mirror of
https://github.com/mudler/LocalAI.git
synced 2026-07-31 10:28:43 -04:00
feat/vllm-cpp-darwin-mlx
27 Commits
| Author | SHA1 | Message | Date | |
|---|---|---|---|---|
Richard Palethorpe
|
49ef40a187 |
feat(classifier/VAD): support voice control on low power devices (#10804)
* feat(llama-cpp): route Score through the slot loop Score previously bypassed the slot loop with a direct llama_decode: a conflict guard aborted the whole process if scoring raced generation, the config validator had to reject score alongside chat/completion/embeddings, and every candidate re-decoded the full shared prompt. Add SERVER_TASK_TYPE_SCORE to the (patched) upstream server so score tasks are scheduled like any other slot work: generation and scoring serialize naturally, the shared prompt is decoded once per call, and the slot's prompt cache carries the conversation prefix across calls. Context checkpoints at the score boundary and at the cache-divergence point keep SWA/hybrid/recurrent models (e.g. LFM2.5) from re-prefilling the whole prompt per candidate: warm-turn scoring on a 6-option set drops from ~8s to ~0.5s on a desktop CPU. The conflict guard and the validation split are removed; declaring score with generation usecases on one config is now supported and shares the slot cache. Assisted-by: Claude:claude-fable-5 Signed-off-by: Richard Palethorpe <io@richiejp.com> * feat(realtime): classifier wire types and pipeline config Wire types and YAML config for realtime classifier mode: sessions carry a localai_classifier extension (options with canned replies/tool calls, softmax threshold, normalization, history trimming, fallback modes, and a deterministic wake-word address gate), mirrored by pipeline.classifier in the model YAML and surfaced in the config-meta registry. The localai.classifier.result server event reports the full score distribution per turn. Assisted-by: Claude:claude-fable-5 Signed-off-by: Richard Palethorpe <io@richiejp.com> * feat(realtime): classifier response flow Classifier-mode responses: instead of autoregressive generation, each user turn is prefill-scored against the option list (router.ScoreClassifier prompt/candidate shapes over the Score primitive) and the winning option's canned reply and tool call are emitted through the existing response machinery. Below-threshold turns take the configured fallback (none / canned reply / generate); empty transcripts and unaddressed turns (wake word not mentioned) skip scoring entirely. The scoring probe defaults to the latest user message only — small scorers echo canned replies from prior turns back as the top option otherwise. Built for hardware that can afford prompt processing but not decode: with slot-based Score the option list stays KV-cached across turns, so a turn costs roughly one forward pass over the new words. session_update_error events now carry the validation cause instead of a generic message. Assisted-by: Claude:claude-fable-5 Signed-off-by: Richard Palethorpe <io@richiejp.com> * fix(realtime): bound the VAD tick's scan window and buffer retention The VAD tick loop re-scanned the entire input buffer every 300ms and only trimmed it on zero-segment ticks or commits. Audio that keeps producing segments without a committing pause (steady noise a mic pipeline lets through, music, continuous speech) grew the buffer toward the 100MB cap with each tick rescanning all of it — O(n^2), measured at ~3.3ms of silero per buffered second: past ~90s retained, ticks run back to back and pin ~4 cores until the stream stops. Silero's recurrent state only carries a few hundred ms of context, so rescanning old audio buys nothing. Clip the slice handed to the VAD to the largest silence the commit test can need to measure (server_vad silence window or the semantic eagerness fallback) plus a warm-up margin, and rebase the returned segment times so every downstream consumer keeps whole-buffer coordinates. An open turn whose clipped window is all silence now commits (the silence outran the window) instead of being discarded as no-speech. Independently, retain at most 90s of raw buffer, rebasing the live-feed and EOU cursors on trim — this also bounds the previously unbounded VAD-error path. Turn boundaries are otherwise unchanged: no forced commits, no new coordinator states. pipeline.turn_detection.vad_window_sec can widen