workspace = true }
-
 # async runtime
 tokio = { workspace = true, features = ["full", "tracing"] }
-futures = { workspace = true }
-
-# utility dependencies
-once_cell = "1.21.3"
-thread_local = "1.1.9"
-util = { workspace = true }
-thiserror = { workspace = true }
-#internment = { workspace = true }
-#recursion = { workspace = true }
-#generativity = { workspace = true }
-#itertools = { workspace = true }
-

 # Tracing
 #tracing = "0.1"
@@ -75,3 +56,4 @@ env_logger = "0.11"

 # Networking
 libp2p = { workspace = true, features = ["full"] }
+futures-lite = "2.6.1"
--- a/rust/exo_pyo3_bindings/exo_pyo3_bindings.pyi
+++ b/rust/exo_pyo3_bindings/exo_pyo3_bindings.pyi
@@ -2,220 +2,39 @@
 # ruff: noqa: E501, F401

 import builtins
-import enum
 import typing

-@typing.final
-class AllQueuesFullError(builtins.Exception):
-    def __new__(cls, *args: typing.Any) -> AllQueuesFullError: ...
-    def __repr__(self) -> builtins.str: ...
-    def __str__(self) -> builtins.str: ...
-
-@typing.final
-class ConnectionUpdate:
-    @property
-    def update_type(self) -> ConnectionUpdateType:
-        r"""
-        Whether this is a connection or disconnection event
-        """
-    @property
-    def peer_id(self) -> PeerId:
-        r"""
-        Identity of the peer that we have connected to or disconnected from.
-        """
-    @property
-    def remote_ipv4(self) -> builtins.str:
-        r"""
-        Remote connection's IPv4 address.
-        """
-    @property
-    def remote_tcp_port(self) -> builtins.int:
-        r"""
-        Remote connection's TCP port.
-        """
-
@typing.final
 class Keypair:
-    r"""
-    Identity keypair of a node.
-    """
    @staticmethod
-    def generate_ed25519() -> Keypair:
+    def generate() -> Keypair:
        r"""
-        Generate a new Ed25519 keypair.
-        """
-    @staticmethod
-    def generate_ecdsa() -> Keypair:
-        r"""
-        Generate a new ECDSA keypair.
-        """
-    @staticmethod
-    def generate_secp256k1() -> Keypair:
-        r"""
-        Generate a new Secp256k1 keypair.
+        Generate a new ed25519 keypair
        """
    @staticmethod
    def from_protobuf_encoding(bytes: bytes) -> Keypair:
        r"""
        Decode a private key from a protobuf structure and parse it as a `Keypair`.
        """
-    @staticmethod
-    def rsa_from_pkcs8(bytes: bytes) -> Keypair:
-        r"""
-        Decode an keypair from a DER-encoded secret key in PKCS#8 `PrivateKeyInfo`
-        format (i.e. unencrypted) as defined in [RFC5208].
-        
-        [RFC5208]: https://tools.ietf.org/html/rfc5208#section-5
-        """
-    @staticmethod
-    def secp256k1_from_der(bytes: bytes) -> Keypair:
-        r"""
-        Decode a keypair from a DER-encoded Secp256k1 secret key in an `ECPrivateKey`
-        structure as defined in [RFC5915].
-        
-        [RFC5915]: https://tools.ietf.org/html/rfc5915
-        """
-    @staticmethod
-    def ed25519_from_bytes(bytes: bytes) -> Keypair: ...
    def to_protobuf_encoding(self) -> bytes:
        r"""
-        Encode a private key as protobuf structure.
-        """
-    def to_peer_id(self) -> PeerId:
-        r"""
-        Convert the `Keypair` into the corresponding `PeerId`.
+        Encode a private key to a protobuf structure.
        """
+    def to_string(self) -> builtins.str: ...

@typing.final
-class Multiaddr:
-    r"""
-    Representation of a Multiaddr.
-    """
+class PyPeer:
    @staticmethod
-    def empty() -> Multiaddr:
-        r"""
-        Create a new, empty multiaddress.
-        """
-    @staticmethod
-    def with_capacity(n: builtins.int) -> Multiaddr:
-        r"""
-        Create a new, empty multiaddress with the given capacity.
-        """
-    @staticmethod
-    def from_bytes(bytes: bytes) -> Multiaddr:
-        r"""
-        Parse a `Multiaddr` value from its byte slice representation.
-        """
-    @staticmethod
-    def from_string(string: builtins.str) -> Multiaddr:
-        r"""
-        Parse a `Multiaddr` value from its string representation.
-        """
-    def len(self) -> builtins.int:
-        r"""
-        Return the length in bytes of this multiaddress.
-        """
-    def is_empty(self) -> builtins.bool:
-        r"""
-        Returns true if the length of this multiaddress is 0.
-        """
-    def to_bytes(self) -> bytes:
-        r"""
-        Return a copy of this [`Multiaddr`]'s byte representation.
-        """
-    def to_string(self) -> builtins.str:
-        r"""
-        Convert a Multiaddr to a string.
-        """
+    def new(kp: Keypair, namespace: builtins.str) -> PyPeer: ...
+    async def subscribe(self, topic: builtins.str) -> None: ...
+    async def unsubscribe(self, topic: builtins.str) -> None: ...
+    async def send(self, topic: builtins.str, payload: bytes) -> None: ...
+    async def run(self) -> None: ...
+    async def recv(self) -> PySwarmEvent: ...

@typing.final
-class NetworkingHandle:
-    def __new__(cls, identity: Keypair) -> NetworkingHandle: ...
-    async def connection_update_recv(self) -> ConnectionUpdate:
-        r"""
-        Receives the next `ConnectionUpdate` from networking.
-        """
-    async def connection_update_recv_many(self, limit: builtins.int) -> builtins.list[ConnectionUpdate]:
-        r"""
-        Receives at most `limit` `ConnectionUpdate`s from networking and returns them.
-        
-        For `limit = 0`, an empty collection of `ConnectionUpdate`s will be returned immediately.
-        For `limit > 0`, if there are no `ConnectionUpdate`s in the channel's queue this method
-        will sleep until a `ConnectionUpdate`s is sent.
-        """
-    async def gossipsub_subscribe(self, topic: builtins.str) -> builtins.bool:
-        r"""
-        Subscribe to a `GossipSub` topic.
-        
-        Returns `True` if the subscription worked. Returns `False` if we were already subscribed.
-        """
-    async def gossipsub_unsubscribe(self, topic: builtins.str) -> builtins.bool:
-        r"""
-        Unsubscribes from a `GossipSub` topic.
-        
-        Returns `True` if we were subscribed to this topic. Returns `False` if we were not subscribed.
-        """
-    async def gossipsub_publish(self, topic: builtins.str, data: bytes) -> None:
-        r"""
-        Publishes a message with multiple topics to the `GossipSub` network.
-        
-        If no peers are found that subscribe to this topic, throws `NoPeersSubscribedToTopicError` exception.
-        """
-    async def gossipsub_recv(self) -> tuple[builtins.str, bytes]:
-        r"""
-        Receives the next message from the `GossipSub` network.
-        """
-    async def gossipsub_recv_many(self, limit: builtins.int) -> builtins.list[tuple[builtins.str, bytes]]:
-        r"""
-        Receives at most `limit` messages from the `GossipSub` network and returns them.
-        
-        For `limit = 0`, an empty collection of messages will be returned immediately.
-        For `limit > 0`, if there are no messages in the channel's queue this method
-        will sleep until a message is sent.
-        """
-
-@typing.final
-class NoPeersSubscribedToTopicError(builtins.Exception):
-    def __new__(cls, *args: typing.Any) -> NoPeersSubscribedToTopicError: ...
-    def __repr__(self) -> builtins.str: ...
-    def __str__(self) -> builtins.str: ...
-
-@typing.final
-class PeerId:
-    r"""
-    Identifier of a peer of the network.
-    
-    The data is a `CIDv0` compatible multihash of the protobuf encoded public key of the peer
-    as specified in [specs/peer-ids](https://github.com/libp2p/specs/blob/master/peer-ids/peer-ids.md).
-    """
-    @staticmethod
-    def random() -> PeerId:
-        r"""
-        Generates a random peer ID from a cryptographically secure PRNG.
-        
-        This is useful for randomly walking on a DHT, or for testing purposes.
-        """
-    @staticmethod
-    def from_bytes(bytes: bytes) -> PeerId:
-        r"""
-        Parses a `PeerId` from bytes.
-        """
-    def to_bytes(self) -> bytes:
-        r"""
-        Returns a raw bytes representation of this `PeerId`.
-        """
-    def to_base58(self) -> builtins.str:
-        r"""
-        Returns a base-58 encoded string of this `PeerId`.
-        """
-    def __repr__(self) -> builtins.str: ...
-    def __str__(self) -> builtins.str: ...
-
-@typing.final
-class ConnectionUpdateType(enum.Enum):
-    r"""
-    Connection or disconnection event discriminant type.
-    """
-    Connected = ...
-    Disconnected = ...
+class PySwarmEvent:
+    def downcast_discovered(self) -> typing.Optional[builtins.str]: ...
+    def downcast_expired(self) -> typing.Optional[builtins.str]: ...
+    def downcast_message(self) -> typing.Optional[tuple[builtins.str, builtins.str, bytes]]: ...

--- a/rust/exo_pyo3_bindings/src/allow_threading.rs
+++ b/rust/exo_pyo3_bindings/src/allow_threading.rs
@@ -1,8 +1,4 @@
-//! SEE: https://pyo3.rs/v0.26.0/async-await.html#detaching-from-the-interpreter-across-await
-//!
-
-use pin_project::pin_project;
-use pyo3::marker::Ungil;
+//! See: <https://pyo3.rs/v0.27.2/async-await.html#detaching-from-the-interpreter-across-await>
 use pyo3::prelude::*;
 use std::{
    future::Future,
@@ -10,31 +6,17 @@ use std::{
    task::{Context, Poll},
 };

-/// SEE: https://pyo3.rs/v0.26.0/async-await.html#detaching-from-the-interpreter-across-await
-#[pin_project]
-#[repr(transparent)]
-pub(crate) struct AllowThreads<F>(#[pin] F);
-
-impl<F> AllowThreads<F>
-where
-    Self: Future,
-{
-    pub fn new(f: F) -> Self {
-        Self(f)
-    }
-}
+pub struct AllowThreads<F>(pub(crate) F);

 impl<F> Future for AllowThreads<F>
 where
-    F: Future + Ungil,
-    F::Output: Ungil,
+    F: Future + Unpin + Send,
+    F::Output: Send,
 {
    type Output = F::Output;

-    fn poll(self: Pin<&mut Self>, cx: &mut Context<'_>) -> Poll<Self::Output> {
+    fn poll(mut self: Pin<&mut Self>, cx: &mut Context<'_>) -> Poll<Self::Output> {
        let waker = cx.waker();
-        Python::with_gil(|py| {
-            py.allow_threads(|| self.project().0.poll(&mut Context::from_waker(waker)))
-        })
+        Python::attach(|py| py.detach(|| pin!(&mut self.0).poll(&mut Context::from_waker(waker))))
    }
 }
--- a/rust/exo_pyo3_bindings/src/examples/mod.rs
+++ b/rust/exo_pyo3_bindings/src/examples/mod.rs
@@ -1,240 +0,0 @@
-//! This module exists to hold examples of some pyo3 patterns that may be too complex to
-//! re-create from scratch, but too inhomogenous to create an abstraction/wrapper around.
-//!
-//! Pattern examples include:
-//!  - Async task handles: with GC-integrated cleanup
-//!  - Sync/async callbacks from python: with propper eventloop handling
-//!
-//! Mutability pattern: https://pyo3.rs/v0.26.0/async-await.html#send--static-constraint
-//!  - Store mutable fields in tokio's `Mutex<T>`
-//!  - For async code: take `&self` and `.lock().await`
-//!  - For sync code: take `&mut self` and `.get_mut()`
-
-use crate::ext::{PyResultExt as _, ResultExt as _, TokioRuntimeExt as _};
-use futures::FutureExt as _;
-use futures::future::BoxFuture;
-use pyo3::exceptions::PyRuntimeError;
-use pyo3::prelude::{PyModule, PyModuleMethods as _};
-use pyo3::{
-    Bound, Py, PyAny, PyErr, PyResult, PyTraverseError, PyVisit, Python, pyclass, pymethods,
-};
-use std::time::Duration;
-use tokio::sync::mpsc;
-use tokio::sync::mpsc::error::TryRecvError;
-
-fn needs_tokio_runtime() {
-    tokio::runtime::Handle::current();
-}
-
-type SyncCallback = Box<dyn Fn() + Send + Sync>;
-type AsyncCallback = Box<dyn Fn() -> BoxFuture<'static, ()> + Send + Sync>;
-
-enum AsyncTaskMessage {
-    SyncCallback(SyncCallback),
-    AsyncCallback(AsyncCallback),
-}
-
-async fn async_task(
-    sender: mpsc::UnboundedSender<()>,
-    mut receiver: mpsc::UnboundedReceiver<AsyncTaskMessage>,
-) {
-    log::info!("RUST: async task started");
-
-    // task state
-    let mut interval = tokio::time::interval(Duration::from_secs(1));
-
-    let mut sync_cbs: Vec<SyncCallback> = vec![];
-    let mut async_cbs: Vec<AsyncCallback> = vec![];
-
-    loop {
-        tokio::select! {
-            // handle incoming messages from task-handle
-            message = receiver.recv() => {
-                // handle closed channel by exiting
-                let Some(message) = message else {
-                    log::info!("RUST: channel closed");
-                    break;
-                };
-
-                // dispatch incoming event
-                match message {
-                    AsyncTaskMessage::SyncCallback(cb) => {
-                        sync_cbs.push(cb);
-                    }
-                    AsyncTaskMessage::AsyncCallback(cb) => {
-                        async_cbs.push(cb);
-                    }
-                }
-            }
-
-            // handle all other events
-            _ = interval.tick() => {
-                log::info!("RUST: async task tick");
-
-                // call back all sync callbacks
-                for cb in &sync_cbs {
-                    cb();
-                }
-
-                // call back all async callbacks
-                for cb in &async_cbs {
-                    cb().await;
-                }
-
-                // send event on unbounded channel
-                sender.send(()).expect("handle receiver cannot be closed/dropped");
-            }
-        }
-    }
-
-    log::info!("RUST: async task stopped");
-}
-
-// #[gen_stub_pyclass]
-#[pyclass(name = "AsyncTaskHandle")]
-#[derive(Debug)]
-struct PyAsyncTaskHandle {
-    sender: Option<mpsc::UnboundedSender<AsyncTaskMessage>>,
-    receiver: mpsc::UnboundedReceiver<()>,
-}
-
-#[allow(clippy::expect_used)]
-impl PyAsyncTaskHandle {
-    const fn sender(&self) -> &mpsc::UnboundedSender<AsyncTaskMessage> {
-        self.sender
-            .as_ref()
-            .expect("The sender should only be None after de-initialization.")
-    }
-
-    const fn sender_mut(&mut self) -> &mpsc::UnboundedSender<AsyncTaskMessage> {
-        self.sender
-            .as_mut()
-            .expect("The sender should only be None after de-initialization.")
-    }
-
-    const fn new(
-        sender: mpsc::UnboundedSender<AsyncTaskMessage>,
-        receiver: mpsc::UnboundedReceiver<()>,
-    ) -> Self {
-        Self {
-            sender: Some(sender),
-            receiver,
-        }
-    }
-}
-
-// #[gen_stub_pymethods]
-#[pymethods]
-impl PyAsyncTaskHandle {
-    #[new]
-    fn py_new(py: Python<'_>) -> PyResult<Self> {
-        use pyo3_async_runtimes::tokio::get_runtime;
-
-        // create communication channel TOWARDS our task
-        let (h_sender, t_receiver) = mpsc::unbounded_channel::<AsyncTaskMessage>();
-
-        // create communication channel FROM our task
-        let (t_sender, h_receiver) = mpsc::unbounded_channel::<()>();
-
-        // perform necessary setup within tokio context - or it crashes
-        let () = get_runtime().block_on(async { needs_tokio_runtime() });
-
-        // spawn tokio task with this thread's task-locals - without this, async callbacks on the new threads will not work!!
-        _ = get_runtime().spawn_with_scope(py, async move {
-            async_task(t_sender, t_receiver).await;
-        });
-        Ok(Self::new(h_sender, h_receiver))
-    }
-
-    /// NOTE: exceptions in callbacks are silently ignored until end of execution
-    fn add_sync_callback(
-        &self,
-        // #[gen_stub(override_type(
-        //     type_repr="collections.abc.Callable[[], None]",
-        //     imports=("collections.abc")
-        // ))]
-        callback: Py<PyAny>,
-    ) -> PyResult<()> {
-        // blocking call to async method -> can do non-blocking if needed
-        self.sender()
-            .send(AsyncTaskMessage::SyncCallback(Box::new(move || {
-                _ = Python::with_gil(|py| callback.call0(py).write_unraisable_with(py));
-            })))
-            .pyerr()?;
-        Ok(())
-    }
-
-    /// NOTE: exceptions in callbacks are silently ignored until end of execution
-    fn add_async_callback(
-        &self,
-        // #[gen_stub(override_type(
-        //     type_repr="collections.abc.Callable[[], collections.abc.Awaitable[None]]",
-        //     imports=("collections.abc")
-        // ))]
-        callback: Py<PyAny>,
-    ) -> PyResult<()> {
-        // blocking call to async method -> can do non-blocking if needed
-        self.sender()
-            .send(AsyncTaskMessage::AsyncCallback(Box::new(move || {
-                let c = Python::with_gil(|py| callback.clone_ref(py));
-                async move {
-                    if let Some(f) = Python::with_gil(|py| {
-                        let coroutine = c.call0(py).write_unraisable_with(py)?;
-                        pyo3_async_runtimes::tokio::into_future(coroutine.into_bound(py))
-                            .write_unraisable_with(py)
-                    }) {
-                        _ = f.await.write_unraisable();
-                    }
-                }
-                .boxed()
-            })))
-            .pyerr()?;
-        Ok(())
-    }
-
-    async fn receive_unit(&mut self) -> PyResult<()> {
-        self.receiver
-            .recv()
-            .await
-            .ok_or(PyErr::new::<PyRuntimeError, _>(
-                "cannot receive unit on closed channel",
-            ))
-    }
-
-    fn drain_units(&mut self) -> PyResult<i32> {
-        let mut cnt = 0;
-        loop {
-            match self.receiver.try_recv() {
-                Err(TryRecvError::Disconnected) => {
-                    return Err(PyErr::new::<PyRuntimeError, _>(
-                        "cannot receive unit on closed channel",
-                    ));
-                }
-                Err(TryRecvError::Empty) => return Ok(cnt),
-                Ok(()) => {
-                    cnt += 1;
-                    continue;
-                }
-            }
-        }
-    }
-
-    // #[gen_stub(skip)]
-    const fn __traverse__(&self, _visit: PyVisit<'_>) -> Result<(), PyTraverseError> {
-        Ok(()) // This is needed purely so `__clear__` can work
-    }
-
-    // #[gen_stub(skip)]
-    fn __clear__(&mut self) {
-        // TODO: may or may not need to await a "kill-signal" oneshot channel message,
-        //       to ensure that the networking task is done BEFORE exiting the clear function...
-        //       but this may require GIL?? and it may not be safe to call GIL here??
-        self.sender = None; // Using Option<T> as a trick to force `sender` channel to be dropped
-    }
-}
-
-pub fn examples_submodule(m: &Bound<'_, PyModule>) -> PyResult<()> {
-    m.add_class::<PyAsyncTaskHandle>()?;
-
-    Ok(())
-}
--- a/rust/exo_pyo3_bindings/src/lib.rs
+++ b/rust/exo_pyo3_bindings/src/lib.rs
@@ -1,216 +1,42 @@
 //! TODO: crate documentation
-//!
-//! this is here as a placeholder documentation
-//!
-//!
+pub(crate) mod allow_threading;

-// enable Rust-unstable features for convenience
-#![feature(trait_alias)]
-#![feature(tuple_trait)]
-#![feature(unboxed_closures)]
-// #![feature(stmt_expr_attributes)]
-// #![feature(assert_matches)]
-// #![feature(async_fn_in_dyn_trait)]
-// #![feature(async_for_loop)]
-// #![feature(auto_traits)]
-// #![feature(negative_impls)]
-
-extern crate core;
-mod allow_threading;
-mod examples;
 pub(crate) mod networking;
-pub(crate) mod pylibp2p;
+pub(crate) mod take_once {
+    use tokio::sync::Mutex;
+
+    pub struct TakeOnce<T>(Mutex<Option<T>>);
+    impl<T> TakeOnce<T> {
+        pub fn new(t: T) -> Self {
+            Self(Mutex::new(Some(t)))
+        }
+        pub fn take(&self) -> Option<T> {
+            match self.0.try_lock() {
+                Ok(mut o) => o.take(),
+                Err(_) => None,
+            }
+        }
+    }
+}

-use crate::networking::networking_submodule;
-use crate::pylibp2p::ident::ident_submodule;
-use crate::pylibp2p::multiaddr::multiaddr_submodule;
-use pyo3::prelude::PyModule;
 use pyo3::prelude::*;
-use pyo3::{Bound, PyResult, pyclass, pymodule};
+
 use pyo3_stub_gen::define_stub_info_gatherer;

