mirror of
https://github.com/mudler/LocalAI.git
synced 2026-04-17 05:18:53 -04:00
6f0051301b8e65d4f86cd5eea5a1d8cfd02cf7dd
6 Commits
| Author | SHA1 | Message | Date | |
|---|---|---|---|---|
Ettore Di Giacinto
|
016da02845 |
feat: refactor shared helpers and enhance MLX backend functionality (#9335)
* refactor(backends): extract python_utils + add mlx_utils shared helpers
Move parse_options() and messages_to_dicts() out of vllm_utils.py into a
new framework-agnostic python_utils.py, and re-export them from vllm_utils
so existing vllm / vllm-omni imports keep working.
Add mlx_utils.py with split_reasoning() and parse_tool_calls() — ported
from mlx_vlm/server.py's process_tool_calls. These work with any
mlx-lm / mlx-vlm tool module (anything exposing tool_call_start,
tool_call_end, parse_tool_call). Used by the mlx and mlx-vlm backends in
later commits to emit structured ChatDelta.tool_calls without
reimplementing per-model parsing.
Shared smoke tests confirm:
- parse_options round-trips bool/int/float/string
- vllm_utils re-exports are identity-equal to python_utils originals
- mlx_utils parse_tool_calls handles <tool_call>...</tool_call> with a
shim module and produces a correctly-indexed list with JSON arguments
- mlx_utils split_reasoning extracts <think> blocks and leaves clean
content
* feat(mlx): wire native tool parsers + ChatDelta + token usage + logprobs
Bring the MLX backend up to the same structured-output contract as vLLM
and llama.cpp: emit Reply.chat_deltas so the OpenAI HTTP layer sees
tool_calls and reasoning_content, not just raw text.
Key insight: mlx_lm.load() returns a TokenizerWrapper that already auto-
detects the right tool parser from the model's chat template
(_infer_tool_parser in mlx_lm/tokenizer_utils.py). The wrapper exposes
has_tool_calling, has_thinking, tool_parser, tool_call_start,
tool_call_end, think_start, think_end — no user configuration needed,
unlike vLLM.
Changes in backend/python/mlx/backend.py:
- Imports: replace inline parse_options / messages_to_dicts with the
shared helpers from python_utils. Pull split_reasoning / parse_tool_calls
from the new mlx_utils shared module.
- LoadModel: log the auto-detected has_tool_calling / has_thinking /
tool_parser_type for observability. Drop the local is_float / is_int
duplicates.
- _prepare_prompt: run request.Messages through messages_to_dicts so
tool_call_id / tool_calls / reasoning_content survive the conversion,
and pass tools=json.loads(request.Tools) + enable_thinking=True (when
request.Metadata says so) to apply_chat_template. Falls back on
TypeError for tokenizers whose template doesn't accept those kwargs.
- _build_generation_params: return an additional (logits_params,
stop_words) pair. Maps RepetitionPenalty / PresencePenalty /
FrequencyPenalty to mlx_lm.sample_utils.make_logits_processors and
threads StopPrompts through to post-decode truncation.
- New _tool_module_from_tokenizer / _finalize_output / _truncate_at_stop
helpers. _finalize_output runs split_reasoning when has_thinking is
true and parse_tool_calls (using a SimpleNamespace shim around the
wrapper's tool_parser callable) when has_tool_calling is true, then
extracts prompt_tokens, generation_tokens and (best-effort) logprobs
from the last GenerationResponse chunk.
- Predict: use make_logits_processors, accumulate text + last_response,
finalize into a structured Reply carrying chat_deltas,
prompt_tokens, tokens, logprobs. Early-stops on user stop sequences.
- PredictStream: per-chunk Reply still carries raw message bytes for
back-compat but now also emits chat_deltas=[ChatDelta(content=delta)].
On loop exit, emit a terminal Reply with structured
reasoning_content / tool_calls / token counts / logprobs — so the Go
side sees tool calls without needing the regex fallback.
- TokenizeString RPC: uses the TokenizerWrapper's encode(); returns
length + tokens or FAILED_PRECONDITION if the model isn't loaded.
- Free RPC: drops model / tokenizer / lru_cache, runs gc.collect(),
calls mx.metal.clear_cache() when available, and best-effort clears
torch.cuda as a belt-and-suspenders.