the scan window; values below the automatic floor are ignored. The tick body is extracted into vadTick so specs can drive turn detection synchronously (same shape as classifySoundWindow); the babble reproduction that pinned 4 cores now plateaus under 10% of one core. Assisted-by: Claude:claude-fable-5 Signed-off-by: Richard Palethorpe <io@richiejp.com> * fix(backend): let per-model threads override the global default ModelOptions overrode a set per-model threads value with the app-level --threads whenever the latter was non-zero — and WithThreads defaults it to the physical core count, so it always was. The YAML threads: knob has been dead config: a tiny VAD model could never opt down from the global pool size. SetDefaults already fills an unset per-model value from the app config, which is the intended precedence; resolve threads through a helper that honors it (explicit threads: 0 still means unset). Assisted-by: Claude:claude-fable-5 Signed-off-by: Richard Palethorpe <io@richiejp.com> * chore(gallery): single-thread the silero VAD Silero is a ~2MB recurrent model with no exploitable graph parallelism: measured per-call latency is identical at 1 and 10 ORT threads, while every extra pool thread just spin-waits between the realtime loop's frequent tiny inferences. Assisted-by: Claude:claude-fable-5 Signed-off-by: Richard Palethorpe <io@richiejp.com> * docs(realtime): classifier mode, VAD scan window, threads precedence Document the realtime classifier mode (options, threshold guidance, wake-word address gate, empty-transcript handling), the VAD scan window and 90s buffer retention (pipeline.turn_detection.vad_window_sec), the per-model threads precedence, and the M3 classifier note in the realtime state-machine design doc. Assisted-by: Claude:claude-fable-5 Signed-off-by: Richard Palethorpe <io@richiejp.com> * perf(llama-cpp): score all candidates in one batched decode One scoring call is now a single SERVER_TASK_TYPE_SCORE task: the slot decodes the shared prefix (prompt + longest common candidate token prefix) once, then forks one sequence per candidate off it (metadata-only for the unified KV cache, copy-on-write for recurrent state) and decodes every candidate's unique tail in one llama_decode. Previously each candidate was its own task that restored the boundary checkpoint and re-decoded its full tail sequentially, paying per-candidate task and decode overhead. The context reserves SERVER_SCORE_FORK_SEQS extra sequence ids (and recurrent-state cells) beyond the parallel slots via the new common_params::n_seq_score_forks. Forking requires the unified KV cache (already this backend's default) since per-sequence streams would shrink n_ctx_seq; an explicit kv_unified:false disables forking and Score calls that need it fail cleanly. Candidates beyond the fork/output budget decode in successive chunks. Wire contract and scores are unchanged: per-token logprobs are stitched from the shared region and the forked tails. Verified bitwise deterministic call-to-call and independent of candidate order (no cross-fork leakage via equal-length candidate swap); ranking matches the per-candidate implementation on the drone battery (winner softmax 0.99996 vs 0.99997), and >16-candidate chunking, prefix-of-another and empty candidates all pass. Measured on a desktop CPU: warm /api/score calls 0.52s -> 0.23s; warm realtime classifier turns 196-303ms. The 9-candidate drone turn decodes ~17 unique tail tokens in one batch instead of nine sequential ~220ms checkpoint-restore tasks. Assisted-by: Claude:claude-fable-5 Signed-off-by: Richard Palethorpe <io@richiejp.com> * fix(realtime): gate scoring capacity by model usecase Reserve llama.cpp scoring slots only for models that explicitly declare the score usecase, while allowing score to coexist with chat and completion. Reject incompatible unified-KV settings and classifier activation on models without scoring capacity. Propagate application defaults when resolving realtime and preload pipeline stages so unset thread counts are resolved consistently without overriding explicit model settings. Assisted-by: Codex:gpt-5 Signed-off-by: Richard Palethorpe <io@richiejp.com> * fix(ci): honor APT mirrors in the prebuilt llama-cpp compile step The builder-prebuilt path installs gcc-14 with apt directly and ignored the APT_MIRROR/APT_PORTS_MIRROR build