-/// Namespace for all the constants used by this crate.
-pub(crate) mod r#const {
-    pub const MPSC_CHANNEL_SIZE: usize = 1024;
-}
-
-/// Namespace for all the type/trait aliases used by this crate.
-pub(crate) mod alias {
-    use std::error::Error;
-    use std::marker::Tuple;
-
-    pub trait SendFn<Args: Tuple + Send + 'static, Output> =
-        Fn<Args, Output = Output> + Send + 'static;
-
-    pub type AnyError = Box<dyn Error + Send + Sync + 'static>;
-    pub type AnyResult<T> = Result<T, AnyError>;
-}
-
-/// Namespace for crate-wide extension traits/methods
-pub(crate) mod ext {
-    use crate::allow_threading::AllowThreads;
-    use extend::ext;
-    use pyo3::exceptions::{PyConnectionError, PyRuntimeError};
-    use pyo3::marker::Ungil;
-    use pyo3::types::PyBytes;
-    use pyo3::{Py, PyErr, PyResult, Python};
-    use tokio::runtime::Runtime;
-    use tokio::sync::mpsc;
-    use tokio::sync::mpsc::error::TryRecvError;
-    use tokio::task::JoinHandle;
-
-    #[ext(pub, name = ByteArrayExt)]
-    impl [u8] {
-        fn pybytes(&self) -> Py<PyBytes> {
-            Python::with_gil(|py| PyBytes::new(py, self).unbind())
-        }
-    }
-
-    #[ext(pub, name = ResultExt)]
-    impl<T, E> Result<T, E>
-    where
-        E: ToString,
-    {
-        fn pyerr(self) -> PyResult<T> {
-            self.map_err(|e| PyRuntimeError::new_err(e.to_string()))
-        }
-    }
-
-    pub trait FutureExt: Future + Sized {
-        /// SEE: https://pyo3.rs/v0.26.0/async-await.html#detaching-from-the-interpreter-across-await
-        fn allow_threads_py(self) -> AllowThreads<Self>
-        where
-            AllowThreads<Self>: Future,
-        {
-            AllowThreads::new(self)
-        }
-    }
-
-    impl<T: Future> FutureExt for T {}
-
-    #[ext(pub, name = PyErrExt)]
-    impl PyErr {
-        fn receiver_channel_closed() -> Self {
-            PyConnectionError::new_err("Receiver channel closed unexpectedly")
-        }
-    }
-
-    #[ext(pub, name = PyResultExt)]
-    impl<T> PyResult<T> {
-        fn write_unraisable(self) -> Option<T> {
-            Python::with_gil(|py| self.write_unraisable_with(py))
-        }
-
-        fn write_unraisable_with(self, py: Python<'_>) -> Option<T> {
-            match self {
-                Ok(v) => Some(v),
-                Err(e) => {
-                    // write error back to python
-                    e.write_unraisable(py, None);
-                    None
-                }
-            }
-        }
-    }
-
-    #[ext(pub, name = TokioRuntimeExt)]
-    impl Runtime {
-        fn spawn_with_scope<F>(&self, py: Python<'_>, future: F) -> PyResult<JoinHandle<F::Output>>
-        where
-            F: Future + Send + 'static,
-            F::Output: Send + 'static,
-        {
-            let locals = pyo3_async_runtimes::tokio::get_current_locals(py)?;
-            Ok(self.spawn(pyo3_async_runtimes::tokio::scope(locals, future)))
-        }
-    }
-
-    #[ext(pub, name = TokioMpscSenderExt)]
-    impl<T> mpsc::Sender<T> {
-        /// Sends a value, waiting until there is capacity.
-        ///
-        /// A successful send occurs when it is determined that the other end of the
-        /// channel has not hung up already. An unsuccessful send would be one where
-        /// the corresponding receiver has already been closed.
-        async fn send_py(&self, value: T) -> PyResult<()> {
-            self.send(value)
-                .await
-                .map_err(|_| PyErr::receiver_channel_closed())
-        }
-    }
-
-    #[ext(pub, name = TokioMpscReceiverExt)]
-    impl<T> mpsc::Receiver<T> {
-        /// Receives the next value for this receiver.
-        async fn recv_py(&mut self) -> PyResult<T> {
-            self.recv().await.ok_or_else(PyErr::receiver_channel_closed)
-        }
-
-        /// Receives at most `limit` values for this receiver and returns them.
-        ///
-        /// For `limit = 0`, an empty collection of messages will be returned immediately.
-        /// For `limit > 0`, if there are no messages in the channel's queue this method
-        /// will sleep until a message is sent.
-        async fn recv_many_py(&mut self, limit: usize) -> PyResult<Vec<T>> {
-            // get updates from receiver channel
-            let mut updates = Vec::with_capacity(limit);
-            let received = self.recv_many(&mut updates, limit).await;
-
-            // if we received zero items, then the channel was unexpectedly closed
-            if limit != 0 && received == 0 {
-                return Err(PyErr::receiver_channel_closed());
-            }
-
-            Ok(updates)
-        }
-
-        /// Tries to receive the next value for this receiver.
-        fn try_recv_py(&mut self) -> PyResult<Option<T>> {
-            match self.try_recv() {
-                Ok(v) => Ok(Some(v)),
-                Err(TryRecvError::Empty) => Ok(None),
-                Err(TryRecvError::Disconnected) => Err(PyErr::receiver_channel_closed()),
-            }
-        }
-    }
-}
-
-pub(crate) mod private {
-    use std::marker::Sized;
-
-    /// Sealed traits support
-    pub trait Sealed {}
-    impl<T: ?Sized> Sealed for T {}
-}
-
-/// A wrapper around [`Py`] that implements [`Clone`] using [`Python::with_gil`].
-#[repr(transparent)]
-pub(crate) struct ClonePy<T>(pub Py<T>);
-
-impl<T> Clone for ClonePy<T> {
-    fn clone(&self) -> Self {
-        Python::with_gil(|py| Self(self.0.clone_ref(py)))
-    }
-}
-
 /// A Python module implemented in Rust. The name of this function must match
 /// the `lib.name` setting in the `Cargo.toml`, else Python will not be able to
 /// import the module.
 #[pymodule(name = "exo_pyo3_bindings")]
-fn main_module(m: &Bound<'_, PyModule>) -> PyResult<()> {
+pub fn networking_module(m: &Bound<'_, PyModule>) -> PyResult<()> {
    // install logger
    pyo3_log::init();
+    // setup runtime
+    let mut builder = tokio::runtime::Builder::new_multi_thread();
+    builder.enable_all();
+    pyo3_async_runtimes::tokio::init(builder);

-    // TODO: for now this is all NOT a submodule, but figure out how to make the submodule system
-    //       work with maturin, where the types generate correctly, in the right folder, without
-    //       too many importing issues...
-    ident_submodule(m)?;
-    multiaddr_submodule(m)?;
-    networking_submodule(m)?;
-
-    // top-level constructs
-    // TODO: ...
-
+    m.add_class::<networking::PyPeer>()?;
+    m.add_class::<networking::PyKeypair>()?;
    Ok(())
 }

--- a/rust/exo_pyo3_bindings/src/networking.rs
+++ b/rust/exo_pyo3_bindings/src/networking.rs
@@ -1,571 +1,214 @@
-#![allow(
-    clippy::multiple_inherent_impl,
-    clippy::unnecessary_wraps,
-    clippy::unused_self,
-    clippy::needless_pass_by_value
-)]
+use crate::allow_threading::AllowThreads;
+use crate::take_once::TakeOnce;

-use crate::r#const::MPSC_CHANNEL_SIZE;
-use crate::ext::{ByteArrayExt as _, FutureExt, PyErrExt as _};
-use crate::ext::{ResultExt as _, TokioMpscReceiverExt as _, TokioMpscSenderExt as _};
-use crate::pyclass;
-use crate::pylibp2p::ident::{PyKeypair, PyPeerId};
-use libp2p::futures::StreamExt as _;
-use libp2p::gossipsub::{IdentTopic, Message, MessageId, PublishError};
-use libp2p::swarm::SwarmEvent;
-use libp2p::{gossipsub, mdns};
-use networking::discovery;
-use networking::swarm::create_swarm;
-use pyo3::prelude::{PyModule, PyModuleMethods as _};
-use pyo3::types::PyBytes;
-use pyo3::{Bound, Py, PyErr, PyResult, PyTraverseError, PyVisit, Python, pymethods};
-use pyo3_stub_gen::derive::{gen_stub_pyclass, gen_stub_pyclass_enum, gen_stub_pymethods};
-use std::net::IpAddr;
-use tokio::sync::{Mutex, mpsc, oneshot};
-use util::ext::VecExt as _;
+use std::pin::pin;

-mod exception {
-    use pyo3::types::PyTuple;
-    use pyo3::{PyErrArguments, exceptions::PyException, prelude::*};
-    use pyo3_stub_gen::derive::*;
-
-    #[gen_stub_pyclass]
-    #[pyclass(frozen, extends=PyException, name="NoPeersSubscribedToTopicError")]
-    pub struct PyNoPeersSubscribedToTopicError {}
-
-    impl PyNoPeersSubscribedToTopicError {
-        const MSG: &'static str = "\
-        No peers are currently subscribed to receive messages on this topic. \
-        Wait for peers to subscribe or check your network connectivity.";
-
-        ///   Creates a new  [ `PyErr` ]  of this type.
-        ///
-        ///   [`PyErr`] :  https://docs.rs/pyo3/latest/pyo3/struct.PyErr.html   "PyErr in pyo3"
-        pub(crate) fn new_err() -> PyErr {
-            PyErr::new::<Self, _>(()) // TODO: check if this needs to be replaced???
-        }
-    }
-
-    #[gen_stub_pymethods]
-    #[pymethods]
-    impl PyNoPeersSubscribedToTopicError {
-        #[new]
-        #[pyo3(signature = (*args))]
-        #[allow(unused_variables)]
-        pub(crate) fn new(args: &Bound<'_, PyTuple>) -> Self {
-            Self {}
-        }
-
-        fn __repr__(&self) -> String {
-            format!("PeerId(\"{}\")", Self::MSG)
-        }
-
-        fn __str__(&self) -> String {
-            Self::MSG.to_string()
-        }
-    }
-
-    #[gen_stub_pyclass]
-    #[pyclass(frozen, extends=PyException, name="AllQueuesFullError")]
-    pub struct PyAllQueuesFullError {}
-
-    impl PyAllQueuesFullError {
-        const MSG: &'static str =
-            "All libp2p peers are unresponsive, resend the message or reconnect.";
-
-        ///   Creates a new  [ `PyErr` ]  of this type.
-        ///
-        ///   [`PyErr`] :  https://docs.rs/pyo3/latest/pyo3/struct.PyErr.html   "PyErr in pyo3"
-        pub(crate) fn new_err() -> PyErr {
-            PyErr::new::<Self, _>(()) // TODO: check if this needs to be replaced???
-        }
-    }
-
-    #[gen_stub_pymethods]
-    #[pymethods]
-    impl PyAllQueuesFullError {
-        #[new]
-        #[pyo3(signature = (*args))]
-        #[allow(unused_variables)]
-        pub(crate) fn new(args: &Bound<'_, PyTuple>) -> Self {
-            Self {}
-        }
-
-        fn __repr__(&self) -> String {
-            format!("PeerId(\"{}\")", Self::MSG)
-        }
-
-        fn __str__(&self) -> String {
-            Self::MSG.to_string()
-        }
-    }
-}
-
-/// Connection or disconnection event discriminant type.
-#[gen_stub_pyclass_enum]
-#[pyclass(eq, eq_int, name = "ConnectionUpdateType")]
-#[derive(Debug, Clone, PartialEq)]
-enum PyConnectionUpdateType {
-    Connected = 0,
-    Disconnected,
-}
+use futures_lite::FutureExt;
+use libp2p::{gossipsub::PublishError, identity::Keypair};
+use networking::{FromSwarm, Peer, ToSwarm};
+use pyo3::{
+    coroutine::CancelHandle,
+    exceptions::{PyConnectionError, PyRuntimeError, PyValueError},
+    prelude::*,
+    types::PyBytes,
+};
+use pyo3_stub_gen::{
+    derive::{gen_methods_from_python, gen_stub_pyclass, gen_stub_pymethods},
+    inventory::submit,
+};
+use tokio::sync::{Mutex, mpsc};

 #[gen_stub_pyclass]
-#[pyclass(frozen, name = "ConnectionUpdate")]
-#[derive(Debug, Clone)]
-struct PyConnectionUpdate {
-    /// Whether this is a connection or disconnection event
-    #[pyo3(get)]
-    update_type: PyConnectionUpdateType,
+#[pyclass(name = "Keypair", frozen)]
+#[derive(Clone)]
+pub struct PyKeypair(Keypair);

-    /// Identity of the peer that we have connected to or disconnected from.
-    #[pyo3(get)]
-    peer_id: PyPeerId,
+#[gen_stub_pymethods]
+#[pymethods]
+impl PyKeypair {
+    /// Generate a new ed25519 keypair
+    #[staticmethod]
+    fn generate() -> Self {
+        Self(Keypair::generate_ed25519())
+    }

-    /// Remote connection's IPv4 address.
-    #[pyo3(get)]
-    remote_ipv4: String,
+    /// Decode a private key from a protobuf structure and parse it as a `Keypair`.
+    #[staticmethod]
+    fn from_protobuf_encoding(bytes: &Bound<'_, PyBytes>) -> Self {
+        let bytes = Vec::from(bytes.as_bytes());
+        Self(Keypair::from_protobuf_encoding(&bytes).expect("todo"))
+    }

-    /// Remote connection's TCP port.
-    #[pyo3(get)]
-    remote_tcp_port: u16,
-}
-
-enum ToTask {
-    GossipsubSubscribe {
-        topic: String,
-        result_tx: oneshot::Sender<PyResult<bool>>,
-    },
-    GossipsubUnsubscribe {
-        topic: String,
-        result_tx: oneshot::Sender<bool>,
-    },
-    GossipsubPublish {
-        topic: String,
-        data: Vec<u8>,
-        result_tx: oneshot::Sender<PyResult<MessageId>>,
-    },
-}
-
-#[allow(clippy::enum_glob_use)]
-async fn networking_task(
-    mut swarm: networking::swarm::Swarm,
-    mut to_task_rx: mpsc::Receiver<ToTask>,
-    connection_update_tx: mpsc::Sender<PyConnectionUpdate>,
-    gossipsub_message_tx: mpsc::Sender<(String, Vec<u8>)>,
-) {
-    use SwarmEvent::*;
-    use ToTask::*;
-    use mdns::Event::*;
-    use networking::swarm::BehaviourEvent::*;
-
-    log::info!("RUST: networking task started");
-
-    loop {
-        tokio::select! {
-            message = to_task_rx.recv() => {
-                // handle closed channel
-                let Some(message) = message else {
-                    log::info!("RUST: channel closed");
-                    break;
-                };
-
-                // dispatch incoming messages
-                match message {
-                    GossipsubSubscribe { topic, result_tx } => {
-                        // try to subscribe
-                        let result = swarm.behaviour_mut()
-                            .gossipsub.subscribe(&IdentTopic::new(topic));
-
-                        // send response oneshot
-                        if let Err(e) = result_tx.send(result.pyerr()) {
-                            log::error!("RUST: could not subscribe to gossipsub topic since channel already closed: {e:?}");
-                            continue;
-                        }
-                    }
-                    GossipsubUnsubscribe { topic, result_tx } => {
-                        // try to unsubscribe from the topic
-                        let result = swarm.behaviour_mut()
-                            .gossipsub.unsubscribe(&IdentTopic::new(topic));
-
-                        // send response oneshot (or exit if connection closed)
-                        if let Err(e) = result_tx.send(result) {
-                            log::error!("RUST: could not unsubscribe from gossipsub topic since channel already closed: {e:?}");
-                            continue;
-                        }
-                    }
-                    GossipsubPublish { topic, data, result_tx } => {
-                        // try to publish the data -> catch NoPeersSubscribedToTopic error & convert to correct exception
-                        let result = swarm.behaviour_mut().gossipsub.publish(
-                            IdentTopic::new(topic), data);
-                        let pyresult: PyResult<MessageId> = if let Err(PublishError::NoPeersSubscribedToTopic) = result {
-                            Err(exception::PyNoPeersSubscribedToTopicError::new_err())
-                        } else if let Err(PublishError::AllQueuesFull(_)) = result {
-                            Err(exception::PyAllQueuesFullError::new_err())
-                        } else {
-                            result.pyerr()
-                        };
-
-                        // send response oneshot (or exit if connection closed)
-                        if let Err(e) = result_tx.send(pyresult) {
-                            log::error!("RUST: could not publish gossipsub message since channel already closed: {e:?}");
-                            continue;
-                        }
-                    }
-                }
-            }
-
-            // architectural solution to this problem:
-            // create keep_alive behavior who's job it is to dial peers discovered by mDNS (and drop when expired)
-            //   -> it will emmit TRUE connected/disconnected events consumable elsewhere
-            //
-            // gossipsub will feed off-of dial attempts created by networking, and that will bootstrap its' peers list
-            // then for actual communication it will dial those peers if need-be
-            swarm_event = swarm.select_next_some() => {
-                match swarm_event {
-                    Behaviour(Gossipsub(gossipsub::Event::Message {
-                        message: Message {
-                            topic,
-                            data,
-                            ..
-                        },
-                        ..
-                    })) => {
-                        // topic-ID is just the topic hash!!! (since we used identity hasher)
-                        let message = (topic.into_string(), data);
-
-                        // send incoming message to channel (or exit if connection closed)
-                        if let Err(e) = gossipsub_message_tx.send(message).await {
-                            log::error!("RUST: could not send incoming gossipsub message since channel already closed: {e}");
-                            continue;
-                        }
-                    },
-                    Behaviour(Discovery(discovery::Event::ConnectionEstablished { peer_id, remote_ip, remote_tcp_port, .. })) => {
-                        // grab IPv4 string
-                        let remote_ipv4 = match remote_ip {
-                            IpAddr::V4(ip) => ip.to_string(),
-                            IpAddr::V6(ip) => {
-                                log::warn!("RUST: ignoring connection to IPv6 address: {ip}");
-                                continue;
-                            }
-                        };
-
-                        // send connection event to channel (or exit if connection closed)
-                        if let Err(e) = connection_update_tx.send(PyConnectionUpdate {
-                            update_type: PyConnectionUpdateType::Connected,
-                            peer_id: PyPeerId(peer_id),
-                            remote_ipv4,
-                            remote_tcp_port,
-                        }).await {
-                            log::error!("RUST: could not send connection update since channel already closed: {e}");
-                            continue;
-                        }
-                    },
-                    Behaviour(Discovery(discovery::Event::ConnectionClosed { peer_id, remote_ip, remote_tcp_port, .. })) => {
-                        // grab IPv4 string
-                        let remote_ipv4 = match remote_ip {
-                            IpAddr::V4(ip) => ip.to_string(),
-                            IpAddr::V6(ip) => {
-                                log::warn!("RUST: ignoring disconnection from IPv6 address: {ip}");
-                                continue;
-                            }
-                        };
-
-                        // send disconnection event to channel (or exit if connection closed)
-                        if let Err(e) = connection_update_tx.send(PyConnectionUpdate {
-                            update_type: PyConnectionUpdateType::Disconnected,
-                            peer_id: PyPeerId(peer_id),
-                            remote_ipv4,
-                            remote_tcp_port,
-                        }).await {
-                            log::error!("RUST: could not send connection update since channel already closed: {e}");
-                            continue;
-                        }
-                    },
-                    e => {
-                        log::info!("RUST: other event {e:?}");
-                    }
-                }
-            }
+    /// Encode a private key to a protobuf structure.
+    fn to_protobuf_encoding<'py>(&self, py: Python<'py>) -> PyResult<Bound<'py, PyBytes>> {
+        match self.0.to_protobuf_encoding() {
+            Ok(bytes) => Ok(PyBytes::new(py, &bytes)),
+            Err(e) => Err(PyValueError::new_err(e.to_string())),
        }
    }

-    log::info!("RUST: networking task stopped");
+    fn to_string(&self) -> String {
+        self.0.public().to_peer_id().to_base58()
+    }
 }

+struct PeerBuilder(
+    String,
+    Keypair,
+    mpsc::Sender<FromSwarm>,
+    mpsc::Receiver<ToSwarm>,
+);
+
 #[gen_stub_pyclass]
-#[pyclass(name = "NetworkingHandle")]
-#[derive(Debug)]
-struct PyNetworkingHandle {
-    // channels
-    to_task_tx: Option<mpsc::Sender<ToTask>>,
-    connection_update_rx: Mutex<mpsc::Receiver<PyConnectionUpdate>>,
-    gossipsub_message_rx: Mutex<mpsc::Receiver<(String, Vec<u8>)>>,
-}
-
-impl Drop for PyNetworkingHandle {
-    fn drop(&mut self) {
-        // TODO: may or may not need to await a "kill-signal" oneshot channel message,
-        //       to ensure that the networking task is done BEFORE exiting the clear function...
-        //       but this may require GIL?? and it may not be safe to call GIL here??
-        self.to_task_tx = None; // Using Option<T> as a trick to force channel to be dropped
-    }
-}
-
-#[allow(clippy::expect_used)]
-impl PyNetworkingHandle {
-    fn new(
-        to_task_tx: mpsc::Sender<ToTask>,
-        connection_update_rx: mpsc::Receiver<PyConnectionUpdate>,
-        gossipsub_message_rx: mpsc::Receiver<(String, Vec<u8>)>,
-    ) -> Self {
-        Self {
-            to_task_tx: Some(to_task_tx),
-            connection_update_rx: Mutex::new(connection_update_rx),
-            gossipsub_message_rx: Mutex::new(gossipsub_message_rx),
-        }
-    }
-
-    const fn to_task_tx(&self) -> &mpsc::Sender<ToTask> {
-        self.to_task_tx
-            .as_ref()
-            .expect("The sender should only be None after de-initialization.")
-    }
+#[pyclass]
+pub struct PyPeer {
+    peer: TakeOnce<PeerBuilder>,
+    to_swarm: mpsc::Sender<ToSwarm>,
+    from_swarm: Mutex<mpsc::Receiver<FromSwarm>>,
 }

 #[gen_stub_pymethods]
 #[pymethods]
-impl PyNetworkingHandle {
-    // NOTE: `async fn`s here that use `.await` will wrap the future in `.allow_threads_py()`
-    //       immediately beforehand to release the interpreter.
-    //       SEE: https://pyo3.rs/v0.26.0/async-await.html#detaching-from-the-interpreter-across-await
-
-    // ---- Lifecycle management methods ----
-
-    #[new]
-    fn py_new(identity: Bound<'_, PyKeypair>) -> PyResult<Self> {
-        use pyo3_async_runtimes::tokio::get_runtime;
-
-        // create communication channels
-        let (to_task_tx, to_task_rx) = mpsc::channel(MPSC_CHANNEL_SIZE);
-        let (connection_update_tx, connection_update_rx) = mpsc::channel(MPSC_CHANNEL_SIZE);
-        let (gossipsub_message_tx, gossipsub_message_rx) = mpsc::channel(MPSC_CHANNEL_SIZE);
-
-        // get identity
-        let identity = identity.borrow().0.clone();
-
-        // create networking swarm (within tokio context!! or it crashes)
-        let swarm = get_runtime()
-            .block_on(async { create_swarm(identity) })
-            .pyerr()?;
-
-        // spawn tokio task running the networking logic
-        get_runtime().spawn(async move {
-            networking_task(
-                swarm,
-                to_task_rx,
-                connection_update_tx,
-                gossipsub_message_tx,
-            )
-            .await;
-        });
-        Ok(Self::new(
-            to_task_tx,
-            connection_update_rx,
-            gossipsub_message_rx,
-        ))
+impl PyPeer {
+    #[staticmethod]
+    fn new(kp: PyKeypair, namespace: String) -> PyResult<Self> {
+        let (to_client, from_swarm) = mpsc::channel(1024);
+        let (to_swarm, from_client) = mpsc::channel(1024);
+        Ok(Self {
+            peer: TakeOnce::new(PeerBuilder(namespace, kp.0, to_client, from_client)),
+            to_swarm,
+            from_swarm: Mutex::new(from_swarm),
+        })
    }

    #[gen_stub(skip)]
-    const fn __traverse__(&self, _visit: PyVisit<'_>) -> Result<(), PyTraverseError> {
-        Ok(()) // This is needed purely so `__clear__` can work
+    async fn run(&self, #[pyo3(cancel_handle)] mut cancel: CancelHandle) -> PyResult<()> {
+        let builder = self
+            .peer
+            .take()
+            .ok_or_else(|| PyRuntimeError::new_err("tried to run peer twice"))?;
+        let jh = pyo3_async_runtimes::tokio::get_runtime()
+            .spawn(async move {
+                let mut peer =
+                    Peer::new(builder.0, builder.1, builder.2, builder.3).map_err(|_| {
+                        PyConnectionError::new_err("peer failed to listen on default address")
+                    })?;
+                peer.run()
+                    .await
+                    .map_err(|()| PyConnectionError::new_err("peer communication closed"))
+            })
+            .or(async {
+                cancel.cancelled().await;
+                Ok(Ok(()))
+            });
+        match AllowThreads(pin!(jh)).await {
+            Err(e) if e.is_cancelled() => Ok(()),
+            Err(e) if e.is_panic() => Err(PyRuntimeError::new_err(format!("tokio panic {e}"))),
+            Err(_) => unreachable!(),
+            Ok(res) => res,
+        }
+    }
+
+    async fn subscribe(&self, topic: String) -> PyResult<()> {
+        self.to_swarm
+            .send(ToSwarm::Subscribe(topic))
+            .await
+            .map_err(|_| PyRuntimeError::new_err("swarm communication closed"))
+    }
+    async fn unsubscribe(&self, topic: String) -> PyResult<()> {
+        self.to_swarm
+            .send(ToSwarm::Unsubscribe(topic))
+            .await
+            .map_err(|_| PyRuntimeError::new_err("swarm communication closed"))
+    }
+    async fn send(&self, topic: String, payload: Py<PyBytes>) -> PyResult<()> {
+        // this function attaches to the python interpreter synchronously to avoid holding the GIL
+        let bytes = Python::attach(|py| Vec::from(payload.bind(py).as_bytes()));
+        self.to_swarm
+            .send(ToSwarm::Message(topic, bytes))
+            .await
+            .map_err(|_| PyRuntimeError::new_err("swarm communication closed"))
    }