* feat(mlx-vlm): mirror MLX parity (tool parsers + ChatDelta + samplers)
Same treatment as the MLX backend: emit structured Reply.chat_deltas,
tool_calls, reasoning_content, token counts and logprobs, and extend
sampling parameter coverage beyond the temp/top_p pair the backend
used to handle.
- Imports: drop the inline is_float/is_int helpers, pull parse_options /
messages_to_dicts from python_utils and split_reasoning /
parse_tool_calls from mlx_utils. Also import make_sampler and
make_logits_processors from mlx_lm.sample_utils — mlx-vlm re-uses them.
- LoadModel: use parse_options; call mlx_vlm.tool_parsers._infer_tool_parser
/ load_tool_module to auto-detect a tool module from the processor's
chat_template. Stash think_start / think_end / has_thinking so later
finalisation can split reasoning blocks without duck-typing on each
call. Logs the detected parser type.
- _prepare_prompt: convert proto Messages via messages_to_dicts (so
tool_call_id / tool_calls survive), pass tools=json.loads(request.Tools)
and enable_thinking=True to apply_chat_template when present, fall
back on TypeError for older mlx-vlm versions. Also handle the
prompt-only + media and empty-prompt + media paths consistently.
- _build_generation_params: return (max_tokens, sampler_params,
logits_params, stop_words). Maps repetition_penalty / presence_penalty /
frequency_penalty and passes them through make_logits_processors.
- _finalize_output / _truncate_at_stop: common helper used by Predict
and PredictStream to split reasoning, run parse_tool_calls against the
auto-detected tool module, build ToolCallDelta list, and extract token
counts + logprobs from the last GenerationResult.
- Predict / PredictStream: switch from mlx_vlm.generate to mlx_vlm.stream_generate
in both paths, accumulate text + last_response, pass sampler and
logits_processors through, emit content-only ChatDelta per streaming
chunk followed by a terminal Reply carrying reasoning_content,
tool_calls, prompt_tokens, tokens and logprobs. Non-streaming Predict
returns the same structured Reply shape.
- New helper _collect_media extracted from the duplicated base64 image /
audio decode loop.
- New TokenizeString RPC using the processor's tokenizer.encode and
Free RPC that drops model/processor/config, runs gc + Metal cache
clear + best-effort torch.cuda cache clear.
* feat(importer/mlx): auto-set tool_parser/reasoning_parser on import
Mirror what core/gallery/importers/vllm.go does: after applying the
shared inference defaults, look up the model URI in parser_defaults.json
and append matching tool_parser:/reasoning_parser: entries to Options.
The MLX backends auto-detect tool parsers from the chat template at
runtime so they don't actually consume these options — but surfacing
them in the generated YAML:
- keeps the import experience consistent with vllm
- gives users a single visible place to override
- documents the intended parser for a given model family
* test(mlx): add helper unit tests + TokenizeString/Free + e2e make targets
- backend/python/mlx/test.py: add TestSharedHelpers with server-less
unit tests for parse_options, messages_to_dicts, split_reasoning and
parse_tool_calls (using a SimpleNamespace shim to fake a tool module
without requiring a model). Plus test_tokenize_string and test_free
RPC tests that load a tiny MLX-quantized Llama and exercise the new
RPCs end-to-end.
- backend/python/mlx-vlm/test.py: same helper unit tests + cleanup of
the duplicated import block at the top of the file.
- Makefile: register BACKEND_MLX and BACKEND_MLX_VLM (they were missing
from the docker-build-target eval list — only mlx-distributed had a
generated target before). Add test-extra-backend-mlx and
test-extra-backend-mlx-vlm convenience targets that build the
respective image and run tests/e2e-backends with the tools capability
against mlx-community/Qwen2.5-0.5B-Instruct-4bit. The MLX backend
auto-detects the tool parser from the chat template so no
BACKEND_TEST_OPTIONS is needed (unlike vllm).