args the from-source path already honors, so an ubuntu mirror outage broke every arm64 backend build. Pass the args into the stage and run apt-mirror.sh (already in the build context via COPY . /LocalAI) before the apt step. Assisted-by: Claude:claude-fable-5 Signed-off-by: Richard Palethorpe <io@richiejp.com> * feat(realtime): classifier argument slots via constrained completion Hybrid classify-then-complete: a classifier option's canned tool call can declare typed argument slots (number | enum | string, with defaults and prompt hints) referenced as "{{name}}" in the arguments template. When the option wins, the slots are filled by a short grammar-constrained completion that continues the exact scoring prompt — rendered by the same cached ScoreClassifier, so the llama.cpp prompt cache is already warm — with the chosen route JSON re-opened at the first slot field. A GBNF grammar pins the field skeleton and frees only the values; temperature 0, a couple dozen tokens at most (~300ms on a desktop CPU for two slots). Slot declarations and hints ride the option descriptions in the shared system prompt, informing scoring and the fill alike at no per-turn token cost. The localai.classifier.result event carries the final arguments and a fill_latency_ms. On inference failure the slots' defaults apply; a slot without a default fails the response (or falls through with fallback.mode: generate). Slot filling requires completion alongside score in the scoring model's known_usecases. Verified end-to-end on the Pi drone demo: "fly forward three meters" in distance mode classifies forward and infers {"distance": 3, "units": "meters"} in ~310ms, and the drone flies exactly 3 units. Assisted-by: Claude:claude-fable-5 Signed-off-by: Richard Palethorpe <io@richiejp.com> * feat(realtime): splice filled slot values into classifier replies A classifier option's spoken reply can now reference its tool's argument slots ("Going forward {{distance}} {{units}}."): the values inferred by the slot-fill completion — or the recovery defaults — are spliced into the reply as plain text before it is emitted, so what the assistant says confirms what it actually inferred. Placeholders without a value stay literal, and options without slots are untouched. FillToolArguments now returns the raw slot values alongside the spliced arguments JSON to make the reply templating possible. Assisted-by: Claude:claude-fable-5 Signed-off-by: Richard Palethorpe <io@richiejp.com> * fix(realtime): harden classifier slot completion Reserve context for constrained slot filling, size completions from their encoded output, and encode enum grammar literals as valid JSON. Reject empty enum values and cover the failure modes with regression tests. Assisted-by: Codex:gpt-5 Signed-off-by: Richard Palethorpe <io@richiejp.com> * feat(realtime): prewarm the classifier scoring prompt on registration Swapping a session's classifier option list (a voice-switched command mode, for instance) made the next turns pay a full re-prefill of the new option-list prompt — measured 2.4s vs 0.3s warm on a desktop CPU, and worse: on hybrid-memory models like LFM2.5, whose state cannot be partially rewound (llama.cpp can only restore checkpoints), *every* probe change re-prefilled from scratch whenever the last checkpoint missed the probe boundary, so even same-list turns intermittently cost full prefills. Registering an option list (pipeline seed or session.update) now fires a best-effort background prewarm: two throwaway scores with distinct probes. The first prefills the new option-list prompt; the second, diverging exactly where per-turn probe text starts, plants the backend's rewind point (KV checkpoint) at the stable-prefix boundary that every real turn reuses. The prewarm hides behind the canned mode-switch reply — by the time it finishes speaking, the cache is warm. Idempotent per option set, detached from the registering request's lifetime. Measured on the drone demo (LFM2.5-1.2B, desktop CPU): first turn after a mode switch 2374ms -> 340ms; intermittent same-list full prefills (1.3-2.1s) all -> under 0.5s. For clients that swap lists frequently, options: [parallel:2] on the scoring model additionally keeps one slot per list via prefix-similarity routing (+26MB RSS, unified KV). Assisted-by: Claude:claude-fable-5 Signed-off-by: Richard Palethorpe <io@richiejp.com> * perf(llama-cpp): checkpoint scoring at the caller-declared stable prefix Hybrid-memory models (LFM2.5 shortconv, Qwen3.5 deltanet — where new small models are headed) cannot rewind their state, so any prompt-cache reuse that needs a rewind falls back to a full re-prefill. For classifier scoring that meant every probe change re-processed the whole option-list prompt: the server's checkpoints were placed reactively (at wherever the previous task happened to diverge), so a checkpoint past the next divergence was erased rather than restored — measured as intermittent 2-10s turns on prompts with a 95%+ common prefix. The classifier now computes the probe-invariant prompt prefix once (the byte-wise common prefix of two synthetic probe renders) and declares its length with every Score request; the server maps it to a token boundary and forces a KV checkpoint exactly there on each score prefill. That checkpoint sits at or before every future divergence under the same option list, so it always survives and always restores — repeat scoring costs probe+candidates regardless of how the probe changes. Also: - prewarm reruns on every option-list registration instead of memoizing per list: with boundary checkpoints a redundant rewarm costs two probe-sized decodes, while skipping one after a slot eviction (three lists sharing fewer slots evict in LRU cascades) silently moves a full re-prefill onto the user's next turn - new llama.cpp backend option rs_seq:N exposes bounded recurrent-state rollback outside speculative decoding; measured impractical for deltanet-scale states (65GB for 64 snapshots on Qwen3.5-4B) but cheap insurance for small-state models - docs: the multi-list recipe (parallel:N + sps:0.5 — the default slot similarity threshold funnels distinct lists onto one slot) Measured on the drone demo (LFM2.5-1.2B scorer, desktop CPU), steady state: every turn 285-421ms including mode switches, vs 2.4s post-switch and intermittent 1.3-2.9s re-prefills before. Assisted-by: Claude:claude-fable-5 Signed-off-by: Richard Palethorpe <io@richiejp.com> * fix(realtime): align classifier cache guidance Document the single-score prewarm behavior and clean the vendored score patch formatting. Assisted-by: Codex:gpt-5 Signed-off-by: Richard Palethorpe <io@richiejp.com> * fix(llama-cpp): guard score task for fork backends TurboQuant and Bonsai reuse the primary gRPC server against llama.cpp forks that do not carry LocalAI's slot-based Score patches. Compile the Score integration only for the patched primary backend and return UNIMPLEMENTED from fork builds instead of referencing absent task types and common_params fields. Assisted-by: Codex:gpt-5 [gh] Signed-off-by: Richard Palethorpe <io@richiejp.com> * fix(dev): generate gRPC code before commit lint The coverage phase regenerates ignored protobuf bindings, but lint runs first and can fail against missing or stale output. Generate the pinned bindings before lint so the gate always type-checks the current schema. Assisted-by: Codex:gpt-5 Signed-off-by: Richard Palethorpe <io@richiejp.com> --------- Signed-off-by: Richard Palethorpe <io@richiejp.com> |
||
|
mudler's LocalAI [bot]
|
d7e04dcc32 |
fix(openresponses): make responses visible and cancellable across replicas (#11000)
In distributed mode the Open Responses store is process-local: a sync.OnceValue over a map behind an RWMutex. With several frontend replicas behind a round-robin load balancer, every request that lands on a replica other than the creator misses. Measured on a live 2-replica cluster (#10993): the same response id returns 200 on the creating replica and 404 on its peer, and a cancel on the peer returns 404 without ever invoking CancelFunc, so generation runs to completion on the other replica while the caller is told the response does not exist. previous_response_id chaining fails through the same lookup. Split the state by what can actually cross a process boundary: - Replicated: response metadata (request, response resource, owner, expiry, stream/background flags) via syncstate.SyncedMap, the same component finetune, quantization and agent tasks already use. A local miss in Get/FindItem now falls back to it and returns a read-only remote view, so polling and chaining resolve on any replica. - Delegated: cancellation. context.CancelFunc is a function pointer and exists only in the creating process, so a cancel that lands elsewhere is broadcast on responses.