    #[gen_stub(skip)]
-    fn __clear__(&mut self) {
-        // TODO: may or may not need to await a "kill-signal" oneshot channel message,
-        //       to ensure that the networking task is done BEFORE exiting the clear function...
-        //       but this may require GIL?? and it may not be safe to call GIL here??
-        self.to_task_tx = None; // Using Option<T> as a trick to force channel to be dropped
+    async fn recv(
+        &self,
+        #[pyo3(cancel_handle)] mut cancel: CancelHandle,
+    ) -> PyResult<PySwarmEvent> {
+        loop {
+            return match AllowThreads(pin!(
+                self.from_swarm
+                    .try_lock()
+                    .map_err(|_| PyRuntimeError::new_err("tried to recv twice"))?
+                    .recv()
+                    .or(async {
+                        cancel.cancelled().await;
+                        None
+                    })
+            ))
+            .await
+            {
+                Some(FromSwarm::PublishError(p)) => match p {
+                    PublishError::AllQueuesFull(_) => {
+                        Err(PyConnectionError::new_err("swarm overloaded"))
+                    }
+                    PublishError::MessageTooLarge => {
+                        Err(PyValueError::new_err("message too large"))
+                    }
+                    PublishError::NoPeersSubscribedToTopic => {
+                        continue;
+                    }
+                    // TODO(evan): logs here
+                    _ => continue,
+                },
+                None => Err(PyRuntimeError::new_err("swarm communication closed")),
+                Some(fs) => Ok(PySwarmEvent(fs)),
+            };
+        }
    }
-
-    // ---- Connection update receiver methods ----
-
-    /// Receives the next `ConnectionUpdate` from networking.
-    async fn connection_update_recv(&self) -> PyResult<PyConnectionUpdate> {
-        self.connection_update_rx
-            .lock()
-            .allow_threads_py() // allow-threads-aware async call
-            .await
-            .recv_py()
-            .allow_threads_py() // allow-threads-aware async call
-            .await
-    }
-
-    /// Receives at most `limit` `ConnectionUpdate`s from networking and returns them.
-    ///
-    /// For `limit = 0`, an empty collection of `ConnectionUpdate`s will be returned immediately.
-    /// For `limit > 0`, if there are no `ConnectionUpdate`s in the channel's queue this method
-    /// will sleep until a `ConnectionUpdate`s is sent.
-    async fn connection_update_recv_many(&self, limit: usize) -> PyResult<Vec<PyConnectionUpdate>> {
-        self.connection_update_rx
-            .lock()
-            .allow_threads_py() // allow-threads-aware async call
-            .await
-            .recv_many_py(limit)
-            .allow_threads_py() // allow-threads-aware async call
-            .await
-    }
-
-    // TODO: rn this blocks main thread if anything else is awaiting the channel (bc its a mutex)
-    //       so its too dangerous to expose just yet. figure out a better semantics for handling this,
-    //       so things don't randomly block
-    // /// Tries to receive the next `ConnectionUpdate` from networking.
-    // fn connection_update_try_recv(&self) -> PyResult<Option<PyConnectionUpdate>> {
-    //     self.connection_update_rx.blocking_lock().try_recv_py()
-    // }
-    //
-    // /// Checks if the `ConnectionUpdate` channel is empty.
-    // fn connection_update_is_empty(&self) -> bool {
-    //     self.connection_update_rx.blocking_lock().is_empty()
-    // }
-    //
-    // /// Returns the number of `ConnectionUpdate`s in the channel.
-    // fn connection_update_len(&self) -> usize {
-    //     self.connection_update_rx.blocking_lock().len()
-    // }
-
-    // ---- Gossipsub management methods ----
-
-    /// Subscribe to a `GossipSub` topic.
-    ///
-    /// Returns `True` if the subscription worked. Returns `False` if we were already subscribed.
-    async fn gossipsub_subscribe(&self, topic: String) -> PyResult<bool> {
-        let (tx, rx) = oneshot::channel();
-
-        // send off request to subscribe
-        self.to_task_tx()
-            .send_py(ToTask::GossipsubSubscribe {
-                topic,
-                result_tx: tx,
-            })
-            .allow_threads_py() // allow-threads-aware async call
-            .await?;
-
-        // wait for response & return any errors
-        rx.allow_threads_py() // allow-threads-aware async call
-            .await
-            .map_err(|_| PyErr::receiver_channel_closed())?
-    }
-
-    /// Unsubscribes from a `GossipSub` topic.
-    ///
-    /// Returns `True` if we were subscribed to this topic. Returns `False` if we were not subscribed.
-    async fn gossipsub_unsubscribe(&self, topic: String) -> PyResult<bool> {
-        let (tx, rx) = oneshot::channel();
-
-        // send off request to unsubscribe
-        self.to_task_tx()
-            .send_py(ToTask::GossipsubUnsubscribe {
-                topic,
-                result_tx: tx,
-            })
-            .allow_threads_py() // allow-threads-aware async call
-            .await?;
-
-        // wait for response & convert any errors
-        rx.allow_threads_py() // allow-threads-aware async call
-            .await
-            .map_err(|_| PyErr::receiver_channel_closed())
-    }
-
-    /// Publishes a message with multiple topics to the `GossipSub` network.
-    ///
-    /// If no peers are found that subscribe to this topic, throws `NoPeersSubscribedToTopicError` exception.
-    async fn gossipsub_publish(&self, topic: String, data: Py<PyBytes>) -> PyResult<()> {
-        let (tx, rx) = oneshot::channel();
-
-        // send off request to subscribe
-        let data = Python::with_gil(|py| Vec::from(data.as_bytes(py)));
-        self.to_task_tx()
-            .send_py(ToTask::GossipsubPublish {
-                topic,
-                data,
-                result_tx: tx,
-            })
-            .allow_threads_py() // allow-threads-aware async call
-            .await?;
-
-        // wait for response & return any errors => ignore messageID for now!!!
-        let _ = rx
-            .allow_threads_py() // allow-threads-aware async call
-            .await
-            .map_err(|_| PyErr::receiver_channel_closed())??;
-        Ok(())
-    }
-
-    // ---- Gossipsub message receiver methods ----
-
-    /// Receives the next message from the `GossipSub` network.
-    async fn gossipsub_recv(&self) -> PyResult<(String, Py<PyBytes>)> {
-        self.gossipsub_message_rx
-            .lock()
-            .allow_threads_py() // allow-threads-aware async call
-            .await
-            .recv_py()
-            .allow_threads_py() // allow-threads-aware async call
-            .await
-            .map(|(t, d)| (t, d.pybytes()))
-    }
-
-    /// Receives at most `limit` messages from the `GossipSub` network and returns them.
-    ///
-    /// For `limit = 0`, an empty collection of messages will be returned immediately.
-    /// For `limit > 0`, if there are no messages in the channel's queue this method
-    /// will sleep until a message is sent.
-    async fn gossipsub_recv_many(&self, limit: usize) -> PyResult<Vec<(String, Py<PyBytes>)>> {
-        Ok(self
-            .gossipsub_message_rx
-            .lock()
-            .allow_threads_py() // allow-threads-aware async call
-            .await
-            .recv_many_py(limit)
-            .allow_threads_py() // allow-threads-aware async call
-            .await?
-            .map(|(t, d)| (t, d.pybytes())))
-    }
-
-    // TODO: rn this blocks main thread if anything else is awaiting the channel (bc its a mutex)
-    //       so its too dangerous to expose just yet. figure out a better semantics for handling this,
-    //       so things don't randomly block
-    // /// Tries to receive the next message from the `GossipSub` network.
-    // fn gossipsub_try_recv(&self) -> PyResult<Option<(String, Py<PyBytes>)>> {
-    //     Ok(self
-    //         .gossipsub_message_rx
-    //         .blocking_lock()
-    //         .try_recv_py()?
-    //         .map(|(t, d)| (t, d.pybytes())))
-    // }
-    //
-    // /// Checks if the `GossipSub` message channel is empty.
-    // fn gossipsub_is_empty(&self) -> bool {
-    //     self.gossipsub_message_rx.blocking_lock().is_empty()
-    // }
-    //
-    // /// Returns the number of `GossipSub` messages in the channel.
-    // fn gossipsub_len(&self) -> usize {
-    //     self.gossipsub_message_rx.blocking_lock().len()
-    // }
 }

-pub fn networking_submodule(m: &Bound<'_, PyModule>) -> PyResult<()> {
-    m.add_class::<exception::PyNoPeersSubscribedToTopicError>()?;
-    m.add_class::<exception::PyAllQueuesFullError>()?;
-
-    m.add_class::<PyConnectionUpdateType>()?;
-    m.add_class::<PyConnectionUpdate>()?;
-    m.add_class::<PyConnectionUpdateType>()?;
-    m.add_class::<PyNetworkingHandle>()?;
-
-    Ok(())
+// Manually submit the run()/recv() stub because the cancelhandle is poorly understood
+submit! {
+    gen_methods_from_python! {
+        r#"
+        class PyPeer:
+            async def run(self): ...
+            async def recv(self) -> PySwarmEvent: ...
+        "#
+    }
+}
+
+#[gen_stub_pyclass]
+#[pyclass]
+pub struct PySwarmEvent(FromSwarm);
+
+#[gen_stub_pymethods]
+#[pymethods]
+impl PySwarmEvent {
+    // probably a better way to do this, but...
+    fn downcast_discovered(&self) -> Option<String> {
+        if let FromSwarm::Discovered(peer_id) = self.0 {
+            Some(peer_id.to_base58())
+        } else {
+            None
+        }
+    }
+    fn downcast_expired(&self) -> Option<String> {
+        if let FromSwarm::Expired(peer_id) = self.0 {
+            Some(peer_id.to_base58())
+        } else {
+            None
+        }
+    }
+    fn downcast_message<'py>(
+        &self,
+        py: Python<'py>,
+    ) -> Option<(String, String, Bound<'py, PyBytes>)> {
+        if let FromSwarm::Message(peer_id, topic, data) = &self.0 {
+            Some((peer_id.to_base58(), topic.clone(), PyBytes::new(py, data)))
+        } else {
+            None
+        }
+    }
 }
--- a/rust/exo_pyo3_bindings/src/pylibp2p/ident.rs
+++ b/rust/exo_pyo3_bindings/src/pylibp2p/ident.rs
@@ -1,159 +0,0 @@
-use crate::ext::ResultExt as _;
-use libp2p::PeerId;
-use libp2p::identity::Keypair;
-use pyo3::prelude::{PyBytesMethods as _, PyModule, PyModuleMethods as _};
-use pyo3::types::PyBytes;
-use pyo3::{Bound, PyResult, Python, pyclass, pymethods};
-use pyo3_stub_gen::derive::{gen_stub_pyclass, gen_stub_pymethods};
-
-/// Identity keypair of a node.
-#[gen_stub_pyclass]
-#[pyclass(name = "Keypair", frozen)]
-#[repr(transparent)]
-pub struct PyKeypair(pub Keypair);
-
-#[gen_stub_pymethods]
-#[pymethods]
-#[allow(clippy::needless_pass_by_value)]
-impl PyKeypair {
-    /// Generate a new Ed25519 keypair.
-    #[staticmethod]
-    fn generate_ed25519() -> Self {
-        Self(Keypair::generate_ed25519())
-    }
-
-    /// Generate a new ECDSA keypair.
-    #[staticmethod]
-    fn generate_ecdsa() -> Self {
-        Self(Keypair::generate_ecdsa())
-    }
-
-    /// Generate a new Secp256k1 keypair.
-    #[staticmethod]
-    fn generate_secp256k1() -> Self {
-        Self(Keypair::generate_secp256k1())
-    }
-
-    /// Decode a private key from a protobuf structure and parse it as a `Keypair`.
-    #[staticmethod]
-    fn from_protobuf_encoding(bytes: Bound<'_, PyBytes>) -> PyResult<Self> {
-        let bytes = Vec::from(bytes.as_bytes());
-        Ok(Self(Keypair::from_protobuf_encoding(&bytes).pyerr()?))
-    }
-
-    /// Decode an keypair from a DER-encoded secret key in PKCS#8 `PrivateKeyInfo`
-    /// format (i.e. unencrypted) as defined in [RFC5208].
-    ///
-    /// [RFC5208]: https://tools.ietf.org/html/rfc5208#section-5
-    #[staticmethod]
-    fn rsa_from_pkcs8(bytes: Bound<'_, PyBytes>) -> PyResult<Self> {
-        let mut bytes = Vec::from(bytes.as_bytes());
-        Ok(Self(Keypair::rsa_from_pkcs8(&mut bytes).pyerr()?))
-    }
-
-    /// Decode a keypair from a DER-encoded Secp256k1 secret key in an `ECPrivateKey`
-    /// structure as defined in [RFC5915].
-    ///
-    /// [RFC5915]: https://tools.ietf.org/html/rfc5915
-    #[staticmethod]
-    fn secp256k1_from_der(bytes: Bound<'_, PyBytes>) -> PyResult<Self> {
-        let mut bytes = Vec::from(bytes.as_bytes());
-        Ok(Self(Keypair::secp256k1_from_der(&mut bytes).pyerr()?))
-    }
-
-    #[staticmethod]
-    fn ed25519_from_bytes(bytes: Bound<'_, PyBytes>) -> PyResult<Self> {
-        let mut bytes = Vec::from(bytes.as_bytes());
-        Ok(Self(Keypair::ed25519_from_bytes(&mut bytes).pyerr()?))
-    }
-
-    /// Encode a private key as protobuf structure.
-    fn to_protobuf_encoding<'py>(&self, py: Python<'py>) -> PyResult<Bound<'py, PyBytes>> {
-        let bytes = self.0.to_protobuf_encoding().pyerr()?;
-        Ok(PyBytes::new(py, &bytes))
-    }
-
-    /// Convert the `Keypair` into the corresponding `PeerId`.
-    fn to_peer_id(&self) -> PyPeerId {
-        PyPeerId(self.0.public().to_peer_id())
-    }
-
-    // /// Hidden constructor for pickling support. TODO: figure out how to do pickling...
-    // #[gen_stub(skip)]
-    // #[new]
-    // fn py_new(bytes: Bound<'_, PyBytes>) -> PyResult<Self> {
-    //     Self::from_protobuf_encoding(bytes)
-    // }
-    //
-    // #[gen_stub(skip)]
-    // fn __setstate__(&mut self, state: Bound<'_, PyBytes>) -> PyResult<()> {
-    //     *self = Self::from_protobuf_encoding(state)?;
-    //     Ok(())
-    // }
-    //
-    // #[gen_stub(skip)]
-    // fn __getstate__<'py>(&self, py: Python<'py>) -> PyResult<Bound<'py, PyBytes>> {
-    //     self.to_protobuf_encoding(py)
-    // }
-    //
-    // #[gen_stub(skip)]
-    // pub fn __getnewargs__<'py>(&self, py: Python<'py>) -> PyResult<(Bound<'py, PyBytes>,)> {
-    //     Ok((self.to_protobuf_encoding(py)?,))
-    // }
-}
-
-/// Identifier of a peer of the network.
-///
-/// The data is a `CIDv0` compatible multihash of the protobuf encoded public key of the peer
-/// as specified in [specs/peer-ids](https://github.com/libp2p/specs/blob/master/peer-ids/peer-ids.md).
-#[gen_stub_pyclass]
-#[pyclass(name = "PeerId", frozen)]
-#[derive(Debug, Clone)]
-#[repr(transparent)]
-pub struct PyPeerId(pub PeerId);
-
-#[gen_stub_pymethods]
-#[pymethods]
-#[allow(clippy::needless_pass_by_value)]
-impl PyPeerId {
-    /// Generates a random peer ID from a cryptographically secure PRNG.
-    ///
-    /// This is useful for randomly walking on a DHT, or for testing purposes.
-    #[staticmethod]
-    fn random() -> Self {
-        Self(PeerId::random())
-    }
-
-    /// Parses a `PeerId` from bytes.
-    #[staticmethod]
-    fn from_bytes(bytes: Bound<'_, PyBytes>) -> PyResult<Self> {
-        let bytes = Vec::from(bytes.as_bytes());
-        Ok(Self(PeerId::from_bytes(&bytes).pyerr()?))
-    }
-
-    /// Returns a raw bytes representation of this `PeerId`.
-    fn to_bytes<'py>(&self, py: Python<'py>) -> Bound<'py, PyBytes> {
-        let bytes = self.0.to_bytes();
-        PyBytes::new(py, &bytes)
-    }
-
-    /// Returns a base-58 encoded string of this `PeerId`.
-    fn to_base58(&self) -> String {
-        self.0.to_base58()
-    }
-
-    fn __repr__(&self) -> String {
-        format!("PeerId({})", self.to_base58())
-    }
-
-    fn __str__(&self) -> String {
-        self.to_base58()
-    }
-}
-
-pub fn ident_submodule(m: &Bound<'_, PyModule>) -> PyResult<()> {
-    m.add_class::<PyKeypair>()?;
-    m.add_class::<PyPeerId>()?;
-
-    Ok(())
-}
--- a/rust/exo_pyo3_bindings/src/pylibp2p/mod.rs
+++ b/rust/exo_pyo3_bindings/src/pylibp2p/mod.rs
@@ -1,8 +0,0 @@
-//! A module for exposing Rust's libp2p datatypes over Pyo3
-//!
-//! TODO: right now we are coupled to libp2p's identity, but eventually we want to create our own
-//!       independent identity type of some kind or another. This may require handshaking.
-//!
-
-pub mod ident;
-pub mod multiaddr;
--- a/rust/exo_pyo3_bindings/src/pylibp2p/multiaddr.rs
+++ b/rust/exo_pyo3_bindings/src/pylibp2p/multiaddr.rs
@@ -1,81 +0,0 @@
-use crate::ext::ResultExt as _;
-use libp2p::Multiaddr;
-use pyo3::prelude::{PyBytesMethods as _, PyModule, PyModuleMethods as _};
-use pyo3::types::PyBytes;
-use pyo3::{Bound, PyResult, Python, pyclass, pymethods};
-use pyo3_stub_gen::derive::{gen_stub_pyclass, gen_stub_pymethods};
-use std::str::FromStr as _;
-
-/// Representation of a Multiaddr.
-#[gen_stub_pyclass]
-#[pyclass(name = "Multiaddr", frozen)]
-#[derive(Debug, Clone)]
-#[repr(transparent)]
-pub struct PyMultiaddr(pub Multiaddr);
-
-#[gen_stub_pymethods]
-#[pymethods]
-#[allow(clippy::needless_pass_by_value)]
-impl PyMultiaddr {
-    /// Create a new, empty multiaddress.
-    #[staticmethod]
-    fn empty() -> Self {
-        Self(Multiaddr::empty())
-    }
-
-    /// Create a new, empty multiaddress with the given capacity.
-    #[staticmethod]
-    fn with_capacity(n: usize) -> Self {
-        Self(Multiaddr::with_capacity(n))
-    }
-
-    /// Parse a `Multiaddr` value from its byte slice representation.
-    #[staticmethod]
-    fn from_bytes(bytes: Bound<'_, PyBytes>) -> PyResult<Self> {
-        let bytes = Vec::from(bytes.as_bytes());
-        Ok(Self(Multiaddr::try_from(bytes).pyerr()?))
-    }
-
-    /// Parse a `Multiaddr` value from its string representation.
-    #[staticmethod]
-    fn from_string(string: String) -> PyResult<Self> {
-        Ok(Self(Multiaddr::from_str(&string).pyerr()?))
-    }
-
-    /// Return the length in bytes of this multiaddress.
-    fn len(&self) -> usize {
-        self.0.len()
-    }
-
-    /// Returns true if the length of this multiaddress is 0.
-    fn is_empty(&self) -> bool {
-        self.0.is_empty()
-    }
-
-    /// Return a copy of this [`Multiaddr`]'s byte representation.
-    fn to_bytes<'py>(&self, py: Python<'py>) -> Bound<'py, PyBytes> {
-        let bytes = self.0.to_vec();
-        PyBytes::new(py, &bytes)
-    }
-
-    /// Convert a Multiaddr to a string.
-    fn to_string(&self) -> String {
-        self.0.to_string()
-    }
-
-    #[gen_stub(skip)]
-    fn __repr__(&self) -> String {
-        format!("Multiaddr({})", self.0)
-    }
-
-    #[gen_stub(skip)]
-    fn __str__(&self) -> String {
-        self.to_string()
-    }
-}
-
-pub fn multiaddr_submodule(m: &Bound<'_, PyModule>) -> PyResult<()> {
-    m.add_class::<PyMultiaddr>()?;
-
-    Ok(())
-}
--- a/rust/exo_pyo3_bindings/tests/dummy.rs
+++ b/rust/exo_pyo3_bindings/tests/dummy.rs
@@ -1,54 +0,0 @@
-#[cfg(test)]
-mod tests {
-    use core::mem::drop;
-    use core::option::Option::Some;
-    use core::time::Duration;
-    use tokio;
-    use tokio::sync::mpsc;
-
-    #[tokio::test]
-    async fn test_drop_channel() {
-        struct Ping;
-
-        let (tx, mut rx) = mpsc::channel::<Ping>(10);
-
-        let _ = tokio::spawn(async move {
-            println!("TASK: entered");
-
-            loop {
-                tokio::select! {
-                    result = rx.recv() => {
-                        match result {
-                            Some(_) => {
-                                println!("TASK: pinged");
-                            }
-                            None => {
-                                println!("TASK: closing channel");
-                                break;
-                            }
-                        }
-                    }
-                    _ = tokio::time::sleep(Duration::from_secs_f32(0.1)) => {
-                        println!("TASK: heartbeat");
-                    }
-                }
-            }
-
-            println!("TASK: exited");
-        });
-
-        let tx2 = tx.clone();
-
-        tokio::time::sleep(Duration::from_secs_f32(0.11)).await;
-
-        tx.send(Ping).await.expect("Should not fail");
-        drop(tx);
-
-        tokio::time::sleep(Duration::from_secs_f32(0.11)).await;
-
-        tx2.send(Ping).await.expect("Should not fail");
-        drop(tx2);
-
-        tokio::time::sleep(Duration::from_secs_f32(0.11)).await;
-    }
-}
--- a/rust/networking/Cargo.toml
+++ b/rust/networking/Cargo.toml
@@ -13,32 +13,14 @@ path = "src/lib.rs"
 workspace = true

 [dependencies]
-# datastructures
-either = { workspace = true }
-
-# macro dependencies
-extend = { workspace = true }
-delegate = { workspace = true }
-impl-trait-for-tuples = { workspace = true }
-derive_more = { workspace = true }
-
 # async
 tokio = { workspace = true, features = ["full"] }
-futures = { workspace = true }
-futures-timer = { workspace = true }