* fix(libbackend): don't pass --copies to venv unless PORTABLE_PYTHON=true
backend/python/common/libbackend.sh:ensureVenv() always invoked
'python -m venv --copies', but macOS system python (and some other
builds) refuses with:
Error: This build of python cannot create venvs without using symlinks
--copies only matters when _makeVenvPortable later relocates the venv,
which only happens when PORTABLE_PYTHON=true. Make --copies conditional
on that flag and fall back to default (symlinked) venv otherwise.
Caught while bringing up the mlx backend on Apple Silicon — the same
build path is used by every Python backend with USE_PIP=true.
* fix(mlx): support mlx-lm 0.29.x tool calling + drop deprecated clear_cache
The released mlx-lm 0.29.x ships a much simpler tool-calling API than
HEAD: TokenizerWrapper detects the <tool_call>...</tool_call> markers
from the tokenizer vocab and exposes has_tool_calling /
tool_call_start / tool_call_end, but does NOT expose a tool_parser
callable on the wrapper and does NOT ship a mlx_lm.tool_parsers
subpackage at all (those only exist on main).
Caught while running the smoke test on Apple Silicon with the
released mlx-lm 0.29.1: tokenizer.tool_parser raised AttributeError
(falling through to the underlying HF tokenizer), so
_tool_module_from_tokenizer always returned None and tool calls slipped
through as raw <tool_call>...</tool_call> text in Reply.message instead
of being parsed into ChatDelta.tool_calls.
Fix: when has_tool_calling is True but tokenizer.tool_parser is missing,
default the parse_tool_call callable to json.loads(body.strip()) — that's
exactly what mlx_lm.tool_parsers.json_tools.parse_tool_call does on HEAD
and covers the only format 0.29 detects (<tool_call>JSON</tool_call>).
Future mlx-lm releases that ship more parsers will be picked up
automatically via the tokenizer.tool_parser attribute when present.
Also tighten the LoadModel logging — the old log line read
init_kwargs.get('tool_parser_type') which doesn't exist on 0.29 and
showed None even when has_tool_calling was True. Log the actual
tool_call_start / tool_call_end markers instead.
While here, switch Free()'s Metal cache clear from the deprecated
mx.metal.clear_cache to mx.clear_cache (mlx >= 0.30), with a
fallback for older releases. Mirrored to the mlx-vlm backend.
* feat(mlx-distributed): mirror MLX parity (tool calls + ChatDelta + sampler)
Same treatment as the mlx and mlx-vlm backends: emit Reply.chat_deltas
with structured tool_calls / reasoning_content / token counts /
logprobs, expand sampling parameter coverage beyond temp+top_p, and
add the missing TokenizeString and Free RPCs.
Notes specific to mlx-distributed:
- Rank 0 is the only rank that owns a sampler — workers participate in
the pipeline-parallel forward pass via mx.distributed and don't
re-implement sampling. So the new logits_params (repetition_penalty,
presence_penalty, frequency_penalty) and stop_words apply on rank 0
only; we don't need to extend coordinator.broadcast_generation_params,
which still ships only max_tokens / temperature / top_p to workers
(everything else is a rank-0 concern).
- Free() now broadcasts CMD_SHUTDOWN to workers when a coordinator is
active, so they release the model on their end too. The constant is
already defined and handled by the existing worker loop in
backend.py:633 (CMD_SHUTDOWN = -1).
- Drop the locally-defined is_float / is_int / parse_options trio in
favor of python_utils.parse_options, re-exported under the module
name for back-compat with anything that imported it directly.
- _prepare_prompt: route through messages_to_dicts so tool_call_id /
tool_calls / reasoning_content survive, pass tools=json.loads(
request.Tools) and enable_thinking=True to apply_chat_template, fall
back on TypeError for templates that don't accept those kwargs.
- New _tool_module_from_tokenizer (with the json.loads fallback for
mlx-lm 0.29.x), _finalize_output, _truncate_at_stop helpers — same
contract as the mlx backend.
- LoadModel logs the auto-detected has_tool_calling / has_thinking /
tool_call_start / tool_call_end so users can see what the wrapper
picked up for the loaded model.