<id>.cancel and applied by whichever replica holds the function. The broadcast is fire-and-forget rather than request/reply: if the owner crashed or was scaled down nobody answers, and the handler must not block on a reply that will never come. The replicated status moves to cancelled either way, which is truthful, since a dead owner's generation died with its process. - Refused: streaming resume. The resume buffer is a byte log plus a live notification channel and cannot be replicated without shipping every token over the bus. A resume that reaches the wrong replica now returns HTTP 409 naming the owning replica via the new ErrResponseNotLocal, instead of an empty event list that looks like a finished stream. It is deliberately distinct from ErrOffsetLost, which means the owner's buffer evicted the requested events. Standalone deployments never call EnableDistributed and keep exactly the previous process-local behaviour. Fixes #10993 Assisted-by: Claude Code:claude-opus-4-8[1m] [Read] [Edit] [Bash] Signed-off-by: Ettore Di Giacinto <mudler@localai.io> Co-authored-by: Ettore Di Giacinto <mudler@localai.io> |
||
|
mudler's LocalAI [bot]
|
40d35c0385 |
docs: onboarding overhaul, dedup, and error docs (#7711) (#10895)
* docs: fix CPU image tag (latest, not latest-cpu) Signed-off-by: Ettore Di Giacinto <mudler@localai.io> * docs: use canonical localai/localai registry in models guide Signed-off-by: Ettore Di Giacinto <mudler@localai.io> * docs: replace dead llama-stable backend with llama-cpp Signed-off-by: Ettore Di Giacinto <mudler@localai.io> * docs: correct mitm-proxy intercept config and redaction tier Signed-off-by: Ettore Di Giacinto <mudler@localai.io> * docs: fix text-to-audio endpoint and broken notice block Signed-off-by: Ettore Di Giacinto <mudler@localai.io> * docs: fix VAD example, stale FAQ, broken link, CLI list, whats-new dump Signed-off-by: Ettore Di Giacinto <mudler@localai.io> * docs: render advanced/reference section indexes (consolidate _index) Signed-off-by: Ettore Di Giacinto <mudler@localai.io> * docs: remove duplicate getting-started build/kubernetes pages Signed-off-by: Ettore Di Giacinto <mudler@localai.io> * docs: fold container image reference into installation/containers Signed-off-by: Ettore Di Giacinto <mudler@localai.io> * docs: remove stale advanced fine-tuning page (superseded by features/fine-tuning) Signed-off-by: Ettore Di Giacinto <mudler@localai.io> * docs: fold distribution/longcat/sound pages into their parents Signed-off-by: Ettore Di Giacinto <mudler@localai.io> * docs: make getting-started index accurate and complete Signed-off-by: Ettore Di Giacinto <mudler@localai.io> * docs: carry one concrete model through the getting-started path Signed-off-by: Ettore Di Giacinto <mudler@localai.io> * docs: add end-to-end 'build your first agent' walkthrough Signed-off-by: Ettore Di Giacinto <mudler@localai.io> * docs: add runtime errors reference; consolidate troubleshooting from FAQ Signed-off-by: Ettore Di Giacinto <mudler@localai.io> * docs: add agent actions catalog Signed-off-by: Ettore Di Giacinto <mudler@localai.io> * docs: agent-scoped MCP, skills walkthrough, agentic disambiguation Signed-off-by: Ettore Di Giacinto <mudler@localai.io> * docs: add concrete gallery install lines to media feature pages Signed-off-by: Ettore Di Giacinto <mudler@localai.io> * docs: merge installation into getting-started (URLs preserved via aliases) Signed-off-by: Ettore Di Giacinto <mudler@localai.io> * docs: add Operations section; move operator pages and P2P API reference Signed-off-by: Ettore Di Giacinto <mudler@localai.io> * docs: journey-ordered top nav and grouped feature sections Signed-off-by: Ettore Di Giacinto <mudler@localai.io> * docs: add docs-with-code process gate (PR template + agent instructions) Signed-off-by: Ettore Di Giacinto <mudler@localai.io> * docs: remove em/en dashes from documentation prose Signed-off-by: Ettore Di Giacinto <mudler@localai.io> --------- Signed-off-by: Ettore Di Giacinto <mudler@localai.io> Co-authored-by: Ettore Di Giacinto <mudler@localai.io> |
||
|
LocalAI [bot]
|
40dae953f4 |
feat: interleaved thinking with tool calls (reasoning_content alias + Anthropic thinking blocks) (#10744)