 # utility dependencies
-util = { workspace = true }
-thiserror = { workspace = true }
-#internment = { workspace = true }
-#recursion = { workspace = true }
-#generativity = { workspace = true }
-#itertools = { workspace = true }
 tracing-subscriber = { version = "0.3.19", features = ["default", "env-filter"] }
-keccak-const = { workspace = true }

 # tracing/logging
 log = { workspace = true }

 # networking
-libp2p = { workspace = true, features = ["full"] }
+libp2p = { workspace = true, features = ["full"] }
--- a/rust/networking/examples/chatroom.rs
+++ b/rust/networking/examples/chatroom.rs
@@ -1,6 +1,6 @@
-use futures::stream::StreamExt as _;
-use libp2p::{gossipsub, identity, swarm::SwarmEvent};
-use networking::{discovery, swarm};
+use libp2p::identity;
+use networking::{self, FromSwarm, ToSwarm};
+use tokio::sync::mpsc;
 use tokio::{io, io::AsyncBufReadExt as _, select};
 use tracing_subscriber::EnvFilter;
 use tracing_subscriber::filter::LevelFilter;
@@ -12,63 +12,51 @@ async fn main() {
        .try_init();

    // Configure swarm
-    let mut swarm =
-        swarm::create_swarm(identity::Keypair::generate_ed25519()).expect("Swarm creation failed");
+    let (to_client, mut from_swarm) = mpsc::channel(20);
+    let (to_swarm, from_client) = mpsc::channel(20);
+    let mut peer = networking::Peer::new(
+        "chatroom!".to_string(),
+        identity::Keypair::generate_ed25519(),
+        to_client,
+        from_client,
+    )
+    .expect("listen error");

    // Create a Gossipsub topic & subscribe
-    let topic = gossipsub::IdentTopic::new("test-net");
-    swarm
-        .behaviour_mut()
-        .gossipsub
-        .subscribe(&topic)
-        .expect("Subscribing to topic failed");
-
    // Read full lines from stdin
    let mut stdin = io::BufReader::new(io::stdin()).lines();
    println!("Enter messages via STDIN and they will be sent to connected peers using Gossipsub");

+    let jh = tokio::spawn(async move { peer.run().await });
+    _ = to_swarm
+        .send(ToSwarm::Subscribe("chatting".to_string()))
+        .await;
+
    // Kick it off
    loop {
        select! {
            // on gossipsub outgoing
            Ok(Some(line)) = stdin.next_line() => {
-                if let Err(e) = swarm
-                    .behaviour_mut().gossipsub
-                    .publish(topic.clone(), line.as_bytes()) {
-                    println!("Publish error: {e:?}");
-                }
+                _ = to_swarm.send(ToSwarm::Message("chatting".to_string(), line.into_bytes())).await;
            }
-            event = swarm.select_next_some() => match event {
+            event = from_swarm.recv() => match event {
                // on gossipsub incoming
-                SwarmEvent::Behaviour(swarm::BehaviourEvent::Gossipsub(gossipsub::Event::Message {
-                    propagation_source: peer_id,
-                    message_id: id,
-                    message,
-                })) => println!(
-                        "\n\nGot message: '{}' with id: {id} from peer: {peer_id}\n\n",
-                        String::from_utf8_lossy(&message.data),
+                Some(FromSwarm::Message(peer_id,_, data)) => println!(
+                        "\n\nGot message: '{}' from peer: {peer_id}\n\n",
+                        String::from_utf8_lossy(&data),
                    ),

                // on discovery
-                SwarmEvent::Behaviour(swarm::BehaviourEvent::Discovery(e)) => match e {
-                    discovery::Event::ConnectionEstablished {
-                        peer_id, connection_id, remote_ip, remote_tcp_port
-                    } => {
-                        println!("\n\nConnected to: {peer_id}; connection ID: {connection_id}; remote IP: {remote_ip}; remote TCP port: {remote_tcp_port}\n\n");
-                    }
-                    discovery::Event::ConnectionClosed {
-                        peer_id, connection_id, remote_ip, remote_tcp_port
-                    } => {
-                        eprintln!("\n\nDisconnected from: {peer_id}; connection ID: {connection_id}; remote IP: {remote_ip}; remote TCP port: {remote_tcp_port}\n\n");
-                    }
+                Some(FromSwarm::Discovered(peer_id)) => {
+                    println!("\n\nConnected to: {peer_id}\n\n");
                }
-
-                // ignore outgoing errors: those are normal
-                e@SwarmEvent::OutgoingConnectionError { .. } => { log::debug!("Outgoing connection error: {e:?}"); }
-
-                // otherwise log any other event
-                e => { log::info!("Other event {e:?}"); }
+                Some(FromSwarm::Expired(peer_id)) => {
+                    println!("\n\nDisconnected from: {peer_id}\n\n");
+                }
+                Some(FromSwarm::PublishError(e)) => eprintln!("\n\nError {e:?}\n\n"),
+                None => break,
            }
        }
    }
+    _ = jh.await;
 }
--- a/rust/networking/examples/chatroom_manual.rs
+++ b/rust/networking/examples/chatroom_manual.rs
@@ -1,127 +0,0 @@
-// Copyright 2018 Parity Technologies (UK) Ltd.
-//
-// Permission is hereby granted, free of charge, to any person obtaining a
-// copy of this software and associated documentation files (the "Software"),
-// to deal in the Software without restriction, including without limitation
-// the rights to use, copy, modify, merge, publish, distribute, sublicense,
-// and/or sell copies of the Software, and to permit persons to whom the
-// Software is furnished to do so, subject to the following conditions:
-//
-// The above copyright notice and this permission notice shall be included in
-// all copies or substantial portions of the Software.
-//
-// THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
-// OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
-// FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
-// AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
-// LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING
-// FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER
-// DEALINGS IN THE SOFTWARE.
-
-use futures::stream::StreamExt;
-use libp2p::{
-    gossipsub, mdns, noise,
-    swarm::{NetworkBehaviour, SwarmEvent},
-    tcp, yamux,
-};
-use std::time::Duration;
-use std::{error::Error, hash::Hash};
-use tokio::{io, io::AsyncBufReadExt, select};
-use tracing_subscriber::EnvFilter;
-
-// We create a custom network behaviour that combines Gossipsub and Mdns.
-#[derive(NetworkBehaviour)]
-struct MyBehaviour {
-    gossipsub: gossipsub::Behaviour,
-    mdns: mdns::tokio::Behaviour,
-}
-
-#[tokio::main]
-async fn main() -> Result<(), Box<dyn Error>> {
-    let _ = tracing_subscriber::fmt()
-        .with_env_filter(EnvFilter::from_default_env())
-        .try_init();
-
-    let mut swarm = libp2p::SwarmBuilder::with_new_identity()
-        .with_tokio()
-        .with_tcp(
-            tcp::Config::default(),
-            noise::Config::new,
-            yamux::Config::default,
-        )?
-        .with_behaviour(|key| {
-            // Set a custom gossipsub configuration
-            let gossipsub_config = gossipsub::ConfigBuilder::default()
-                .heartbeat_interval(Duration::from_secs(10))
-                .validation_mode(gossipsub::ValidationMode::Strict) // This sets the kind of message validation. The default is Strict (enforce message signing)
-                .build()
-                .map_err(io::Error::other)?; // Temporary hack because `build` does not return a proper `std::error::Error`.
-
-            // build a gossipsub network behaviour
-            let gossipsub = gossipsub::Behaviour::new(
-                gossipsub::MessageAuthenticity::Signed(key.clone()),
-                gossipsub_config,
-            )?;
-
-            let mdns =
-                mdns::tokio::Behaviour::new(mdns::Config::default(), key.public().to_peer_id())?;
-            Ok(MyBehaviour { gossipsub, mdns })
-        })?
-        .build();
-
-    println!("Running swarm with identity {}", swarm.local_peer_id());
-
-    // Create a Gossipsub topic
-    let topic = gossipsub::IdentTopic::new("test-net");
-    // subscribes to our topic
-    swarm.behaviour_mut().gossipsub.subscribe(&topic)?;
-
-    // Read full lines from stdin
-    let mut stdin = io::BufReader::new(io::stdin()).lines();
-
-    // Listen on all interfaces and whatever port the OS assigns
-    swarm.listen_on("/ip4/0.0.0.0/tcp/0".parse()?)?;
-
-    println!("Enter messages via STDIN and they will be sent to connected peers using Gossipsub");
-
-    // Kick it off
-    loop {
-        select! {
-            Ok(Some(line)) = stdin.next_line() => {
-                if let Err(e) = swarm
-                    .behaviour_mut().gossipsub
-                    .publish(topic.clone(), line.as_bytes()) {
-                    println!("Publish error: {e:?}");
-                }
-            }
-            event = swarm.select_next_some() => match event {
-                SwarmEvent::Behaviour(MyBehaviourEvent::Mdns(mdns::Event::Discovered(list))) => {
-                    for (peer_id, multiaddr) in list {
-                        println!("mDNS discovered a new peer: {peer_id} on {multiaddr}");
-                        swarm.behaviour_mut().gossipsub.add_explicit_peer(&peer_id);
-                    }
-                },
-                SwarmEvent::Behaviour(MyBehaviourEvent::Mdns(mdns::Event::Expired(list))) => {
-                    for (peer_id, multiaddr) in list {
-                        println!("mDNS discover peer has expired: {peer_id} on {multiaddr}");
-                        swarm.behaviour_mut().gossipsub.remove_explicit_peer(&peer_id);
-                    }
-                },
-                SwarmEvent::Behaviour(MyBehaviourEvent::Gossipsub(gossipsub::Event::Message {
-                    propagation_source: peer_id,
-                    message_id: id,
-                    message,
-                })) => println!(
-                        "Got message: '{}' with id: {id} from peer: {peer_id}",
-                        String::from_utf8_lossy(&message.data),
-                    ),
-                SwarmEvent::NewListenAddr { address, .. } => {
-                    println!("Local node is listening on {address}");
-                }
-                e => {
-                    println!("Other swarm event: {:?}", e);
-                }
-            }
-        }
-    }
-}
--- a/rust/networking/src/RESEARCH_NOTES.txt
+++ b/rust/networking/src/RESEARCH_NOTES.txt
@@ -1,44 +0,0 @@
-https://github.com/ml-explore/mlx/commit/3fe98bacc7640d857acf3539f1d21b47a32e5609
-^raw sockets distributed -> `<net/ndrv.h>` -> https://newosxbook.com/code/xnu-3247.1.106/bsd/net/ndrv.h.auto.html
--> header file for a networking component found in the macOS kernel (XNU) that defines structures for network device driver registration, specifically the ndrv_demux_desc and ndrv_protocol_desc structures used for demultiplexing protocol data at the network interface level. It specifies how to describe protocol data, such as an Ethernet type or a SNAP header, and how to associate these descriptions with a specific protocol family to receive matching packets.
--> Used to bind an NDRV socket so that packets that match given protocol demux descriptions can be received.
--> An NDRV socket is a special kind of socket in the Darwin/macOS operating system's XNU kernel, used for low-level network packet manipulation and binding to specific protocols for packet processing. It allows user-space applications or drivers to directly write Layer 2 (L2) network packets or interact with the network stack at a lower level, often by binding to protocol descriptors like the ndrv_protocol_desc. This type of socket is used for functions such as capturing and injecting packets, especially in network infrastructure software like routers or for kernel-level network monitoring and security tools.
--> also called PF_NDRV sockets --> https://newosxbook.com/bonus/vol1ch16.html
----> they are conceptually similar to https://scapy.disruptivelabs.in/networking/socket-interface PF_RAW or PF_PACKET
-
-https://stackoverflow.com/questions/17169298/af-packet-on-osx
-^AF_PACKET duplicates the packets as soon as it receives them from the physical layer (for incoming packets) or just before sending them out to the physical layer (for outgoing packets). -> this is on Linux only
-^it doesn't exist on OS X so you can use /dev/bpfX (Berkeley Packet Filter) for sniffing
-
-https://www.unix.com/man_page/mojave/4/ip/
-^OS X manpages for IP
-
-https://developer.apple.com/documentation/kernel/implementing_drivers_system_extensions_and_kexts
-^driver kit, system extensions & kexts for macOS
-
----
-
-To set up a Linux system to use a Thunderbolt connection as a network device, connect the two computers with a Thunderbolt cable, load the thunderbolt-net kernel module (usually automatic but modprobe is an option for manual loading), and then the operating system will create virtual Ethernet interfaces (e.g., thunderbolt0) for networking. You can then use standard tools like ifconfig or your desktop environment's network manager to configure these new interfaces for a link-local network.
--> https://gist.github.com/geosp/80fbd39e617b7d1d9421683df4ea224a
----> here is a guide on how to set up thunderbolt-ethernet on linux
----> I may be able to steal the thunderbolt-net code ideas to implement a kernel module for MacOS
-
-https://chatgpt.com/s/t_68af8e41a8548191993281a014f846a7
-^GPT discussion about making socket interface
-
-https://chatgpt.com/s/t_68afb798a85c8191973c02a0fa7a48a3 --> link-local address,,??
-https://chatgpt.com/s/t_68afb02987e08191b2b0044d3667ece2
-^GPT discussion about accessing TB on MacOS low level interactions
-
--------------------------------
-
-https://www.intel.com/content/www/us/en/support/articles/000098893/software.html
-^Thunderbolt Share & Thunderbolt Networking Mode => intel's equivalent of thunderbolt bridge
-
-
---------------------------------
-
-https://www.zerotier.com/blog/how-zerotier-eliminated-kernel-extensions-on-macos/
-->fake ethernet devices on MacOS -> omg??? we can detect thunderbolt bridge, then bind to it, then re-expose it as fake ethernet??
-->ps: https://chatgpt.com/s/t_68afb2b25fb881919526763fb5d7359c, AF/PF_NDRV are one and the same!!!
-->https://github.com/zerotier/ZeroTierOne/blob/dev/osdep/MacEthernetTapAgent.c
--- a/rust/networking/src/discovery.rs
+++ b/rust/networking/src/discovery.rs
@@ -1,383 +0,0 @@
-use crate::ext::MultiaddrExt;
-use crate::keep_alive;
-use delegate::delegate;
-use either::Either;
-use futures::FutureExt;
-use futures_timer::Delay;
-use libp2p::core::transport::PortUse;
-use libp2p::core::{ConnectedPoint, Endpoint};
-use libp2p::swarm::behaviour::ConnectionEstablished;
-use libp2p::swarm::dial_opts::DialOpts;
-use libp2p::swarm::{
-    CloseConnection, ConnectionClosed, ConnectionDenied, ConnectionHandler,
-    ConnectionHandlerSelect, ConnectionId, FromSwarm, NetworkBehaviour, THandler, THandlerInEvent,
-    THandlerOutEvent, ToSwarm, dummy,
-};
-use libp2p::{Multiaddr, PeerId, identity, mdns};
-use std::collections::{BTreeSet, HashMap};
-use std::convert::Infallible;
-use std::io;
-use std::net::IpAddr;
-use std::task::{Context, Poll};
-use std::time::Duration;
-use util::wakerdeque::WakerDeque;
-
-const RETRY_CONNECT_INTERVAL: Duration = Duration::from_secs(5);
-
-mod managed {
-    use libp2p::swarm::NetworkBehaviour;
-    use libp2p::{identity, mdns, ping};
-    use std::io;
-    use std::time::Duration;
-
-    const MDNS_RECORD_TTL: Duration = Duration::from_secs(2_500);
-    const MDNS_QUERY_INTERVAL: Duration = Duration::from_secs(1_500);
-    const PING_TIMEOUT: Duration = Duration::from_millis(2_500);
-    const PING_INTERVAL: Duration = Duration::from_millis(2_500);
-
-    #[derive(NetworkBehaviour)]
-    pub struct Behaviour {
-        mdns: mdns::tokio::Behaviour,
-        ping: ping::Behaviour,
-    }
-
-    impl Behaviour {
-        pub fn new(keypair: &identity::Keypair) -> io::Result<Self> {
-            Ok(Self {
-                mdns: mdns_behaviour(keypair)?,
-                ping: ping_behaviour(),
-            })
-        }
-    }
-
-    fn mdns_behaviour(keypair: &identity::Keypair) -> io::Result<mdns::tokio::Behaviour> {
-        use mdns::{Config, tokio};
-
-        // mDNS config => enable IPv6
-        let mdns_config = Config {
-            ttl: MDNS_RECORD_TTL,
-            query_interval: MDNS_QUERY_INTERVAL,
-
-            // enable_ipv6: true, // TODO: for some reason, TCP+mDNS don't work well with ipv6?? figure out how to make work
-            ..Default::default()
-        };
-
-        let mdns_behaviour = tokio::Behaviour::new(mdns_config, keypair.public().to_peer_id());
-        Ok(mdns_behaviour?)
-    }
-
-    fn ping_behaviour() -> ping::Behaviour {
-        ping::Behaviour::new(
-            ping::Config::new()
-                .with_timeout(PING_TIMEOUT)
-                .with_interval(PING_INTERVAL),
-        )
-    }
-}
-
-/// Events for when a listening connection is truly established and truly closed.
-#[derive(Debug, Clone)]
-pub enum Event {
-    ConnectionEstablished {
-        peer_id: PeerId,
-        connection_id: ConnectionId,
-        remote_ip: IpAddr,
-        remote_tcp_port: u16,
-    },
-    ConnectionClosed {
-        peer_id: PeerId,
-        connection_id: ConnectionId,
-        remote_ip: IpAddr,
-        remote_tcp_port: u16,
-    },
-}
-
-/// Discovery behavior that wraps mDNS to produce truly discovered durable peer-connections.
-///
-/// The behaviour operates as such:
-///  1) All true (listening) connections/disconnections are tracked, emitting corresponding events
-///     to the swarm.
-///  1) mDNS discovered/expired peers are tracked; discovered but not connected peers are dialed
-///     immediately, and expired but connected peers are disconnected from immediately.
-///  2) Every fixed interval: discovered but not connected peers are dialed, and expired but
-///     connected peers are disconnected from.
-pub struct Behaviour {
-    // state-tracking for managed behaviors & mDNS-discovered peers
-    managed: managed::Behaviour,
-    mdns_discovered: HashMap<PeerId, BTreeSet<Multiaddr>>,
-
-    retry_delay: Delay, // retry interval
-
-    // pending events to emmit => waker-backed Deque to control polling
-    pending_events: WakerDeque<ToSwarm<Event, Infallible>>,
-}
-
-impl Behaviour {
-    pub fn new(keypair: &identity::Keypair) -> io::Result<Self> {
-        Ok(Self {
-            managed: managed::Behaviour::new(keypair)?,
-            mdns_discovered: HashMap::new(),
-            retry_delay: Delay::new(RETRY_CONNECT_INTERVAL),
-            pending_events: WakerDeque::new(),
-        })
-    }
-
-    fn dial(&mut self, peer_id: PeerId, addr: Multiaddr) {
-        self.pending_events.push_back(ToSwarm::Dial {
-            opts: DialOpts::peer_id(peer_id).addresses(vec![addr]).build(),
-        })
-    }
-
-    fn close_connection(&mut self, peer_id: PeerId, connection: ConnectionId) {
-        // push front to make this IMMEDIATE
-        self.pending_events.push_front(ToSwarm::CloseConnection {
-            peer_id,
-            connection: CloseConnection::One(connection),
-        })
-    }
-
-    fn handle_mdns_discovered(&mut self, peers: Vec<(PeerId, Multiaddr)>) {
-        for (p, ma) in peers {
-            self.dial(p, ma.clone()); // always connect
-
-            // get peer's multi-addresses or insert if missing
-            let Some(mas) = self.mdns_discovered.get_mut(&p) else {
-                self.mdns_discovered.insert(p, BTreeSet::from([ma]));
-                continue;
-            };
-
-            // multiaddress should never already be present - else something has gone wrong
-            let is_new_addr = mas.insert(ma);
-            assert!(is_new_addr, "cannot discover a discovered peer");
-        }
-    }
-
-    fn handle_mdns_expired(&mut self, peers: Vec<(PeerId, Multiaddr)>) {
-        for (p, ma) in peers {
-            // at this point, we *must* have the peer
-            let mas = self
-                .mdns_discovered
-                .get_mut(&p)
-                .expect("nonexistent peer cannot expire");
-
-            // at this point, we *must* have the multiaddress
-            let was_present = mas.remove(&ma);
-            assert!(was_present, "nonexistent multiaddress cannot expire");
-
-            // if empty, remove the peer-id entirely
-            if mas.is_empty() {
-                self.mdns_discovered.remove(&p);
-            }
-        }
-    }
-
-    fn on_connection_established(
-        &mut self,
-        peer_id: PeerId,
-        connection_id: ConnectionId,
-        remote_ip: IpAddr,
-        remote_tcp_port: u16,
-    ) {
-        // send out connected event
-        self.pending_events
-            .push_back(ToSwarm::GenerateEvent(Event::ConnectionEstablished {
-                peer_id,
-                connection_id,
-                remote_ip,
-                remote_tcp_port,
-            }));
-    }
-
-    fn on_connection_closed(
-        &mut self,
-        peer_id: PeerId,
-        connection_id: ConnectionId,
-        remote_ip: IpAddr,
-        remote_tcp_port: u16,
-    ) {
-        // send out disconnected event
-        self.pending_events
-            .push_back(ToSwarm::GenerateEvent(Event::ConnectionClosed {
-                peer_id,
-                connection_id,
-                remote_ip,
-                remote_tcp_port,
-            }));
-    }
-}
-
-impl NetworkBehaviour for Behaviour {
-    type ConnectionHandler =
-        ConnectionHandlerSelect<dummy::ConnectionHandler, THandler<managed::Behaviour>>;
-    type ToSwarm = Event;
-
-    // simply delegate to underlying mDNS behaviour
-
-    delegate! {
-        to self.managed {
-            fn handle_pending_inbound_connection(&mut self, connection_id: ConnectionId, local_addr: &Multiaddr, remote_addr: &Multiaddr) -> Result<(), ConnectionDenied>;
-            fn handle_pending_outbound_connection(&mut self, connection_id: ConnectionId, maybe_peer: Option<PeerId>, addresses: &[Multiaddr], effective_role: Endpoint) -> Result<Vec<Multiaddr>, ConnectionDenied>;
-        }
-    }
-
-    fn handle_established_inbound_connection(
-        &mut self,
-        connection_id: ConnectionId,
-        peer: PeerId,
-        local_addr: &Multiaddr,
-        remote_addr: &Multiaddr,
-    ) -> Result<THandler<Self>, ConnectionDenied> {
-        Ok(ConnectionHandler::select(
-            dummy::ConnectionHandler,
-            self.managed.handle_established_inbound_connection(
-                connection_id,
-                peer,
-                local_addr,
-                remote_addr,
-            )?,
-        ))
-    }
-
-    #[allow(clippy::needless_question_mark)]
-    fn handle_established_outbound_connection(
-        &mut self,
-        connection_id: ConnectionId,
-        peer: PeerId,
-        addr: &Multiaddr,
-        role_override: Endpoint,
-        port_use: PortUse,
-    ) -> Result<THandler<Self>, ConnectionDenied> {
-        Ok(ConnectionHandler::select(
-            dummy::ConnectionHandler,
-            self.managed.handle_established_outbound_connection(
-                connection_id,
-                peer,
-                addr,
-                role_override,
-                port_use,
-            )?,
-        ))
-    }
-
-    fn on_connection_handler_event(
-        &mut self,
-        peer_id: PeerId,
-        connection_id: ConnectionId,
-        event: THandlerOutEvent<Self>,
-    ) {
-        match event {
-            Either::Left(ev) => libp2p::core::util::unreachable(ev),
-            Either::Right(ev) => {
-                self.managed
-                    .on_connection_handler_event(peer_id, connection_id, ev)
-            }
-        }
-    }
-
-    // hook into these methods to drive behavior
-
-    fn on_swarm_event(&mut self, event: FromSwarm) {
-        self.managed.on_swarm_event(event); // let mDNS handle swarm events
-
-        // handle swarm events to update internal state:
-        match event {
-            FromSwarm::ConnectionEstablished(ConnectionEstablished {
-                peer_id,
-                connection_id,
-                endpoint,
-                ..
-            }) => {
-                let remote_address = match endpoint {
-                    ConnectedPoint::Dialer { address, .. } => address,
-                    ConnectedPoint::Listener { send_back_addr, .. } => send_back_addr,
-                };
-
-                if let Some((ip, port)) = remote_address.try_to_tcp_addr() {
-                    // handle connection established event which is filtered correctly
-                    self.on_connection_established(peer_id, connection_id, ip, port)
-                }
-            }
-            FromSwarm::ConnectionClosed(ConnectionClosed {
-                peer_id,
-                connection_id,
-                endpoint,
-                ..
-            }) => {
-                let remote_address = match endpoint {
-                    ConnectedPoint::Dialer { address, .. } => address,
-                    ConnectedPoint::Listener { send_back_addr, .. } => send_back_addr,
-                };
-
-                if let Some((ip, port)) = remote_address.try_to_tcp_addr() {
-                    // handle connection closed event which is filtered correctly
-                    self.on_connection_closed(peer_id, connection_id, ip, port)
-                }
-            }
-
-            // since we are running TCP/IP transport layer, we are assuming that
-            // no address changes can occur, hence encountering one is a fatal error
-            FromSwarm::AddressChange(a) => {
-                unreachable!("unhandlable: address change encountered: {:?}", a)
-            }
-            _ => {}
-        }
-    }
-
-    fn poll(&mut self, cx: &mut Context) -> Poll<ToSwarm<Self::ToSwarm, THandlerInEvent<Self>>> {
-        // delegate to managed behaviors for any behaviors they need to perform
-        match self.managed.poll(cx) {
-            Poll::Ready(ToSwarm::GenerateEvent(e)) => {
-                match e {
-                    // handle discovered and expired events from mDNS
-                    managed::BehaviourEvent::Mdns(e) => match e.clone() {
-                        mdns::Event::Discovered(peers) => {
-                            self.handle_mdns_discovered(peers);
-                        }
-                        mdns::Event::Expired(peers) => {
-                            self.handle_mdns_expired(peers);
-                        }
-                    },
-
-                    // handle ping events => if error then disconnect
-                    managed::BehaviourEvent::Ping(e) => {
-                        if let Err(_) = e.result {
-                            self.close_connection(e.peer, e.connection.clone())
-                        }
-                    }
-                }
-
-                // since we just consumed an event, we should immediately wake just in case
-                // there are more events to come where that came from
-                cx.waker().wake_by_ref();
-            }
-
-            // forward any other mDNS event to the swarm or its connection handler(s)
-            Poll::Ready(e) => {
-                return Poll::Ready(
-                    e.map_out(|_| unreachable!("events returning to swarm already handled"))
-                        .map_in(Either::Right),
-                );
-            }
-
-            Poll::Pending => {}
-        }
-
-        // retry connecting to all mDNS peers periodically (fails safely if already connected)
-        if self.retry_delay.poll_unpin(cx).is_ready() {
-            for (p, mas) in self.mdns_discovered.clone() {
-                for ma in mas {
-                    self.dial(p, ma)
-                }
-            }
-            self.retry_delay.reset(RETRY_CONNECT_INTERVAL) // reset timeout
-        }
-
-        // send out any pending events from our own service
-        if let Some(e) = self.pending_events.pop_front(cx) {
-            return Poll::Ready(e.map_in(Either::Left));
-        }
-
-        // wait for pending events
-        Poll::Pending
-    }
-}
--- a/rust/networking/src/keep_alive.rs
+++ b/rust/networking/src/keep_alive.rs
@@ -1,44 +0,0 @@
-use delegate::delegate;
-use libp2p::swarm::handler::ConnectionEvent;
-use libp2p::swarm::{ConnectionHandlerEvent, SubstreamProtocol, dummy, handler};
-use std::task::{Context, Poll};
-
-/// An implementation of [`ConnectionHandler`] that doesn't handle any protocols, but it keeps
-/// the connection alive.
-#[derive(Clone)]
-#[repr(transparent)]
-pub struct ConnectionHandler(dummy::ConnectionHandler);
-
-impl ConnectionHandler {
-    pub fn new() -> Self {
-        ConnectionHandler(dummy::ConnectionHandler)
-    }
-}
-
-impl handler::ConnectionHandler for ConnectionHandler {
-    // delegate types and implementation mostly to dummy handler
-    type FromBehaviour = <dummy::ConnectionHandler as handler::ConnectionHandler>::FromBehaviour;
-    type ToBehaviour = <dummy::ConnectionHandler as handler::ConnectionHandler>::ToBehaviour;
-    type InboundProtocol =
-        <dummy::ConnectionHandler as handler::ConnectionHandler>::InboundProtocol;
-    type OutboundProtocol =
-        <dummy::ConnectionHandler as handler::ConnectionHandler>::OutboundProtocol;
-    type InboundOpenInfo =
-        <dummy::ConnectionHandler as handler::ConnectionHandler>::InboundOpenInfo;
-    type OutboundOpenInfo =
-        <dummy::ConnectionHandler as handler::ConnectionHandler>::OutboundOpenInfo;
-
-    delegate! {
-        to self.0 {
-            fn listen_protocol(&self) -> SubstreamProtocol<Self::InboundProtocol, Self::InboundOpenInfo>;
-            fn poll(&mut self, cx: &mut Context<'_>) -> Poll<ConnectionHandlerEvent<Self::OutboundProtocol, Self::OutboundOpenInfo, Self::ToBehaviour>>;
-            fn on_behaviour_event(&mut self, event: Self::FromBehaviour);
-            fn on_connection_event(&mut self, event: ConnectionEvent<Self::InboundProtocol, Self::OutboundProtocol, Self::InboundOpenInfo, Self::OutboundOpenInfo>);
-        }
-    }
-
-    // specifically override this to force connection to stay alive
-    fn connection_keep_alive(&self) -> bool {
-        true
-    }
-}
--- a/rust/networking/src/lib.rs
+++ b/rust/networking/src/lib.rs
@@ -1,64 +1,299 @@
-//! TODO: crate documentation
-//!
-//! this is here as a placeholder documentation
-//!
-//!
+use libp2p::{
+    Multiaddr, PeerId,
+    futures::StreamExt,
+    gossipsub::{self, TopicHash},
+    identify,
+    identity::Keypair,
+    mdns,
+    swarm::{NetworkBehaviour, SwarmEvent, dial_opts::DialOpts},
+};
+use std::collections::HashMap;
+use tokio::sync::mpsc;