- backend/python/mlx-distributed/test.py: add the same TestSharedHelpers
unit tests (parse_options, messages_to_dicts, split_reasoning,
parse_tool_calls) that exist for mlx and mlx-vlm.
|
||
|
Ettore Di Giacinto
|
59108fbe32 |
feat: add distributed mode (#9124)
* feat: add distributed mode (experimental) Signed-off-by: Ettore Di Giacinto <mudler@localai.io> * fix data races, mutexes, transactions Signed-off-by: Ettore Di Giacinto <mudler@localai.io> * refactorings Signed-off-by: Ettore Di Giacinto <mudler@localai.io> * fixups Signed-off-by: Ettore Di Giacinto <mudler@localai.io> * fix events and tool stream in agent chat Signed-off-by: Ettore Di Giacinto <mudler@localai.io> * use ginkgo Signed-off-by: Ettore Di Giacinto <mudler@localai.io> * refactoring and consolidation Signed-off-by: Ettore Di Giacinto <mudler@localai.io> * refactoring and consolidation Signed-off-by: Ettore Di Giacinto <mudler@localai.io> * refactoring and consolidation Signed-off-by: Ettore Di Giacinto <mudler@localai.io> * refactoring and consolidation Signed-off-by: Ettore Di Giacinto <mudler@localai.io> * refactoring and consolidation Signed-off-by: Ettore Di Giacinto <mudler@localai.io> * refactoring and consolidation Signed-off-by: Ettore Di Giacinto <mudler@localai.io> * refactoring and consolidation Signed-off-by: Ettore Di Giacinto <mudler@localai.io> * refactoring and consolidation Signed-off-by: Ettore Di Giacinto <mudler@localai.io> * fix(cron): compute correctly time boundaries avoiding re-triggering Signed-off-by: Ettore Di Giacinto <mudler@localai.io> * enhancements, refactorings Signed-off-by: Ettore Di Giacinto <mudler@localai.io> * do not flood of healthy checks Signed-off-by: Ettore Di Giacinto <mudler@localai.io> * do not list obvious backends as text backends Signed-off-by: Ettore Di Giacinto <mudler@localai.io> * tests fixups Signed-off-by: Ettore Di Giacinto <mudler@localai.io> * refactoring and consolidation Signed-off-by: Ettore Di Giacinto <mudler@localai.io> * Drop redundant healthcheck Signed-off-by: Ettore Di Giacinto <mudler@localai.io> * enhancements, refactorings Signed-off-by: Ettore Di Giacinto <mudler@localai.io> --------- Signed-off-by: Ettore Di Giacinto <mudler@localai.io> |
||
|
blightbow
|
67baf66555 |
feat(mlx): add thread-safe LRU prompt cache and min_p/top_k sampling (#7556)
* feat(mlx): add thread-safe LRU prompt cache Port mlx-lm's LRUPromptCache to fix race condition where concurrent requests corrupt shared KV cache state. The previous implementation used a single prompt_cache instance shared across all requests. Changes: - Add backend/python/common/mlx_cache.py with ThreadSafeLRUPromptCache - Modify backend.py to use per-request cache isolation via fetch/insert - Add prefix matching for cache reuse across similar prompts - Add LRU eviction (default 10 entries, configurable) - Add concurrency and cache unit tests The cache uses a trie-based structure for efficient prefix matching, allowing prompts that share common prefixes to reuse cached KV states. Thread safety is provided via threading.Lock. New configuration options: - max_cache_entries: Maximum LRU cache entries (default: 10) - max_kv_size: Maximum KV cache size per entry (default: None) 🤖 Generated with [Claude Code](https://claude.com/claude-code) Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com> Signed-off-by: Blightbow <blightbow@users.noreply.github.com> * feat(mlx): add min_p and top_k sampler support Add MinP field to proto (field 52) following the precedent set by other non-OpenAI sampling parameters like TopK, TailFreeSamplingZ, TypicalP, and Mirostat. Changes: - backend.proto: Add float MinP field for min-p sampling - backend.py: Extract and pass min_p and top_k to mlx_lm sampler (top_k was in proto but not being passed) - test.py: Fix test_sampling_params to use valid proto fields and switch to MLX-compatible model (mlx-community/Llama-3.2-1B-Instruct) 🤖 Generated with [Claude Code](https://claude.com/claude-code) Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com> Signed-off-by: Blightbow <blightbow@users.noreply.github.com> * refactor(mlx): move mlx_cache.py from common to mlx backend The ThreadSafeLRUPromptCache is only used by the mlx backend. After evaluating mlx-vlm, it was determined that the cache cannot be shared because mlx-vlm's generate/stream_generate functions don't support the prompt_cache parameter that mlx_lm provides. - Move mlx_cache.py from backend/python/common/ to backend/python/mlx/ - Remove sys.path manipulation from backend.py and test.py - Fix test assertion to expect "MLX model loaded successfully" 🤖 Generated with [Claude Code](https://claude.com/claude-code) Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com> Signed-off-by: Blightbow <blightbow@users.noreply.github.com> * test(mlx): add comprehensive cache tests and document upstream behavior Added comprehensive unit tests (test_mlx_cache.py) covering all cache operation modes: - Exact match - Shorter prefix match - Longer prefix match with trimming - No match scenarios - LRU eviction and access order - Reference counting and deep copy behavior - Multi-model namespacing - Thread safety with data integrity verification Documents upstream mlx_lm/server.py behavior: single-token prefixes are deliberately not matched (uses > 0, not >= 0) to allow longer cached sequences to be preferred for trimming. This is acceptable because real prompts with chat templates are always many tokens. Removed weak unit tests from test.py that only verified "no exception thrown" rather than correctness. 🤖 Generated with [Claude Code](https://claude.com/claude-code) Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com> Signed-off-by: Blightbow <blightbow@users.noreply.github.com> * chore(mlx): remove unused MinP proto field The MinP field was added to PredictOptions but is not populated by the Go frontend/API. The MLX backend uses getattr with a default value, so it works without the proto field. 🤖 Generated with [Claude Code](https://claude.com/claude-code) Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com> Signed-off-by: Blightbow <blightbow@users.noreply.github.com> --------- Signed-off-by: Blightbow <blightbow@users.noreply.github.com> Co-authored-by: Blightbow <blightbow@users.noreply.github.com> Co-authored-by: Claude Opus 4.5 <noreply@anthropic.com> |
||
|
Ettore Di Giacinto
|
660bd45be8 |
fix(python): make option check uniform across backends (#6314)
Signed-off-by: Ettore Di Giacinto <mudler@localai.io> |
||
|
Ettore Di Giacinto
|
9621edb4c5 |
feat(diffusers): add support for wan2.2 (#6153)
* feat(diffusers): add support for wan2.2
Signed-off-by: Ettore Di Giacinto <mudler@localai.io>
* chore(ci): use ttl.sh for PRs
Signed-off-by: Ettore Di Giacinto <mudler@localai.io>
* Add ftfy deps
Signed-off-by: Ettore Di Giacinto <mudler@localai.io>
* Revert "chore(ci): use ttl.sh for PRs"
This reverts commit
|
||
|
Ettore Di Giacinto
|
1d830ce7dd |
feat(mlx): add mlx backend (#6049)
* chore: allow to install with pip Signed-off-by: Ettore Di Giacinto <mudler@localai.io> * WIP Signed-off-by: Ettore Di Giacinto <mudler@localai.io> * Make the backend to build and actually work Signed-off-by: Ettore Di Giacinto <mudler@localai.io> * List models from system only Signed-off-by: Ettore Di Giacinto <mudler@localai.io> * Add script to build darwin python backends Signed-off-by: Ettore Di Giacinto <mudler@localai.io> * Run protogen in libbackend Signed-off-by: Ettore Di Giacinto <mudler@localai.io> * Detect if mps is available across python backends Signed-off-by: Ettore Di Giacinto <mudler@localai.io> * CI: try to build backend Signed-off-by: Ettore Di Giacinto <mudler@localai.io> * Debug CI Signed-off-by: Ettore Di Giacinto <mudler@localai.io> * Fixups Signed-off-by: Ettore Di Giacinto <mudler@localai.io> * Fixups Signed-off-by: Ettore Di Giacinto <mudler@localai.io> * Index mlx-vlm Signed-off-by: Ettore Di Giacinto <mudler@localai.io> * Remove mlx-vlm Signed-off-by: Ettore Di Giacinto <mudler@localai.io> * Drop CI test Signed-off-by: Ettore Di Giacinto <mudler@localai.io> --------- Signed-off-by: Ettore Di Giacinto <mudler@localai.io> |