-// enable Rust-unstable features for convenience
-#![feature(trait_alias)]
-// #![feature(stmt_expr_attributes)]
-// #![feature(unboxed_closures)]
-// #![feature(assert_matches)]
-// #![feature(async_fn_in_dyn_trait)]
-// #![feature(async_for_loop)]
-// #![feature(auto_traits)]
-// #![feature(negative_impls)]
+#[derive(Debug)]
+pub struct ListenError;

-pub mod discovery;
-pub mod keep_alive;
-pub mod swarm;
-
-/// Namespace for all the type/trait aliases used by this crate.
-pub(crate) mod alias {
-    use std::error::Error;
-
-    pub type AnyError = Box<dyn Error + Send + Sync + 'static>;
-    pub type AnyResult<T> = Result<T, AnyError>;
+pub enum FromSwarm {
+    PublishError(gossipsub::PublishError),
+    Discovered(PeerId),
+    Expired(PeerId),
+    Message(PeerId, String, Vec<u8>),
+}
+pub enum ToSwarm {
+    Message(String, Vec<u8>),
+    Subscribe(String),
+    Unsubscribe(String),
 }

-/// Namespace for crate-wide extension traits/methods
-pub(crate) mod ext {
-    use extend::ext;
-    use libp2p::Multiaddr;
-    use libp2p::multiaddr::Protocol;
-    use std::net::IpAddr;
+pub struct Peer {
+    pub swarm: libp2p::Swarm<Behaviour>,
+    to_client: mpsc::Sender<FromSwarm>,
+    from_client: mpsc::Receiver<ToSwarm>,
+    namespace: String,
+    known_peers: HashMap<PeerId, Vec<Multiaddr>>,
+}
+impl Peer {
+    pub fn new(
+        namespace: String,
+        kp: Keypair,
+        to_client: mpsc::Sender<FromSwarm>,
+        from_client: mpsc::Receiver<ToSwarm>,
+    ) -> Result<Self, ListenError> {
+        let mut swarm = libp2p::SwarmBuilder::with_existing_identity(kp)
+            .with_tokio()
+            .with_quic()
+            // TODO(evan) .with_bandwidth_metrics()
+            .with_behaviour(|kp| Behaviour::new(namespace.clone(), kp))
+            .expect("invalid swarm behaviour")
+            .build();

-    #[ext(pub, name = MultiaddrExt)]
-    impl Multiaddr {
-        /// If the multiaddress corresponds to a TCP address, extracts it
-        fn try_to_tcp_addr(&self) -> Option<(IpAddr, u16)> {
-            let mut ps = self.into_iter();
-            let ip = if let Some(p) = ps.next() {
-                match p {
-                    Protocol::Ip4(ip) => IpAddr::V4(ip),
-                    Protocol::Ip6(ip) => IpAddr::V6(ip),
-                    _ => return None,
+        swarm
+            .listen_on("/ip6/::/udp/0/quic-v1".parse().expect("invalid multiaddr"))
+            .map_err(|_| ListenError)?;
+        swarm
+            .listen_on(
+                "/ip4/0.0.0.0/udp/0/quic-v1"
+                    .parse()
+                    .expect("invalid multiaddr"),
+            )
+            .map_err(|_| ListenError)?;
+        Ok(Self {
+            swarm,
+            to_client,
+            from_client,
+            namespace,
+            known_peers: HashMap::default(),
+        })
+    }
+    pub async fn run(&mut self) -> Result<(), ()> {
+        loop {
+            tokio::select! {
+                event = self.swarm.next() => self.handle_event(event.ok_or(())?).await?,
+                msg = self.from_client.recv() => self.handle_message(msg.ok_or(())?).await?,
+            }
+        }
+    }
+    async fn handle_message(&mut self, message: ToSwarm) -> Result<(), ()> {
+        match message {
+            ToSwarm::Message(topic, data) => {
+                if let Err(e) = self
+                    .swarm
+                    .behaviour_mut()
+                    .gossipsub
+                    .publish(TopicHash::from_raw(topic), data)
+                {
+                    self.to_client
+                        .send(FromSwarm::PublishError(e))
+                        .await
+                        .map_err(|_| ())?;
                }
-            } else {
-                return None;
-            };
-            let Some(Protocol::Tcp(port)) = ps.next() else {
-                return None;
-            };
-            Some((ip, port))
+            }
+            ToSwarm::Subscribe(topic) => {
+                match self
+                    .swarm
+                    .behaviour_mut()
+                    .gossipsub
+                    .subscribe(&gossipsub::IdentTopic::new(topic))
+                {
+                    Ok(_) => {}
+                    Err(gossipsub::SubscriptionError::NotAllowed) => {
+                        unreachable!("subscription filter hit")
+                    }
+                    Err(gossipsub::SubscriptionError::PublishError(e)) => self
+                        .to_client
+                        .send(FromSwarm::PublishError(e))
+                        .await
+                        .map_err(|_| ())?,
+                }
+            }
+            ToSwarm::Unsubscribe(topic) => {
+                self.swarm
+                    .behaviour_mut()
+                    .gossipsub
+                    .unsubscribe(&gossipsub::IdentTopic::new(topic));
+            }
+        }
+        Ok(())
+    }
+    async fn handle_event(&mut self, event: SwarmEvent<BehaviourEvent>) -> Result<(), ()> {
+        let SwarmEvent::Behaviour(event) = event else {
+            return Ok(());
+        };
+        match event {
+            BehaviourEvent::Gossipsub(gossipsub::Event::Message { message, .. }) => {
+                if let Some(source) = message.source {
+                    self.to_client
+                        .send(FromSwarm::Message(
+                            source,
+                            message.topic.into_string(),
+                            message.data,
+                        ))
+                        .await
+                        .map_err(|_| ())?;
+                }
+            }
+            BehaviourEvent::Identify(identify::Event::Received { peer_id, info, .. }) => {
+                log::debug!(
+                    "identify from {peer_id}: protocol_version='{}' agent_version='{}' (local namespace='{}')",
+                    info.protocol_version,
+                    info.agent_version,
+                    self.namespace
+                );
+                if info.protocol_version == self.namespace {
+                    self.passed_namespace(peer_id);
+                    self.to_client
+                        .send(FromSwarm::Discovered(peer_id))
+                        .await
+                        .map_err(|_| ())?;
+                } else {
+                    self.failed_namespace(peer_id);
+                }
+            }
+            BehaviourEvent::Mdns(mdns::Event::Discovered(v)) => {
+                for (peer_id, addr) in v {
+                    self.known_peers.entry(peer_id).or_default().push(addr);
+                }
+                for (peer_id, addrs) in &self.known_peers {
+                    // dialopts handles rate limiting, we should check errors if we want to blacklist earlier
+                    let _ = self
+                        .swarm
+                        .dial(DialOpts::peer_id(*peer_id).addresses(addrs.clone()).build());
+                }
+            }
+            BehaviourEvent::Mdns(mdns::Event::Expired(v)) => {
+                for (peer_id, addr) in v {
+                    let addrs = self.known_peers.entry(peer_id).or_default();
+                    addrs.retain(|a| *a != addr);
+                    if addrs.is_empty() {
+                        self.known_peers.remove(&peer_id);
+                        self.swarm
+                            .behaviour_mut()
+                            .gossipsub
+                            .remove_explicit_peer(&peer_id);
+                        self.to_client
+                            .send(FromSwarm::Expired(peer_id))
+                            .await
+                            .map_err(|_| ())?;
+                    }
+                }
+            }
+            _ => {}
+        }
+        Ok(())
+    }
+    fn passed_namespace(&mut self, peer_id: PeerId) {
+        self.swarm
+            .behaviour_mut()
+            .gossipsub
+            .remove_blacklisted_peer(&peer_id);
+        self.swarm
+            .behaviour_mut()
+            .gossipsub
+            .add_explicit_peer(&peer_id);
+    }
+    fn failed_namespace(&mut self, peer_id: PeerId) {
+        self.swarm
+            .behaviour_mut()
+            .gossipsub
+            .blacklist_peer(&peer_id);
+        self.swarm
+            .behaviour_mut()
+            .gossipsub
+            .remove_explicit_peer(&peer_id);
+    }
+}
+
+#[derive(NetworkBehaviour)]
+pub struct Behaviour {
+    gossipsub: gossipsub::Behaviour,
+    mdns: mdns::tokio::Behaviour,
+    identify: identify::Behaviour,
+}
+
+impl Behaviour {
+    fn new(namespace: String, kp: &Keypair) -> Self {
+        let mdns = mdns::Behaviour::new(mdns::Config::default(), kp.public().to_peer_id())
+            .expect("mdns behaviour failed to build");
+
+        let identify =
+            identify::Behaviour::new(identify::Config::new_with_signed_peer_record(namespace, kp));
+
+        let gossipsub = gossipsub::Behaviour::new(
+            gossipsub::MessageAuthenticity::Signed(kp.clone()),
+            gossipsub::ConfigBuilder::default()
+                .max_transmit_size(1024 * 1024)
+                .validation_mode(gossipsub::ValidationMode::Strict)
+                .build()
+                .expect("invalid gossipsub configuration"),
+        )
+        .expect("gossipsub behaviour failed ot build");
+
+        Self {
+            gossipsub,
+            mdns,
+            identify,
        }
    }
 }

-pub(crate) mod private {
-    #![allow(dead_code)]
+// TODO: more tests
+#[cfg(test)]
+mod tests {
+    use super::*;
+    use tokio::time::{Duration, timeout};

-    /// Sealed traits support
-    pub trait Sealed {}
-    impl<T: ?Sized> Sealed for T {}
+    fn make_peer(namespace: &str) -> (Peer, mpsc::Receiver<FromSwarm>, mpsc::Sender<ToSwarm>) {
+        let kp = Keypair::generate_ed25519();
+
+        let (to_client_tx, to_client_rx) = mpsc::channel(64);
+        let (to_peer_tx, to_peer_rx) = mpsc::channel(64);
+
+        let peer = Peer::new(namespace.to_string(), kp, to_client_tx, to_peer_rx)
+            .expect("Peer::new should succeed in tests");
+
+        (peer, to_client_rx, to_peer_tx)
+    }
+
+    async fn next_listen_addr(peer: &mut Peer) -> Multiaddr {
+        loop {
+            match peer.swarm.next().await {
+                Some(SwarmEvent::NewListenAddr { address, .. }) => return address,
+                Some(_) => {}
+                None => panic!("swarm stream ended unexpectedly"),
+            }
+        }
+    }
+
+    #[tokio::test]
+    async fn subscribe_and_unsubscribe_do_not_error() {
+        let (mut peer, mut events_rx, commands_tx) = make_peer("ns-test");
+
+        // Drive the swarm just enough to get at least one listen address event,
+        // so the background run loop has something initialized.
+        let _addr = next_listen_addr(&mut peer).await;
+
+        // Run the peer loop in the background.
+        let handle = tokio::spawn(async move {
+            let _ = peer.run().await;
+        });
+
+        commands_tx
+            .send(ToSwarm::Subscribe("topic-a".to_string()))
+            .await
+            .unwrap();
+
+        commands_tx
+            .send(ToSwarm::Unsubscribe("topic-a".to_string()))
+            .await
+            .unwrap();
+
+        // We don't *require* any FromSwarm events here; this is mainly a
+        // smoke test that the message-handling path doesn't panic/hang.
+        // Still, poll briefly to ensure the task is alive.
+        let _ = timeout(Duration::from_millis(200), events_rx.recv()).await;
+
+        // Shut down: dropping the command sender closes the channel, causing run() to return Err.
+        drop(commands_tx);
+        let _ = handle.await;
+    }
 }
--- a/rust/networking/src/swarm.rs
+++ b/rust/networking/src/swarm.rs
@@ -1,145 +0,0 @@
-use crate::alias;
-use crate::swarm::transport::tcp_transport;
-pub use behaviour::{Behaviour, BehaviourEvent};
-use libp2p::{SwarmBuilder, identity};
-
-pub type Swarm = libp2p::Swarm<Behaviour>;
-
-/// The current version of the network: this prevents devices running different versions of the
-/// software from interacting with each other.
-///
-/// TODO: right now this is a hardcoded constant; figure out what the versioning semantics should
-///       even be, and how to inject the right version into this config/initialization. E.g. should
-///       this be passed in as a parameter? What about rapidly changing versions in debug builds?
-///       this is all VERY very hard to figure out and needs to be mulled over as a team.
-pub const NETWORK_VERSION: &[u8] = b"v0.0.1";
-pub const OVERRIDE_VERSION_ENV_VAR: &str = "EXO_LIBP2P_NAMESPACE";
-
-/// Create and configure a swarm which listens to all ports on OS
-pub fn create_swarm(keypair: identity::Keypair) -> alias::AnyResult<Swarm> {
-    let mut swarm = SwarmBuilder::with_existing_identity(keypair)
-        .with_tokio()
-        .with_other_transport(tcp_transport)?
-        .with_behaviour(Behaviour::new)?
-        .build();
-
-    // Listen on all interfaces and whatever port the OS assigns
-    swarm.listen_on("/ip4/0.0.0.0/tcp/0".parse()?)?;
-    Ok(swarm)
-}
-
-mod transport {
-    use crate::alias;
-    use crate::swarm::{NETWORK_VERSION, OVERRIDE_VERSION_ENV_VAR};
-    use futures::{AsyncRead, AsyncWrite};
-    use keccak_const::Sha3_256;
-    use libp2p::core::muxing;
-    use libp2p::core::transport::Boxed;
-    use libp2p::pnet::{PnetError, PnetOutput};
-    use libp2p::{PeerId, Transport, identity, noise, pnet, yamux};
-    use std::{env, sync::LazyLock};
-
-    /// Key used for networking's private network; parametrized on the [`NETWORK_VERSION`].
-    /// See [`pnet_upgrade`] for more.
-    static PNET_PRESHARED_KEY: LazyLock<[u8; 32]> = LazyLock::new(|| {
-        let builder = Sha3_256::new().update(b"exo_discovery_network");
-
-        if let Ok(var) = env::var(OVERRIDE_VERSION_ENV_VAR) {
-            let bytes = var.into_bytes();
-            builder.update(&bytes)
-        } else {
-            builder.update(NETWORK_VERSION)
-        }
-        .finalize()
-    });
-
-    /// Make the Swarm run on a private network, as to not clash with public libp2p nodes and
-    /// also different-versioned instances of this same network.
-    /// This is implemented as an additional "upgrade" ontop of existing [`libp2p::Transport`] layers.
-    async fn pnet_upgrade<TSocket>(
-        socket: TSocket,
-        _: impl Sized,
-    ) -> Result<PnetOutput<TSocket>, PnetError>
-    where
-        TSocket: AsyncRead + AsyncWrite + Send + Unpin + 'static,
-    {
-        use pnet::{PnetConfig, PreSharedKey};
-        PnetConfig::new(PreSharedKey::new(*PNET_PRESHARED_KEY))
-            .handshake(socket)
-            .await
-    }
-
-    /// TCP/IP transport layer configuration.
-    pub fn tcp_transport(
-        keypair: &identity::Keypair,
-    ) -> alias::AnyResult<Boxed<(PeerId, muxing::StreamMuxerBox)>> {
-        use libp2p::{
-            core::upgrade::Version,
-            tcp::{Config, tokio},
-        };
-
-        // `TCP_NODELAY` enabled => avoid latency
-        let tcp_config = Config::default().nodelay(true);
-
-        // V1 + lazy flushing => 0-RTT negotiation
-        let upgrade_version = Version::V1Lazy;
-
-        // Noise is faster than TLS + we don't care much for security
-        let noise_config = noise::Config::new(keypair)?;
-
-        // Use default Yamux config for multiplexing
-        let yamux_config = yamux::Config::default();
-
-        // Create new Tokio-driven TCP/IP transport layer
-        let base_transport = tokio::Transport::new(tcp_config)
-            .and_then(pnet_upgrade)
-            .upgrade(upgrade_version)
-            .authenticate(noise_config)
-            .multiplex(yamux_config);
-
-        // Return boxed transport (to flatten complex type)
-        Ok(base_transport.boxed())
-    }
-}
-
-mod behaviour {
-    use crate::{alias, discovery};
-    use libp2p::swarm::NetworkBehaviour;
-    use libp2p::{gossipsub, identity};
-    use std::time::Duration;
-
-    /// Behavior of the Swarm which composes all desired behaviors:
-    /// Right now its just [`discovery::Behaviour`] and [`gossipsub::Behaviour`].
-    #[derive(NetworkBehaviour)]
-    pub struct Behaviour {
-        pub discovery: discovery::Behaviour,
-        pub gossipsub: gossipsub::Behaviour,
-    }
-
-    impl Behaviour {
-        pub fn new(keypair: &identity::Keypair) -> alias::AnyResult<Self> {
-            Ok(Self {
-                discovery: discovery::Behaviour::new(keypair)?,
-                gossipsub: gossipsub_behaviour(keypair),
-            })
-        }
-    }
-
-    fn gossipsub_behaviour(keypair: &identity::Keypair) -> gossipsub::Behaviour {
-        use gossipsub::{ConfigBuilder, MessageAuthenticity, ValidationMode};
-
-        // build a gossipsub network behaviour
-        //  => signed message authenticity + strict validation mode means the message-ID is
-        //     automatically provided by gossipsub w/out needing to provide custom message-ID function
-        gossipsub::Behaviour::new(
-            MessageAuthenticity::Signed(keypair.clone()),
-            ConfigBuilder::default()
-                .publish_queue_duration(Duration::from_secs(15))
-                .max_transmit_size(1024 * 1024)
-                .validation_mode(ValidationMode::Strict)
-                .build()
-                .expect("the configuration should always be valid"),
-        )
-        .expect("creating gossipsub behavior should always work")
-    }
-}
--- a/rust/networking/tests/dummy.rs
+++ b/rust/networking/tests/dummy.rs
@@ -1,7 +0,0 @@
-// maybe this will hold test in the future...??
-
-#[cfg(test)]
-mod tests {
-    #[test]
-    fn does_nothing() {}
-}
--- a/rust/rust-toolchain.toml
+++ b/rust/rust-toolchain.toml
@@ -1,2 +0,0 @@
-[toolchain]
-channel = "nightly"
--- a/rust/util/Cargo.toml
+++ b/rust/util/Cargo.toml
@@ -1,25 +0,0 @@
-[package]
-name = "util"
-version = { workspace = true }
-edition = { workspace = true }
-publish = false
-
-[lib]
-doctest = false
-name = "util"
-path = "src/lib.rs"
-
-[lints]
-workspace = true
-
-[dependencies]
-# macro dependencies
-extend = { workspace = true }
-
-# utility dependencies
-thiserror = { workspace = true }
-once_cell = { workspace = true }
-internment = { workspace = true }
-derive_more = { workspace = true }
-bon = { workspace = true }
-recursion = { workspace = true }
--- a/rust/util/src/lib.rs
+++ b/rust/util/src/lib.rs
@@ -1,53 +0,0 @@
-//! TODO: crate documentation
-//!
-//! this is here as a placeholder documentation
-//!
-//!
-
-// enable Rust-unstable features for convenience
-#![feature(trait_alias)]
-#![feature(stmt_expr_attributes)]
-#![feature(type_alias_impl_trait)]
-#![feature(specialization)]
-#![feature(unboxed_closures)]
-#![feature(const_trait_impl)]
-#![feature(fn_traits)]
-
-pub mod nonempty;
-pub mod wakerdeque;
-
-pub(crate) mod private {
-    // sealed traits support
-    pub trait Sealed {}
-    impl<T: ?Sized> Sealed for T {}
-}
-
-/// Namespace for all the type/trait aliases used by this crate.
-pub(crate) mod alias {}
-
-/// Namespace for crate-wide extension traits/methods
-pub mod ext {
-    use extend::ext;
-
-    #[ext(pub, name = BoxedSliceExt)]
-    impl<T> Box<[T]> {
-        #[inline]
-        fn map<B, F>(self, f: F) -> Box<[B]>
-        where
-            F: FnMut(T) -> B,
-        {
-            self.into_iter().map(f).collect()
-        }
-    }
-
-    #[ext(pub, name = VecExt)]
-    impl<T> Vec<T> {
-        #[inline]
-        fn map<B, F>(self, f: F) -> Vec<B>
-        where
-            F: FnMut(T) -> B,
-        {
-            self.into_iter().map(f).collect()
-        }
-    }
-}
--- a/rust/util/src/nonempty.rs
+++ b/rust/util/src/nonempty.rs
@@ -1,138 +0,0 @@
-use std::slice::SliceIndex;
-use std::{ops, slice};
-use thiserror::Error;
-
-#[derive(Error, Debug)]
-#[error("Cannot create to `NonemptyArray` because the supplied slice is empty")]
-pub struct EmptySliceError;
-
-/// A pointer to a non-empty fixed-size slice allocated on the heap.
-#[derive(Clone, Debug, PartialEq, Eq, Hash, PartialOrd, Ord)]
-#[repr(transparent)]
-pub struct NonemptyArray<T>(Box<[T]>);
-
-#[allow(clippy::arbitrary_source_item_ordering)]
-impl<T> NonemptyArray<T> {
-    #[inline]
-    pub fn singleton(value: T) -> Self {
-        Self(Box::new([value]))
-    }
-
-    #[allow(clippy::missing_errors_doc)]
-    #[inline]
-    pub fn try_from_boxed_slice<S: Into<Box<[T]>>>(
-        boxed_slice: S,
-    ) -> Result<Self, EmptySliceError> {
-        let boxed_slice = boxed_slice.into();
-        if boxed_slice.is_empty() {
-            Err(EmptySliceError)
-        } else {
-            Ok(Self(boxed_slice))
-        }
-    }
-
-    #[must_use]
-    #[inline]
-    pub fn into_boxed_slice(self) -> Box<[T]> {
-        self.0
-    }
-
-    #[must_use]
-    #[inline]
-    pub fn to_vec(&self) -> Vec<T>
-    where
-        T: Clone,
-    {
-        self.0.to_vec()
-    }
-
-    #[must_use]
-    #[inline]
-    pub const fn as_slice(&self) -> &[T] {
-        &self.0
-    }
-
-    #[allow(clippy::indexing_slicing)]
-    #[must_use]
-    #[inline]
-    pub fn first(&self) -> &T {
-        &self.0[0]
-    }
-
-    #[allow(clippy::indexing_slicing, clippy::arithmetic_side_effects)]
-    #[must_use]
-    #[inline]
-    pub fn last(&self) -> &T {
-        &self.0[self.0.len() - 1]
-    }
-
-    #[must_use]
-    #[inline]
-    pub fn get<I>(&self, index: I) -> Option<&I::Output>
-    where
-        I: SliceIndex<[T]>,
-    {
-        self.0.get(index)
-    }
-
-    #[allow(clippy::len_without_is_empty)]
-    #[must_use]
-    #[inline]
-    pub const fn len(&self) -> usize {
-        self.0.len()
-    }
-
-    #[allow(clippy::iter_without_into_iter)]
-    #[inline]
-    pub fn iter(&self) -> slice::Iter<'_, T> {
-        self.0.iter()
-    }
-
-    #[allow(clippy::iter_without_into_iter)]
-    #[inline]
-    pub fn iter_mut(&mut self) -> slice::IterMut<'_, T> {
-        self.0.iter_mut()
-    }
-
-    #[inline]
-    #[must_use]
-    pub fn map<U, F: FnMut(T) -> U>(self, f: F) -> NonemptyArray<U> {
-        NonemptyArray(self.0.into_iter().map(f).collect())
-    }
-}
-
-impl<T> From<NonemptyArray<T>> for Box<[T]> {
-    #[inline]
-    fn from(value: NonemptyArray<T>) -> Self {
-        value.into_boxed_slice()
-    }
-}
-
-impl<T> ops::Index<usize> for NonemptyArray<T> {
-    type Output = T;
-
-    #[inline]
-    fn index(&self, index: usize) -> &Self::Output {
-        self.0.index(index)
-    }
-}
-
-impl<T> IntoIterator for NonemptyArray<T> {
-    type Item = T;
-    type IntoIter = std::vec::IntoIter<T>;
-
-    #[inline]
-    fn into_iter(self) -> Self::IntoIter {
-        self.into_boxed_slice().into_vec().into_iter()
-    }
-}
-
-impl<'a, T> IntoIterator for &'a NonemptyArray<T> {
-    type Item = &'a T;
-    type IntoIter = slice::Iter<'a, T>;
-
-    #[inline]
-    fn into_iter(self) -> Self::IntoIter {
-        self.iter()
-    }
-}
--- a/rust/util/src/wakerdeque.rs
+++ b/rust/util/src/wakerdeque.rs
@@ -1,55 +0,0 @@
-use std::collections::VecDeque;
-use std::fmt::{Debug, Formatter};
-use std::task::{Context, Waker};
-
-/// A wrapper around [`VecDeque`] which wakes (if it can) on any `push_*` methods,
-/// and updates the internally stored waker by consuming [`Context`] on any `pop_*` methods.
-pub struct WakerDeque<T> {
-    waker: Option<Waker>,
-    deque: VecDeque<T>,
-}
-
-impl<T: Debug> Debug for WakerDeque<T> {
-    fn fmt(&self, f: &mut Formatter<'_>) -> std::fmt::Result {
-        self.deque.fmt(f)
-    }
-}
-
-impl<T> WakerDeque<T> {
-    pub fn new() -> Self {
-        Self {
-            waker: None,
-            deque: VecDeque::new(),
-        }
-    }
-
-    fn update(&mut self, cx: &mut Context<'_>) {
-        self.waker = Some(cx.waker().clone());
-    }
-
-    fn wake(&mut self) {
-        let Some(ref mut w) = self.waker else { return };
-        w.wake_by_ref();
-        self.waker = None;
-    }
-
-    pub fn pop_front(&mut self, cx: &mut Context<'_>) -> Option<T> {
-        self.update(cx);
-        self.deque.pop_front()
-    }
-
-    pub fn pop_back(&mut self, cx: &mut Context<'_>) -> Option<T> {
-        self.update(cx);
-        self.deque.pop_back()
-    }
-
-    pub fn push_front(&mut self, value: T) {
-        self.wake();
-        self.deque.push_front(value);
-    }
-
-    pub fn push_back(&mut self, value: T) {
-        self.wake();
-        self.deque.push_back(value);
-    }
-}
--- a/src/exo/download/coordinator.py
+++ b/src/exo/download/coordinator.py
@@ -53,11 +53,10 @@ class DownloadCoordinator:
    # Internal event channel for forwarding (initialized in __post_init__)
    event_sender: Sender[Event] = field(init=False)
    event_receiver: Receiver[Event] = field(init=False)
-    _tg: TaskGroup = field(init=False)
+    _tg: TaskGroup = field(init=False, default_factory=anyio.create_task_group)

    def __post_init__(self) -> None:
        self.event_sender, self.event_receiver = channel[Event]()
-        self._tg = anyio.create_task_group()

    async def run(self) -> None:
        logger.info("Starting DownloadCoordinator")
--- a/src/exo/main.py
+++ b/src/exo/main.py
@@ -1,4 +1,5 @@
 import argparse
+import importlib.metadata
 import itertools
 import multiprocessing as mp
 import os
@@ -27,7 +28,6 @@ from exo.utils.pydantic_ext import CamelCaseModel
 from exo.worker.main import Worker


-# I marked this as a dataclass as I want trivial constructors.
@dataclass
 class Node:
    router: Router
@@ -45,9 +45,9 @@ class Node:
    @classmethod
    async def create(cls, args: "Args") -> "Self":
        keypair = get_node_id_keypair()
-        node_id = NodeId(keypair.to_peer_id().to_base58())
+        node_id = NodeId(keypair.to_string())
        session_id = SessionId(master_node_id=node_id, election_clock=0)
-        router = Router.create(keypair)
+        router = Router.create(keypair, namespace=args.namespace)
        await router.register_topic(topics.GLOBAL_EVENTS)
        await router.register_topic(topics.LOCAL_EVENTS)
        await router.register_topic(topics.COMMANDS)
@@ -73,7 +73,7 @@ class Node:
        else:
            download_coordinator = None

-        if args.spawn_api:
+        if not args.no_api:
            api = API(
                node_id,
                session_id,
@@ -136,7 +136,6 @@ class Node:

    async def run(self):
        async with self._tg as tg:
-            signal.signal(signal.SIGINT, lambda _, __: self.shutdown())
            tg.start_soon(self.router.run)
            tg.start_soon(self.election.run)
            if self.download_coordinator:
@@ -148,6 +147,8 @@ class Node:
            if self.api:
                tg.start_soon(self.api.run)
            tg.start_soon(self._elect_loop)
+            signal.signal(signal.SIGINT, lambda _, __: self.shutdown())
+            signal.signal(signal.SIGTERM, lambda _, __: self.shutdown())

    def shutdown(self):
        # if this is our second call to shutdown, just sys.exit
@@ -253,7 +254,7 @@ def main():
    # TODO: Refactor the current verbosity system
    logger_setup(EXO_LOG, args.verbosity)
    logger.info("Starting EXO")
-    logger.info(f"EXO_LIBP2P_NAMESPACE: {os.getenv('EXO_LIBP2P_NAMESPACE')}")
+    logger.info(f"Namespace: {args.namespace}")

    # Set FAST_SYNCH override env var for runner subprocesses
    if args.fast_synch is True:
@@ -270,13 +271,13 @@ def main():


 class Args(CamelCaseModel):
-    verbosity: int = 0
-    force_master: bool = False
-    spawn_api: bool = False
-    api_port: PositiveInt = 52415
-    tb_only: bool = False
+    verbosity: int
+    force_master: bool
+    no_api: bool
+    api_port: PositiveInt
    no_worker: bool = False
    no_downloads: bool = False
+    namespace: str
    fast_synch: bool | None = None  # None = auto, True = force on, False = force off

    @classmethod
@@ -306,14 +307,15 @@ class Args(CamelCaseModel):
        )
        parser.add_argument(
            "--no-api",
-            action="store_false",
-            dest="spawn_api",
+            action="store_true",
+            help="Disable the API server for this node",
        )
        parser.add_argument(
            "--api-port",
            type=int,
            dest="api_port",
            default=52415,
+            help="Which port the API server will be available on",
        )
        parser.add_argument(
            "--no-worker",
@@ -324,6 +326,11 @@ class Args(CamelCaseModel):
            action="store_true",
            help="Disable the download coordinator (node won't download models)",
        )
+        parser.add_argument(
+            "--namespace",
+            default=importlib.metadata.version("exo"),
+            help="Set the EXO namespace to run multiple isolated clusters",
+        )
        fast_synch_group = parser.add_mutually_exclusive_group()
        fast_synch_group.add_argument(
            "--fast-synch",
--- a/src/exo/master/adapters/chat_completions.py
+++ b/src/exo/master/adapters/chat_completions.py
@@ -176,7 +176,7 @@ async def generate_chat_stream(
 async def collect_chat_response(
    command_id: CommandId,
    chunk_stream: AsyncGenerator[ErrorChunk | ToolCallChunk | TokenChunk, None],
-) -> AsyncGenerator[str]:
+) -> ChatCompletionResponse:
    """Collect all token chunks and return a single ChatCompletionResponse."""
    text_parts: list[str] = []
    tool_calls: list[ToolCall] = []
@@ -223,7 +223,7 @@ async def collect_chat_response(
    combined_text = "".join(text_parts)
    assert model is not None

-    yield ChatCompletionResponse(
+    return ChatCompletionResponse(
        id=command_id,
        created=int(time.time()),
        model=model,
@@ -241,5 +241,4 @@ async def collect_chat_response(
                finish_reason=finish_reason,
            )
        ],
-    ).model_dump_json()
-    return
+    )
--- a/src/exo/master/api.py
+++ b/src/exo/master/api.py
@@ -123,7 +123,6 @@ from exo.shared.types.commands import (
    PlaceInstance,
    SendInputChunk,
    StartDownload,
-    TaskCancelled,
    TaskFinished,
    TextGeneration,
 )
@@ -530,14 +529,16 @@ class API:
                        break

        except anyio.get_cancelled_exc_class():
-            command = TaskCancelled(cancelled_command_id=command_id)
-            with anyio.CancelScope(shield=True):
-                await self.command_sender.send(
-                    ForwarderCommand(origin=self.node_id, command=command)
-                )
+            # TODO: TaskCancelled
+            """
+            self.command_sender.send_nowait(
+                ForwarderCommand(origin=self.node_id, command=command)
+            )
+            """
            raise
        finally:
-            await self._send(TaskFinished(finished_command_id=command_id))
+            command = TaskFinished(finished_command_id=command_id)
+            await self._send(command)
            if command_id in self._text_generation_queues:
                del self._text_generation_queues[command_id]

@@ -632,14 +633,11 @@ class API:
                    "X-Accel-Buffering": "no",
                },
            )
-        else:
-            return StreamingResponse(
-                collect_chat_response(
-                    command.command_id,
-                    self._token_chunk_stream(command.command_id),
-                ),
-                media_type="application/json",
-            )
+
+        return await collect_chat_response(
+            command.command_id,
+            self._token_chunk_stream(command.command_id),
+        )

    async def bench_chat_completions(
        self, payload: BenchChatCompletionRequest
@@ -655,7 +653,8 @@ class API:
        command = TextGeneration(task_params=task_params)
        await self._send(command)

-        return await self._collect_text_generation_with_stats(command.command_id)
+        response = await self._collect_text_generation_with_stats(command.command_id)
+        return response

    async def _resolve_and_validate_text_model(self, model_id: ModelId) -> ModelId:
        """Validate a text model exists and return the resolved model ID.
@@ -857,11 +856,6 @@ class API:
                        del image_metadata[key]

        except anyio.get_cancelled_exc_class():
-            command = TaskCancelled(cancelled_command_id=command_id)
-            with anyio.CancelScope(shield=True):
-                await self.command_sender.send(
-                    ForwarderCommand(origin=self.node_id, command=command)
-                )
            raise
        finally:
            await self._send(TaskFinished(finished_command_id=command_id))
@@ -943,11 +937,6 @@ class API:

            return (images, stats if capture_stats else None)
        except anyio.get_cancelled_exc_class():
-            command = TaskCancelled(cancelled_command_id=command_id)
-            with anyio.CancelScope(shield=True):
-                await self.command_sender.send(
-                    ForwarderCommand(origin=self.node_id, command=command)
-                )
            raise
        finally:
            await self._send(TaskFinished(finished_command_id=command_id))
@@ -1331,29 +1320,40 @@ class API:
        ]

    async def run(self):
+        shutdown_ev = anyio.Event()
+
+        try:
+            async with create_task_group() as tg:
+                self._tg = tg
+                logger.info("Starting API")
+                tg.start_soon(self._apply_state)
+                tg.start_soon(self._pause_on_new_election)
+                tg.start_soon(self._cleanup_expired_images)
+                print_startup_banner(self.port)
+                tg.start_soon(self.run_api, shutdown_ev)
+                try:
+                    await anyio.sleep_forever()
+                finally:
+                    with anyio.CancelScope(shield=True):
+                        shutdown_ev.set()
+        finally:
+            self.command_sender.close()
+            self.global_event_receiver.close()
+
+    async def run_api(self, ev: anyio.Event):
        cfg = Config()
-        cfg.bind = f"0.0.0.0:{self.port}"
+        cfg.bind = [f"0.0.0.0:{self.port}"]
        # nb: shared.logging needs updating if any of this changes
        cfg.accesslog = None
        cfg.errorlog = "-"
        cfg.logger_class = InterceptLogger
-
-        async with create_task_group() as tg:
-            self._tg = tg
-            logger.info("Starting API")
-            tg.start_soon(self._apply_state)
-            tg.start_soon(self._pause_on_new_election)
-            tg.start_soon(self._cleanup_expired_images)
-            print_startup_banner(self.port)
+        with anyio.CancelScope(shield=True):
            await serve(
                cast(ASGIFramework, self.app),
                cfg,
-                shutdown_trigger=lambda: anyio.sleep_forever(),
+                shutdown_trigger=ev.wait,
            )

-        self.command_sender.close()
-        self.global_event_receiver.close()
-
    async def _apply_state(self):
        with self.global_event_receiver as events:
            async for f_event in events:
--- a/src/exo/master/main.py
+++ b/src/exo/master/main.py
@@ -21,7 +21,6 @@ from exo.shared.types.commands import (
    PlaceInstance,
    RequestEventLog,
    SendInputChunk,
-    TaskCancelled,
    TaskFinished,
    TestCommand,
    TextGeneration,
@@ -37,7 +36,6 @@ from exo.shared.types.events import (
    NodeTimedOut,
    TaskCreated,
    TaskDeleted,
-    TaskStatusUpdated,
    TraceEventData,
    TracesCollected,
    TracesMerged,
@@ -98,16 +96,18 @@ class Master:
    async def run(self):
        logger.info("Starting Master")

-        async with self._tg as tg:
-            tg.start_soon(self._event_processor)
-            tg.start_soon(self._command_processor)
-            tg.start_soon(self._loopback_processor)
-            tg.start_soon(self._plan)
-        self.global_event_sender.close()
-        self.local_event_receiver.close()
-        self.command_receiver.close()
-        self._loopback_event_sender.close()
-        self._loopback_event_receiver.close()
+        try:
+            async with self._tg as tg:
+                tg.start_soon(self._event_processor)
+                tg.start_soon(self._command_processor)
+                tg.start_soon(self._loopback_processor)
+                tg.start_soon(self._plan)
+        finally:
+            self.global_event_sender.close()
+            self.local_event_receiver.close()
+            self.command_receiver.close()
+            self._loopback_event_sender.close()
+            self._loopback_event_receiver.close()

    async def shutdown(self):
        logger.info("Stopping Master")
@@ -278,7 +278,7 @@ class Master:
                        case DeleteInstance():
                            placement = delete_instance(command, self.state.instances)
                            transition_events = get_transition_events(
-                                self.state.instances, placement, self.state.tasks
+                                self.state.instances, placement
                            )
                            generated_events.extend(transition_events)
                        case PlaceInstance():
@@ -290,7 +290,7 @@ class Master:
                                self.state.node_network,
                            )
                            transition_events = get_transition_events(
-                                self.state.instances, placement, self.state.tasks
+                                self.state.instances, placement
                            )
                            generated_events.extend(transition_events)
                        case CreateInstance():
@@ -300,7 +300,7 @@ class Master:
                                self.state.instances,
                            )
                            transition_events = get_transition_events(
-                                self.state.instances, placement, self.state.tasks
+                                self.state.instances, placement
                            )
                            generated_events.extend(transition_events)
                        case SendInputChunk(chunk=chunk):
@@ -310,18 +310,6 @@ class Master:
                                    chunk=chunk,
                                )
                            )
-                        case TaskCancelled():
-                            if (
-                                task_id := self.command_task_mapping.get(
-                                    command.cancelled_command_id
-                                )
-                            ) is not None:
-                                generated_events.append(
-                                    TaskStatusUpdated(
-                                        task_status=TaskStatus.Cancelled,
-                                        task_id=task_id,
-                                    )
-                                )
                        case TaskFinished():
                            generated_events.append(
                                TaskDeleted(
@@ -330,9 +318,10 @@ class Master:
                                    ]
                                )
                            )
-                            self.command_task_mapping.pop(
-                                command.finished_command_id, None
-                            )
+                            if command.finished_command_id in self.command_task_mapping:
+                                del self.command_task_mapping[
+                                    command.finished_command_id
+                                ]
                        case RequestEventLog():
                            # We should just be able to send everything, since other buffers will ignore old messages
                            for i in range(command.since_idx, len(self._event_log)):
@@ -382,7 +371,7 @@ class Master:
                        await self._handle_traces_collected(event)
                        continue

-                    logger.debug(f"Master indexing event: {str(event)[:100]}")
+                    logger.trace(f"Master indexing event: {str(event)[:100]}")
                    indexed = IndexedEvent(event=event, idx=len(self._event_log))
                    self.state = apply(self.state, indexed)

--- a/src/exo/master/placement.py
+++ b/src/exo/master/placement.py
@@ -20,15 +20,9 @@ from exo.shared.types.commands import (
    PlaceInstance,
 )
 from exo.shared.types.common import NodeId
-from exo.shared.types.events import (
-    Event,
-    InstanceCreated,
-    InstanceDeleted,
-    TaskStatusUpdated,
-)
+from exo.shared.types.events import Event, InstanceCreated, InstanceDeleted
 from exo.shared.types.memory import Memory
 from exo.shared.types.profiling import MemoryUsage, NodeNetworkInfo
-from exo.shared.types.tasks import Task, TaskId, TaskStatus
 from exo.shared.types.worker.instances import (
    Instance,
    InstanceId,
@@ -186,7 +180,6 @@ def delete_instance(
 def get_transition_events(
    current_instances: Mapping[InstanceId, Instance],
    target_instances: Mapping[InstanceId, Instance],
-    tasks: Mapping[TaskId, Task],
 ) -> Sequence[Event]:
    events: list[Event] = []

@@ -202,18 +195,6 @@ def get_transition_events(
    # find instances to delete
    for instance_id in current_instances:
        if instance_id not in target_instances:
-            for task in tasks.values():
-                if task.instance_id == instance_id and task.task_status in [
-                    TaskStatus.Pending,
-                    TaskStatus.Running,
-                ]:
-                    events.append(
-                        TaskStatusUpdated(
-                            task_status=TaskStatus.Cancelled,
-                            task_id=task.task_id,
-                        )
-                    )
-
            events.append(
                InstanceDeleted(
                    instance_id=instance_id,
--- a/src/exo/master/tests/test_master.py
+++ b/src/exo/master/tests/test_master.py
@@ -41,7 +41,7 @@ from exo.utils.channels import channel
@pytest.mark.asyncio
 async def test_master():
    keypair = get_node_id_keypair()
-    node_id = NodeId(keypair.to_peer_id().to_base58())
+    node_id = NodeId(keypair.to_string())
    session_id = SessionId(master_node_id=node_id, election_clock=0)

    ge_sender, global_event_receiver = channel[ForwarderEvent]()
@@ -72,7 +72,7 @@ async def test_master():
    async with anyio.create_task_group() as tg:
        tg.start_soon(master.run)

-        sender_node_id = NodeId(f"{keypair.to_peer_id().to_base58()}_sender")
+        sender_node_id = NodeId(f"{keypair.to_string()}_sender")
        # inject a NodeGatheredInfo event
        logger.info("inject a NodeGatheredInfo event")
        await local_event_sender.send(
--- a/src/exo/master/tests/test_placement.py
+++ b/src/exo/master/tests/test_placement.py
@@ -239,7 +239,7 @@ def test_get_transition_events_no_change(instance: Instance):
    target_instances = {instance_id: instance}

    # act
-    events = get_transition_events(current_instances, target_instances, {})
+    events = get_transition_events(current_instances, target_instances)

    # assert
    assert len(events) == 0
@@ -252,7 +252,7 @@ def test_get_transition_events_create_instance(instance: Instance):
    target_instances: dict[InstanceId, Instance] = {instance_id: instance}

    # act
-    events = get_transition_events(current_instances, target_instances, {})
+    events = get_transition_events(current_instances, target_instances)

    # assert
    assert len(events) == 1
@@ -266,7 +266,7 @@ def test_get_transition_events_delete_instance(instance: Instance):
    target_instances: dict[InstanceId, Instance] = {}

    # act
-    events = get_transition_events(current_instances, target_instances, {})
+    events = get_transition_events(current_instances, target_instances)

    # assert
    assert len(events) == 1
--- a/src/exo/routing/connection_message.py
+++ b/src/exo/routing/connection_message.py
@@ -1,37 +1,9 @@
-from enum import Enum
-
-from exo_pyo3_bindings import ConnectionUpdate, ConnectionUpdateType
-
 from exo.shared.types.common import NodeId
 from exo.utils.pydantic_ext import CamelCaseModel

 """Serialisable types for Connection Updates/Messages"""


-class ConnectionMessageType(Enum):
-    Connected = 0
-    Disconnected = 1
-
-    @staticmethod
-    def from_update_type(update_type: ConnectionUpdateType):
-        match update_type:
-            case ConnectionUpdateType.Connected:
-                return ConnectionMessageType.Connected
-            case ConnectionUpdateType.Disconnected:
-                return ConnectionMessageType.Disconnected
-
-
 class ConnectionMessage(CamelCaseModel):
    node_id: NodeId
-    connection_type: ConnectionMessageType
-    remote_ipv4: str
-    remote_tcp_port: int
-
-    @classmethod
-    def from_update(cls, update: ConnectionUpdate) -> "ConnectionMessage":
-        return cls(
-            node_id=NodeId(update.peer_id.to_base58()),
-            connection_type=ConnectionMessageType.from_update_type(update.update_type),
-            remote_ipv4=update.remote_ipv4,
-            remote_tcp_port=update.remote_tcp_port,
-        )
+    expired: bool
--- a/src/exo/routing/router.py
+++ b/src/exo/routing/router.py
@@ -1,5 +1,5 @@
 from copy import copy
-from itertools import count
+from dataclasses import dataclass, field
 from math import inf
 from os import PathLike
 from pathlib import Path
@@ -9,19 +9,19 @@ from anyio import (
    BrokenResourceError,
    ClosedResourceError,
    create_task_group,
+    move_on_after,
    sleep_forever,
 )
 from anyio.abc import TaskGroup
 from exo_pyo3_bindings import (
-    AllQueuesFullError,
    Keypair,
-    NetworkingHandle,
-    NoPeersSubscribedToTopicError,
+    PyPeer,
 )
 from filelock import FileLock
 from loguru import logger

 from exo.shared.constants import EXO_NODE_ID_KEYPAIR
+from exo.shared.types.common import NodeId
 from exo.utils.channels import Receiver, Sender, channel
 from exo.utils.pydantic_ext import CamelCaseModel

@@ -98,28 +98,32 @@ class TopicRouter[T: CamelCaseModel]:
        )


+@dataclass
 class Router:
-    @classmethod
-    def create(cls, identity: Keypair) -> "Router":
-        return cls(handle=NetworkingHandle(identity))
+    _peer: PyPeer
+    topic_routers: dict[str, TopicRouter[CamelCaseModel]] = field(
+        init=False, default_factory=dict
+    )
+    networking_receiver: Receiver[tuple[str, bytes]] = field(init=False)
+    _tmp_networking_sender: Sender[tuple[str, bytes]] | None = field(init=False)
+    _tg: TaskGroup | None = None

-    def __init__(self, handle: NetworkingHandle):
-        self.topic_routers: dict[str, TopicRouter[CamelCaseModel]] = {}
-        send, recv = channel[tuple[str, bytes]]()
-        self.networking_receiver: Receiver[tuple[str, bytes]] = recv
-        self._net: NetworkingHandle = handle
-        self._tmp_networking_sender: Sender[tuple[str, bytes]] | None = send
-        self._id_count = count()
-        self._tg: TaskGroup | None = None
+    def __post_init__(self):
+        self._tmp_networking_sender, self.networking_receiver = channel()
+
+    @classmethod
+    def create(cls, identity: Keypair, namespace: str) -> "Router":
+        return cls(_peer=PyPeer.new(identity, namespace))

    async def register_topic[T: CamelCaseModel](self, topic: TypedTopic[T]):
-        assert self._tg is None, "Attempted to register topic after setup time"
        send = self._tmp_networking_sender
        if send:
            self._tmp_networking_sender = None
        else:
            send = self.networking_receiver.clone_sender()
        router = TopicRouter[T](topic, send)
+        if self._tg is not None:
+            self._tg.start_soon(router.run)
        self.topic_routers[topic.topic] = cast(TopicRouter[CamelCaseModel], router)
        await self._networking_subscribe(str(topic.topic))

@@ -146,18 +150,24 @@ class Router:

    async def run(self):
        logger.debug("Starting Router")
-        async with create_task_group() as tg:
-            self._tg = tg
-            for topic in self.topic_routers:
-                router = self.topic_routers[topic]
-                tg.start_soon(router.run)
-            tg.start_soon(self._networking_recv)
-            tg.start_soon(self._networking_recv_connection_messages)
-            tg.start_soon(self._networking_publish)
-            # Router only shuts down if you cancel it.
-            await sleep_forever()
-        for topic in self.topic_routers:
-            await self._networking_unsubscribe(str(topic))
+        try:
+            async def _peer_run():
+                await self._peer.run()
+
+            async with create_task_group() as tg:
+                self._tg = tg
+                for topic in self.topic_routers:
+                    router = self.topic_routers[topic]
+                    tg.start_soon(router.run)
+                tg.start_soon(self._networking_recv)
+                tg.start_soon(self._networking_publish)
+                tg.start_soon(_peer_run)
+                # Router only shuts down if you cancel it.
+                await sleep_forever()
+        finally:
+            with move_on_after(1, shield=True):
+                for topic in self.topic_routers:
+                    await self._networking_unsubscribe(str(topic))

    async def shutdown(self):
        logger.debug("Shutting down Router")
@@ -166,47 +176,58 @@ class Router:
        self._tg.cancel_scope.cancel()

    async def _networking_subscribe(self, topic: str):
-        logger.info(f"Subscribing to {topic}")
-        await self._net.gossipsub_subscribe(topic)
+        await self._peer.subscribe(topic)
+        logger.info(f"Subscribed to {topic}")

    async def _networking_unsubscribe(self, topic: str):
-        logger.info(f"Unsubscribing from {topic}")
-        await self._net.gossipsub_unsubscribe(topic)
+        await self._peer.unsubscribe(topic)
+        logger.info(f"Unsubscribed from {topic}")

    async def _networking_recv(self):
        while True:
-            topic, data = await self._net.gossipsub_recv()
-            logger.trace(f"Received message on {topic} with payload {data}")
+            try:
+                swarm_event = await self._peer.recv()
+            except ValueError:
+                logger.error("Message too large for gossipsub, dropped")
+                continue
+            except ConnectionError:
+                logger.error("All peer queues full, network overloaded")
+                continue
+            except RuntimeError:
+                break
+
+            cm = None
+            if (peer_id := swarm_event.downcast_discovered()) is not None:
+                cm = ConnectionMessage(node_id=NodeId(peer_id), expired=False)
+            if (peer_id := swarm_event.downcast_expired()) is not None:
+                cm = ConnectionMessage(node_id=NodeId(peer_id), expired=True)
+
+            if cm is not None:
+                if CONNECTION_MESSAGES.topic in self.topic_routers:
+                    router = self.topic_routers[CONNECTION_MESSAGES.topic]
+                    assert router.topic.model_type == ConnectionMessage
+                    router = cast(TopicRouter[ConnectionMessage], router)
+                    await router.publish(cm)
+                continue
+
+            assert (msg := swarm_event.downcast_message()) is not None
+            _origin, topic, payload = msg
+            logger.debug(f"Received message on {topic} with payload {payload}")
            if topic not in self.topic_routers:
                logger.warning(f"Received message on unknown or inactive topic {topic}")
                continue

            router = self.topic_routers[topic]
-            await router.publish_bytes(data)
-
-    async def _networking_recv_connection_messages(self):
-        while True:
-            update = await self._net.connection_update_recv()
-            message = ConnectionMessage.from_update(update)
-            logger.trace(
-                f"Received message on connection_messages with payload {message}"
-            )
-            if CONNECTION_MESSAGES.topic in self.topic_routers:
-                router = self.topic_routers[CONNECTION_MESSAGES.topic]
-                assert router.topic.model_type == ConnectionMessage
-                router = cast(TopicRouter[ConnectionMessage], router)
-                await router.publish(message)
+            await router.publish_bytes(payload)

    async def _networking_publish(self):
        with self.networking_receiver as networked_items:
            async for topic, data in networked_items:
                try:
                    logger.trace(f"Sending message on {topic} with payload {data}")
-                    await self._net.gossipsub_publish(topic, data)
-                # As a hack, this also catches AllQueuesFull
-                # Need to fix that ASAP.
-                except (NoPeersSubscribedToTopicError, AllQueuesFullError):
-                    pass
+                    await self._peer.send(topic, data)
+                except RuntimeError:
+                    break


 def get_node_id_keypair(
@@ -217,7 +238,7 @@ def get_node_id_keypair(
    Obtain the :class:`PeerId` by from it.
    """
    # TODO(evan): bring back node id persistence once we figure out how to deal with duplicates
-    return Keypair.generate_ed25519()
+    return Keypair.generate()

    def lock_path(path: str | bytes | PathLike[str] | PathLike[bytes]) -> Path:
        return Path(str(path) + ".lock")
--- a/src/exo/shared/election.py
+++ b/src/exo/shared/election.py
@@ -86,28 +86,29 @@ class Election:

    async def run(self):
        logger.info("Starting Election")
-        async with create_task_group() as tg:
-            self._tg = tg
-            tg.start_soon(self._election_receiver)
-            tg.start_soon(self._connection_receiver)
-            tg.start_soon(self._command_counter)
+        try:
+            async with create_task_group() as tg:
+                self._tg = tg
+                tg.start_soon(self._election_receiver)
+                tg.start_soon(self._connection_receiver)
+                tg.start_soon(self._command_counter)

-            # And start an election immediately, that instantly resolves
-            candidates: list[ElectionMessage] = []
-            logger.debug("Starting initial campaign")
-            self._candidates = candidates
-            await self._campaign(candidates, campaign_timeout=0.0)
-            logger.debug("Initial campaign finished")
-
-        # Cancel and wait for the last election to end
-        if self._campaign_cancel_scope is not None:
-            logger.debug("Cancelling campaign")
-            self._campaign_cancel_scope.cancel()
-        if self._campaign_done is not None:
-            logger.debug("Waiting for campaign to finish")
-            await self._campaign_done.wait()
-        logger.debug("Campaign cancelled and finished")
-        logger.info("Election finished")
+                # And start an election immediately, that instantly resolves
+                candidates: list[ElectionMessage] = []
+                logger.debug("Starting initial campaign")
+                self._candidates = candidates
+                await self._campaign(candidates, campaign_timeout=0.0)
+                logger.debug("Initial campaign finished")
+        finally:
+            # Cancel and wait for the last election to end
+            if self._campaign_cancel_scope is not None:
+                logger.debug("Cancelling campaign")
+                self._campaign_cancel_scope.cancel()
+            if self._campaign_done is not None:
+                logger.debug("Waiting for campaign to finish")
+                await self._campaign_done.wait()
+            logger.debug("Campaign cancelled and finished")
+            logger.info("Election shutdown")

    async def elect(self, em: ElectionMessage) -> None:
        logger.debug(f"Electing: {em}")
--- a/src/exo/shared/tests/test_election.py
+++ b/src/exo/shared/tests/test_election.py
@@ -1,7 +1,7 @@
 import pytest
 from anyio import create_task_group, fail_after, move_on_after

-from exo.routing.connection_message import ConnectionMessage, ConnectionMessageType
+from exo.routing.connection_message import ConnectionMessage
 from exo.shared.election import Election, ElectionMessage, ElectionResult
 from exo.shared.types.commands import ForwarderCommand, TestCommand
 from exo.shared.types.common import NodeId, SessionId
@@ -330,9 +330,7 @@ async def test_connection_message_triggers_new_round_broadcast() -> None:
            await cm_tx.send(
                ConnectionMessage(
                    node_id=NodeId(),
-                    connection_type=ConnectionMessageType.Connected,
-                    remote_ipv4="",
-                    remote_tcp_port=0,
+                    expired=False,
                )
            )

--- a/src/exo/shared/types/commands.py
+++ b/src/exo/shared/types/commands.py
@@ -48,10 +48,6 @@ class DeleteInstance(BaseCommand):
    instance_id: InstanceId


-class TaskCancelled(BaseCommand):
-    cancelled_command_id: CommandId
-
-
 class TaskFinished(BaseCommand):
    finished_command_id: CommandId

@@ -88,7 +84,6 @@ Command = (
    | PlaceInstance
    | CreateInstance
    | DeleteInstance
-    | TaskCancelled
    | TaskFinished
    | SendInputChunk
 )
--- a/src/exo/shared/types/tasks.py
+++ b/src/exo/shared/types/tasks.py
@@ -24,7 +24,6 @@ class TaskStatus(str, Enum):
    Complete = "Complete"
    TimedOut = "TimedOut"
    Failed = "Failed"
-    Cancelled = "Cancelled"


 class BaseTask(TaggedModel):
@@ -61,11 +60,6 @@ class TextGeneration(BaseTask):  # emitted by Master
    error_message: str | None = Field(default=None)


-class CancelTask(BaseTask):
-    cancelled_task_id: TaskId
-    runner_id: RunnerId
-
-
 class ImageGeneration(BaseTask):  # emitted by Master
    command_id: CommandId
    task_params: ImageGenerationTaskParams
@@ -93,7 +87,6 @@ Task = (
    | LoadModel
    | StartWarmup
    | TextGeneration
-    | CancelTask
    | ImageGeneration
    | ImageEdits
    | Shutdown
--- a/src/exo/utils/channels.py
+++ b/src/exo/utils/channels.py
@@ -194,9 +194,10 @@ class MpReceiver[T]:
                raise EndOfStream from None
            return item

-    # nb: this function will not cancel particularly well
    async def receive_async(self) -> T:
-        return await to_thread.run_sync(self.receive, limiter=CapacityLimiter(1))
+        return await to_thread.run_sync(
+            self.receive, limiter=CapacityLimiter(1), abandon_on_cancel=True
+        )

    def close(self) -> None:
        if not self._state.closed.is_set():
--- a/src/exo/worker/engines/mlx/generator/generate.py
+++ b/src/exo/worker/engines/mlx/generator/generate.py
@@ -32,6 +32,7 @@ from exo.worker.engines.mlx.constants import (
 )
 from exo.worker.engines.mlx.utils_mlx import (
    apply_chat_template,
+    mx_barrier,
 )
 from exo.worker.runner.bootstrap import logger

@@ -135,6 +136,10 @@ def warmup_inference(

    logger.info("Generated ALL warmup tokens")

+    # TODO: Do we want an mx_barrier?
+    #  At least this version is actively incorrect, as it should use mx_barrier(group)
+    mx_barrier()
+
    return tokens_generated


@@ -398,3 +403,5 @@ def mlx_generate(
        # Limit accumulated_text to what's needed for stop sequence detection
        if max_stop_len > 0 and len(accumulated_text) > max_stop_len:
            accumulated_text = accumulated_text[-max_stop_len:]
+
+        # TODO: Do we want an mx_barrier?
--- a/src/exo/worker/engines/mlx/utils_mlx.py
+++ b/src/exo/worker/engines/mlx/utils_mlx.py
@@ -67,6 +67,8 @@ Group = mx.distributed.Group
 resource.setrlimit(resource.RLIMIT_NOFILE, (2048, 4096))


+# TODO: Test this
+#  ALSO https://github.com/exo-explore/exo/pull/233#discussion_r2549683673
 def get_weights_size(model_shard_meta: ShardMetadata) -> Memory:
    return Memory.from_float_kb(
        (model_shard_meta.end_layer - model_shard_meta.start_layer)
@@ -84,6 +86,30 @@ class ModelLoadingTimeoutError(Exception):
    pass


+def mx_barrier(group: Group | None = None):
+    mx.eval(
+        mx.distributed.all_sum(
+            mx.array(1.0),
+            stream=mx.default_stream(mx.Device(mx.cpu)),
+            group=group,
+        )
+    )
+
+
+def broadcast_from_zero(value: int, group: Group | None = None):
+    if group is None:
+        return value
+
+    if group.rank() == 0:
+        a = mx.array([value], dtype=mx.int32)
+    else:
+        a = mx.array([0], dtype=mx.int32)
+
+    m = mx.distributed.all_sum(a, stream=mx.Device(mx.DeviceType.cpu), group=group)
+    mx.eval(m)
+    return int(m.item())
+
+
 class HostList(RootModel[list[str]]):
    @classmethod
    def from_hosts(cls, hosts: list[Host]) -> "HostList":
@@ -536,23 +562,3 @@ def mlx_cleanup(
    import gc

    gc.collect()
-
-
-def mx_any(bool_: bool, group: Group | None) -> bool:
-    if group is None:
-        return bool_
-    num_true = mx.distributed.all_sum(
-        mx.array(bool_), group=group, stream=mx.default_stream(mx.Device(mx.cpu))
-    )
-    mx.eval(num_true)
-    return num_true.item() > 0
-
-
-def mx_barrier(group: Group | None):
-    if group is None:
-        return
-    mx.eval(
-        mx.distributed.all_sum(
-            mx.array(1.0), group=group, stream=mx.default_stream(mx.Device(mx.cpu))
-        )
-    )
--- a/src/exo/worker/main.py
+++ b/src/exo/worker/main.py
@@ -32,7 +32,6 @@ from exo.shared.types.events import (
 from exo.shared.types.multiaddr import Multiaddr
 from exo.shared.types.state import State
 from exo.shared.types.tasks import (
-    CancelTask,
    CreateRunner,
    DownloadModel,
    ImageEdits,
@@ -99,22 +98,23 @@ class Worker:
        info_send, info_recv = channel[GatheredInfo]()
        info_gatherer: InfoGatherer = InfoGatherer(info_send)

-        async with self._tg as tg:
-            tg.start_soon(info_gatherer.run)
-            tg.start_soon(self._forward_info, info_recv)
-            tg.start_soon(self.plan_step)
-            tg.start_soon(self._resend_out_for_delivery)
-            tg.start_soon(self._event_applier)
-            tg.start_soon(self._forward_events)
-            tg.start_soon(self._poll_connection_updates)
-
-        # Actual shutdown code - waits for all tasks to complete before executing.
-        self.local_event_sender.close()
-        self.command_sender.close()
-        self.download_command_sender.close()
-        async with create_task_group() as tg:
+        try:
+            async with self._tg as tg:
+                tg.start_soon(info_gatherer.run)
+                tg.start_soon(self._forward_info, info_recv)
+                tg.start_soon(self.plan_step)
+                tg.start_soon(self._resend_out_for_delivery)
+                tg.start_soon(self._event_applier)
+                tg.start_soon(self._forward_events)
+                tg.start_soon(self._poll_connection_updates)
+        finally:
+            # Actual shutdown code - waits for all tasks to complete before executing.
+            logger.info("Stopping Worker")
+            self.local_event_sender.close()
+            self.command_sender.close()
+            self.download_command_sender.close()
            for runner in self.runners.values():
-                tg.start_soon(runner.shutdown)
+                runner.shutdown()

    async def _forward_info(self, recv: Receiver[GatheredInfo]):
        with recv as info_stream:
@@ -218,22 +218,15 @@ class Worker:
                        )
                    )
                case Shutdown(runner_id=runner_id):
-                    runner = self.runners.pop(runner_id)
                    try:
                        with fail_after(3):
-                            await runner.start_task(task)
+                            await self.runners.pop(runner_id).start_task(task)
                    except TimeoutError:
                        await self.event_sender.send(
                            TaskStatusUpdated(
                                task_id=task.task_id, task_status=TaskStatus.TimedOut
                            )
                        )
-                    finally:
-                        await runner.shutdown()
-                case CancelTask(
-                    cancelled_task_id=cancelled_task_id, runner_id=runner_id
-                ):
-                    await self.runners[runner_id].cancel_task(cancelled_task_id)
                case ImageEdits() if task.task_params.total_input_chunks > 0:
                    # Assemble image from chunks and inject into task
                    cmd_id = task.command_id
@@ -271,18 +264,18 @@ class Worker:
                        del self.input_chunk_buffer[cmd_id]
                    if cmd_id in self.input_chunk_counts:
                        del self.input_chunk_counts[cmd_id]
-                    await self._start_runner_task(modified_task)
+                    await self.runners[self._task_to_runner_id(task)].start_task(
+                        modified_task
+                    )
                case task:
-                    await self._start_runner_task(task)
+                    await self.runners[self._task_to_runner_id(task)].start_task(task)

    def shutdown(self):
        self._tg.cancel_scope.cancel()

-    async def _start_runner_task(self, task: Task):
-        if (instance := self.state.instances.get(task.instance_id)) is not None:
-            await self.runners[
-                instance.shard_assignments.node_to_runner[self.node_id]
-            ].start_task(task)
+    def _task_to_runner_id(self, task: Task):
+        instance = self.state.instances[task.instance_id]
+        return instance.shard_assignments.node_to_runner[self.node_id]

    async def _nack_request(self, since_idx: int) -> None:
        # We request all events after (and including) the missing index.
@@ -321,6 +314,8 @@ class Worker:
            for event in self.out_for_delivery.copy().values():
                await self.local_event_sender.send(event)

+    ## Op Executors
+
    def _create_supervisor(self, task: CreateRunner) -> RunnerSupervisor:
        """Creates and stores a new AssignedRunner with initial downloading status."""
        runner = RunnerSupervisor.create(
@@ -341,7 +336,7 @@ class Worker:
                    session=self.session_id,
                    event=event,
                )
-                logger.debug(f"Worker published event {idx}: {str(event)[:100]}")
+                logger.trace(f"Worker published event {idx}: {str(event)[:100]}")
                await self.local_event_sender.send(fe)
                self.out_for_delivery[event.event_id] = fe

--- a/src/exo/worker/plan.py
+++ b/src/exo/worker/plan.py
@@ -4,7 +4,6 @@ from collections.abc import Mapping, Sequence

 from exo.shared.types.common import CommandId, NodeId
 from exo.shared.types.tasks import (
-    CancelTask,
    ConnectToGroup,
    CreateRunner,
    DownloadModel,
@@ -54,14 +53,13 @@ def plan(
 ) -> Task | None:
    # Python short circuiting OR logic should evaluate these sequentially.
    return (
-        _cancel_tasks(runners, tasks)
-        or _kill_runner(runners, all_runners, instances)
+        _kill_runner(runners, all_runners, instances)
        or _create_runner(node_id, runners, instances)
        or _model_needs_download(node_id, runners, global_download_status)
        or _init_distributed_backend(runners, all_runners)
        or _load_model(runners, all_runners, global_download_status)
        or _ready_to_warmup(runners, all_runners)
-        or _pending_tasks(runners, tasks, all_runners, input_chunk_buffer or {})
+        or _pending_tasks(runners, tasks, all_runners, input_chunk_buffer)
    )


@@ -272,7 +270,7 @@ def _pending_tasks(
    runners: Mapping[RunnerId, RunnerSupervisor],
    tasks: Mapping[TaskId, Task],
    all_runners: Mapping[RunnerId, RunnerStatus],
-    input_chunk_buffer: Mapping[CommandId, dict[int, str]],
+    input_chunk_buffer: Mapping[CommandId, dict[int, str]] | None = None,
 ) -> Task | None:
    for task in tasks.values():
        # for now, just forward chat completions
@@ -286,7 +284,7 @@ def _pending_tasks(
        if isinstance(task, ImageEdits) and task.task_params.total_input_chunks > 0:
            cmd_id = task.command_id
            expected = task.task_params.total_input_chunks
-            received = len(input_chunk_buffer.get(cmd_id, {}))
+            received = len((input_chunk_buffer or {}).get(cmd_id, {}))
            if received < expected:
                continue  # Wait for all chunks to arrive

@@ -294,33 +292,16 @@ def _pending_tasks(
            if task.instance_id != runner.bound_instance.instance.instance_id:
                continue

-            # the task status _should_ be set to completed by the LAST runner
-            # it is currently set by the first
-            # this is definitely a hack
+            # I have a design point here; this is a state race in disguise as the task status doesn't get updated to completed fast enough
+            # however, realistically the task status should be set to completed by the LAST runner, so this is a true race
+            # the actual solution is somewhat deeper than this bypass - TODO!
            if task.task_id in runner.completed:
                continue

+            # TODO: Check ordering aligns with MLX distributeds expectations.
+
            if isinstance(runner.status, RunnerReady) and all(
                isinstance(all_runners[global_runner_id], (RunnerReady, RunnerRunning))
                for global_runner_id in runner.bound_instance.instance.shard_assignments.runner_to_shard
            ):
                return task
-
-
-def _cancel_tasks(
-    runners: Mapping[RunnerId, RunnerSupervisor],
-    tasks: Mapping[TaskId, Task],
-) -> Task | None:
-    for task in tasks.values():
-        if task.task_status != TaskStatus.Cancelled:
-            continue
-        for runner_id, runner in runners.items():
-            if task.instance_id != runner.bound_instance.instance.instance_id:
-                continue
-            if task.task_id in runner.cancelled:
-                continue
-            return CancelTask(
-                instance_id=task.instance_id,
-                cancelled_task_id=task.task_id,
-                runner_id=runner_id,
-            )
--- a/src/exo/worker/runner/bootstrap.py
+++ b/src/exo/worker/runner/bootstrap.py
@@ -3,7 +3,7 @@ import os
 import loguru

 from exo.shared.types.events import Event, RunnerStatusUpdated
-from exo.shared.types.tasks import Task, TaskId
+from exo.shared.types.tasks import Task
 from exo.shared.types.worker.instances import BoundInstance, MlxJacclInstance
 from exo.shared.types.worker.runners import RunnerFailed
 from exo.utils.channels import ClosedResourceError, MpReceiver, MpSender
@@ -15,7 +15,6 @@ def entrypoint(
    bound_instance: BoundInstance,
    event_sender: MpSender[Event],
    task_receiver: MpReceiver[Task],
-    cancel_receiver: MpReceiver[TaskId],
    _logger: "loguru.Logger",
 ) -> None:
    fast_synch_override = os.environ.get("EXO_FAST_SYNCH")
@@ -39,7 +38,7 @@ def entrypoint(
    try:
        from exo.worker.runner.runner import main

-        main(bound_instance, event_sender, task_receiver, cancel_receiver)
+        main(bound_instance, event_sender, task_receiver)
    except ClosedResourceError:
        logger.warning("Runner communication closed unexpectedly")
    except Exception as e:
--- a/src/exo/worker/runner/runner.py
+++ b/src/exo/worker/runner/runner.py
@@ -1,6 +1,5 @@
 import base64
 import json
-import math
 import time
 from collections.abc import Generator
 from functools import cache
@@ -88,7 +87,6 @@ from exo.worker.engines.mlx.utils_mlx import (
    initialize_mlx,
    load_mlx_items,
    mlx_force_oom,
-    mx_any,
 )
 from exo.worker.runner.bootstrap import logger

@@ -113,7 +111,6 @@ def main(
    bound_instance: BoundInstance,
    event_sender: MpSender[Event],
    task_receiver: MpReceiver[Task],
-    cancel_receiver: MpReceiver[TaskId],
 ):
    instance, runner_id, shard_metadata = (
        bound_instance.instance,
@@ -128,15 +125,11 @@ def main(
        time.sleep(timeout)

    setup_start_time = time.time()
-    cancelled_tasks = set[TaskId]()

-    # type checker was unhappy with me - splitting these fixed it
-    inference_model: Model | None = None
-    image_model: DistributedImageModel | None = None
+    model: Model | DistributedImageModel | None = None
    tokenizer = None
    group = None
    kv_prefix_cache: KVPrefixCache | None = None
-    check_for_cancel_every: int | None = None

    current_status: RunnerStatus = RunnerIdle()
    logger.info("runner created")
@@ -149,7 +142,6 @@ def main(
            if task.task_id in seen:
                logger.warning("repeat task - potential error")
            seen.add(task.task_id)
-            cancelled_tasks.discard(TaskId("CANCEL_CURRENT_TASK"))
            event_sender.send(
                TaskStatusUpdated(task_id=task.task_id, task_status=TaskStatus.Running)
            )
@@ -195,7 +187,7 @@ def main(
                        time.sleep(0.5)

                    if ModelTask.TextGeneration in shard_metadata.model_card.tasks:
-                        inference_model, tokenizer = load_mlx_items(
+                        model, tokenizer = load_mlx_items(
                            bound_instance, group, on_timeout=on_model_load_timeout
                        )
                        logger.info(
@@ -207,7 +199,7 @@ def main(
                        ModelTask.TextToImage in shard_metadata.model_card.tasks
                        or ModelTask.ImageToImage in shard_metadata.model_card.tasks
                    ):
-                        image_model = initialize_image_model(bound_instance)
+                        model = initialize_image_model(bound_instance)
                    else:
                        raise ValueError(
                            f"Unknown model task(s): {shard_metadata.model_card.tasks}"
@@ -215,6 +207,8 @@ def main(
                    current_status = RunnerLoaded()
                    logger.info("runner loaded")
                case StartWarmup() if isinstance(current_status, RunnerLoaded):
+                    assert model
+
                    current_status = RunnerWarmingUp()
                    logger.info("runner warming up")
                    event_sender.send(
@@ -226,30 +220,15 @@ def main(

                    logger.info(f"warming up inference for instance: {instance}")
                    if ModelTask.TextGeneration in shard_metadata.model_card.tasks:
-                        assert inference_model
+                        assert not isinstance(model, DistributedImageModel)
                        assert tokenizer

-                        t = time.perf_counter()
                        toks = warmup_inference(
-                            model=inference_model,
+                            model=model,
                            tokenizer=tokenizer,
+                            # kv_prefix_cache=kv_prefix_cache,  # supply for warmup-time prefix caching
                        )
                        logger.info(f"warmed up by generating {toks} tokens")
-                        check_for_cancel_every = min(
-                            math.ceil(toks / (time.perf_counter() - t)), 100
-                        )
-                        if group is not None:
-                            check_for_cancel_every = int(
-                                mx.max(
-                                    mx.distributed.all_gather(
-                                        mx.array([check_for_cancel_every]), group=group
-                                    )
-                                ).item()
-                            )
-
-                        logger.info(
-                            f"runner checking for cancellation every {check_for_cancel_every} tokens"
-                        )
                        logger.info(
                            f"runner initialized in {time.time() - setup_start_time} seconds"
                        )
@@ -257,8 +236,8 @@ def main(
                        ModelTask.TextToImage in shard_metadata.model_card.tasks
                        or ModelTask.ImageToImage in shard_metadata.model_card.tasks
                    ):
-                        assert image_model
-                        image = warmup_image_generator(model=image_model)
+                        assert isinstance(model, DistributedImageModel)
+                        image = warmup_image_generator(model=model)
                        if image is not None:
                            logger.info(f"warmed up by generating {image.size} image")
                        else:
@@ -278,9 +257,9 @@ def main(
                        )
                    )
                    event_sender.send(TaskAcknowledged(task_id=task.task_id))
-                    assert inference_model
+
+                    assert model and not isinstance(model, DistributedImageModel)
                    assert tokenizer
-                    assert check_for_cancel_every

                    try:
                        _check_for_debug_prompts(task_params)
@@ -290,7 +269,7 @@ def main(

                        # Generate responses using the actual MLX generation
                        mlx_generator = mlx_generate(
-                            model=inference_model,
+                            model=model,
                            tokenizer=tokenizer,
                            task=task_params,
                            prompt=prompt,
@@ -314,11 +293,11 @@ def main(
                            patch_glm_tokenizer(tokenizer)

                        # GPT-OSS specific parsing to match other model formats.
-                        elif isinstance(inference_model, GptOssModel):
+                        elif isinstance(model, GptOssModel):
                            mlx_generator = parse_gpt_oss(mlx_generator)

                        if tokenizer.has_tool_calling and not isinstance(
-                            inference_model, GptOssModel
+                            model, GptOssModel
                        ):
                            assert tokenizer.tool_call_start
                            assert tokenizer.tool_call_end
@@ -331,18 +310,7 @@ def main(
                            )

                        completion_tokens = 0
-                        tokens_since_last_cancel_check = 0
                        for response in mlx_generator:
-                            tokens_since_last_cancel_check += 1
-                            if tokens_since_last_cancel_check >= check_for_cancel_every:
-                                tokens_since_last_cancel_check = 0
-                                cancelled_tasks.update(cancel_receiver.collect())
-                                want_to_cancel = (task.task_id in cancelled_tasks) or (
-                                    TaskId("CANCEL_CURRENT_TASK") in cancelled_tasks
-                                )
-                                if mx_any(want_to_cancel, group):
-                                    break
-
                            match response:
                                case GenerationResponse():
                                    completion_tokens += 1
@@ -414,7 +382,7 @@ def main(
                case ImageGeneration(
                    task_params=task_params, command_id=command_id
                ) if isinstance(current_status, RunnerReady):
-                    assert image_model
+                    assert isinstance(model, DistributedImageModel)
                    logger.info(f"received image generation request: {str(task)[:500]}")
                    current_status = RunnerRunning()
                    logger.info("runner running")
@@ -427,9 +395,7 @@ def main(

                    try:
                        image_index = 0
-                        for response in generate_image(
-                            model=image_model, task=task_params
-                        ):
+                        for response in generate_image(model=model, task=task_params):
                            is_primary_output = _is_primary_output_node(shard_metadata)

                            if is_primary_output:
@@ -479,7 +445,7 @@ def main(
                case ImageEdits(task_params=task_params, command_id=command_id) if (
                    isinstance(current_status, RunnerReady)
                ):
-                    assert image_model
+                    assert isinstance(model, DistributedImageModel)
                    logger.info(f"received image edits request: {str(task)[:500]}")
                    current_status = RunnerRunning()
                    logger.info("runner running")
@@ -492,9 +458,7 @@ def main(

                    try:
                        image_index = 0
-                        for response in generate_image(
-                            model=image_model, task=task_params
-                        ):
+                        for response in generate_image(model=model, task=task_params):
                            if _is_primary_output_node(shard_metadata):
                                match response:
                                    case PartialImageResponse():
@@ -560,7 +524,7 @@ def main(
                RunnerStatusUpdated(runner_id=runner_id, runner_status=current_status)
            )
            if isinstance(current_status, RunnerShutdown):
-                del inference_model, image_model, tokenizer, group
+                del model, tokenizer, group
                mx.clear_cache()
                import gc

--- a/src/exo/worker/runner/runner_supervisor.py
+++ b/src/exo/worker/runner/runner_supervisor.py
@@ -8,10 +8,8 @@ import anyio
 from anyio import (
    BrokenResourceError,
    ClosedResourceError,
-    create_task_group,
    to_thread,
 )
-from anyio.abc import TaskGroup
 from loguru import logger

 from exo.shared.types.events import (
@@ -49,12 +47,9 @@ class RunnerSupervisor:
    _ev_recv: MpReceiver[Event]
    _task_sender: MpSender[Task]
    _event_sender: Sender[Event]
-    _cancel_sender: MpSender[TaskId]
-    _tg: TaskGroup = field(default_factory=create_task_group, init=False)
    status: RunnerStatus = field(default_factory=RunnerIdle, init=False)
    pending: dict[TaskId, anyio.Event] = field(default_factory=dict, init=False)
    completed: set[TaskId] = field(default_factory=set, init=False)
-    cancelled: set[TaskId] = field(default_factory=set, init=False)

    @classmethod
    def create(
@@ -65,8 +60,8 @@ class RunnerSupervisor:
        initialize_timeout: float = 400,
    ) -> Self:
        ev_send, ev_recv = mp_channel[Event]()
+        # A task is kind of a runner command
        task_sender, task_recv = mp_channel[Task]()
-        cancel_sender, cancel_recv = mp_channel[TaskId]()

        runner_process = Process(
            target=entrypoint,
@@ -74,7 +69,6 @@ class RunnerSupervisor:
                bound_instance,
                ev_send,
                task_recv,
-                cancel_recv,
                logger,
            ),
            daemon=True,
@@ -89,7 +83,6 @@ class RunnerSupervisor:
            initialize_timeout=initialize_timeout,
            _ev_recv=ev_recv,
            _task_sender=task_sender,
-            _cancel_sender=cancel_sender,
            _event_sender=event_sender,
        )

@@ -97,42 +90,35 @@ class RunnerSupervisor:

    async def run(self):
        self.runner_process.start()
-        async with self._tg as tg:
-            tg.start_soon(self._forward_events)
+        await self._forward_events()

-        with anyio.CancelScope(shield=True), contextlib.suppress(ClosedResourceError):
-            await self._cancel_sender.send_async(TaskId("CANCEL_CURRENT_TASK"))
+    def shutdown(self):
+        logger.info("Runner supervisor shutting down")
+        self._ev_recv.close()
+        self._task_sender.close()
+        self._event_sender.close()
+        self.runner_process.join(1)
+        if not self.runner_process.is_alive():
+            logger.info("Runner process succesfully terminated")
+            return

-            self._ev_recv.close()
-            self._task_sender.close()
-            self._event_sender.close()
-            self._cancel_sender.close()
+        # This is overkill but it's not technically bad, just unnecessary.
+        logger.warning("Runner process didn't shutdown succesfully, terminating")
+        self.runner_process.terminate()
+        self.runner_process.join(1)
+        if not self.runner_process.is_alive():
+            return

-            await to_thread.run_sync(self.runner_process.join, 10)
-            if not self.runner_process.is_alive():
-                return
+        logger.critical("Runner process didn't respond to SIGTERM, killing")
+        self.runner_process.kill()

-            # This is overkill but it's not technically bad, just unnecessary.
-            logger.warning("Runner process didn't shutdown succesfully, terminating")
-            self.runner_process.terminate()
-            await to_thread.run_sync(self.runner_process.join, 5)
-            if not self.runner_process.is_alive():
-                return
+        self.runner_process.join(1)
+        if not self.runner_process.is_alive():
+            return

-            logger.critical("Runner process didn't respond to SIGTERM, killing")
-            self.runner_process.kill()
-
-            await to_thread.run_sync(self.runner_process.join, 5)
-            if not self.runner_process.is_alive():
-                return
-
-            logger.critical(
-                "Runner process didn't respond to SIGKILL. System resources may have leaked"
-            )
-
-    async def shutdown(self):
-        await self._cancel_sender.send_async(TaskId("CANCEL_CURRENT_TASK"))
-        self._tg.cancel_scope.cancel()
+        logger.critical(
+            "Runner process didn't respond to SIGKILL. System resources may have leaked"
+        )

    async def start_task(self, task: Task):
        if task.task_id in self.pending:
@@ -155,13 +141,6 @@ class RunnerSupervisor:
            return
        await event.wait()

-    async def cancel_task(self, task_id: TaskId):
-        if task_id in self.completed:
-            logger.info(f"Unable to cancel {task_id} as it has been completed")
-            return
-        self.cancelled.add(task_id)
-        await self._cancel_sender.send_async(task_id)
-
    async def _forward_events(self):
        with self._ev_recv as events:
            try:
@@ -226,4 +205,4 @@ class RunnerSupervisor:
                runner_status=RunnerFailed(error_message=f"Terminated ({cause})"),
            )
        )
-        await self.shutdown()
+        self.shutdown()
--- a/src/exo/worker/tests/unittests/test_runner/test_event_ordering.py
+++ b/src/exo/worker/tests/unittests/test_runner/test_event_ordering.py
@@ -2,7 +2,6 @@
 from collections.abc import Iterable
 from typing import Callable

-import mlx.core as mx
 import pytest

 import exo.worker.runner.runner as mlx_runner
@@ -20,7 +19,6 @@ from exo.shared.types.tasks import (
    Shutdown,
    StartWarmup,
    Task,
-    TaskId,
    TaskStatus,
    TextGeneration,
 )
@@ -115,8 +113,6 @@ def patch_out_mlx(monkeypatch: pytest.MonkeyPatch):
    monkeypatch.setattr(mlx_runner, "load_mlx_items", make_nothin((1, MockTokenizer)))
    monkeypatch.setattr(mlx_runner, "warmup_inference", make_nothin(1))
    monkeypatch.setattr(mlx_runner, "_check_for_debug_prompts", nothin)
-    monkeypatch.setattr(mx.distributed, "all_gather", make_nothin(mx.array([1])))
-    monkeypatch.setattr(mlx_runner, "mx_any", make_nothin(False))
    # Mock apply_chat_template since we're using a fake tokenizer (integer 1).
    # Returns a prompt without thinking tag so detect_thinking_prompt_suffix returns None.
    monkeypatch.setattr(mlx_runner, "apply_chat_template", make_nothin("test prompt"))
@@ -167,7 +163,6 @@ def _run(tasks: Iterable[Task]):
    )

    task_sender, task_receiver = mp_channel[Task]()
-    _cancel_sender, cancel_receiver = mp_channel[TaskId]()
    event_sender = EventCollector()

    with task_sender:
@@ -179,7 +174,7 @@ def _run(tasks: Iterable[Task]):
        task_receiver.close = nothin
        task_receiver.join = nothin

-        mlx_runner.main(bound_instance, event_sender, task_receiver, cancel_receiver)  # pyright: ignore[reportArgumentType]
+        mlx_runner.main(bound_instance, event_sender, task_receiver)  # type: ignore[arg-type]

        return event_sender.events

--- a/tests/run_exo_on.sh
+++ b/tests/run_exo_on.sh
@@ -22,7 +22,7 @@ echo "Deploying $commit to $# hosts..."
 hosts=("$@")
 cleanup() {
  for host in "${hosts[@]}"; do
-    ssh -T -o BatchMode=yes "$host@$host" "pkill -SIGINT -of exo-env" &
+    ssh -T -o BatchMode=yes "$host@$host" "pkill -f bin/exo" &
  done
  wait
  jobs -pr | xargs -r kill 2>/dev/null || true
@@ -34,21 +34,13 @@ reset=$'\e[0m'
 i=0
 for host; do
  colour=${colours[i++ % 4]}
-  {
-    ssh -T -o BatchMode=yes -o ServerAliveInterval=30 "$host@$host" \
-      "/nix/var/nix/profiles/default/bin/nix shell nixpkgs#git -c bash -s -- '$commit'" \
-      2>&1 | awk -v p="${colour}[${host}]${reset}" '{ print p $0; fflush() }' &
-  } <<'EOF'
-        set -euo pipefail
-        cd exo
-        git fetch -q origin
-        git checkout -q "$1"
-        EXO_LIBP2P_NAMESPACE="$1" /nix/var/nix/profiles/default/bin/nix run .#exo
-EOF
+  ssh -T -o BatchMode=yes -o ServerAliveInterval=30 "$host@$host" \
+    "/nix/var/nix/profiles/default/bin/nix run github:exo-explore/exo/$commit" |&
+    awk -v p="${colour}[${host}]${reset}" '{ print p $0; fflush() }' &
 done

 for host; do
  echo "Waiting for $host..."
-  until curl -sf "http://$host:52415/models"; do sleep 1; done
+  until curl -sf "http://$host:52415/models" &>/dev/null; do sleep 1; done
 done
 wait