* feat(schema): accept reasoning_content as inbound alias for reasoning Interleaved-thinking clients (cogito, vLLM/DeepSeek-style) emit reasoning_content on assistant turns. Accept it as an inbound alias so reasoning survives the tool-result loop; canonical reasoning wins when both are present. Emission is unchanged (still reasoning). Signed-off-by: Ettore Di Giacinto <mudler@localai.io> * test(schema): pin interleaved reasoning+tool_calls round-trip Signed-off-by: Ettore Di Giacinto <mudler@localai.io> * test(openai): pin reachedTokenBudget truncation detection Signed-off-by: Ettore Di Giacinto <mudler@localai.io> * feat(anthropic): add thinking and signature fields to content blocks Signed-off-by: Ettore Di Giacinto <mudler@localai.io> * feat(anthropic): parse inbound thinking blocks into reasoning Signed-off-by: Ettore Di Giacinto <mudler@localai.io> * feat(anthropic): emit thinking blocks with synthetic signature on tool turns Extract buildAnthropicContentBlocks so non-streaming content assembly is unit-testable, and prepend a thinking block (with an opaque synthetic signature) before text/tool_use blocks when the request opts into thinking. Signed-off-by: Ettore Di Giacinto <mudler@localai.io> * feat(anthropic): stream thinking_delta and signature_delta before tool_use Extract anthropicStreamSequence so the streaming block order is unit-testable, and emit content_block_start(thinking) -> thinking_delta -> signature_delta -> content_block_stop before the tool_use block sequence when thinking is enabled. Signed-off-by: Ettore Di Giacinto <mudler@localai.io> * docs: add interleaved thinking with tool calls guide Add a features guide describing interleaved thinking: an assistant turn carrying reasoning and tool_calls together, the reasoning-round-trip contract (including the reasoning_content inbound alias and Anthropic thinking blocks with a synthetic signature), per-backend enablement (reasoning_format for llama.cpp, reasoning_parser/tool_call_parser for vLLM/SGLang plus the vLLM auto-config hook), a worked request/response example, and known limitations. Cross-link from model-configuration, text-generation, and openai-functions. Signed-off-by: Ettore Di Giacinto <mudler@localai.io> --------- Signed-off-by: Ettore Di Giacinto <mudler@localai.io> Co-authored-by: Ettore Di Giacinto <mudler@localai.io> |
||
|
Roman Mazurenko
|
a3fdfbc0d1 |
feat(llama-cpp): add device selection option (#10724)
Allow llama.cpp model configs to select the backend devices used for offload, matching upstream --device behavior so users can exclude a display or debug GPU. Signed-off-by: rvmzes <rvmzes@rvmzess-MacBook-Pro.local> Co-authored-by: rvmzes <rvmzes@rvmzess-MacBook-Pro.local> |
||
|
Ettore Di Giacinto
|
1152acc167 |
Revert "feat(config): default swa_full:true for sliding-window-attention models" (#10674)
Revert "feat(config): default swa_full:true for sliding-window-attention mode…"
This reverts commit
|
||
|
LocalAI [bot]
|
02b007a31e |
feat(config): default swa_full:true for sliding-window-attention models (#10611)
LocalAI enables a cross-request prompt-prefix cache (cache_reuse, see core/config/serving_defaults.go) so repeated prefixes — system prompts, RAG context, agent scaffolds, multi-turn chat — are not reprocessed every turn. For sliding-window-attention (SWA) models (Gemma 2/3, Cohere2, Llama 4, ...) this silently does nothing: llama.cpp defaults to a reduced SWA KV cache sized to the sliding window, and that reduced cache cannot preserve a prompt prefix across requests, so every turn reprocesses the whole prompt anyway. llama.cpp's --swa-full (params.swa_full, already wired through the LocalAI llama.cpp backend's `swa_full` option) keeps the full KV cache so the shared prefix is reused. Enable it automatically, but only for models that are actually SWA: detection reads the gguf-parser-normalized `<arch>.attention.sliding_window` metadata (which also applies llama.cpp's family rules, e.g. Phi-3 → not SWA), right where the GGUF is already parsed for defaults. It is never applied to dense models (pure memory waste) and never overrides an explicit user `swa_full`/`n_swa` choice. Tradeoff: the full SWA cache scales with context_size, so it costs more memory at large contexts — hence the SWA gating and the documented `swa_full:false` opt-out. Assisted-by: Claude:claude-opus-4-8 [Claude Code] golangci-lint Co-authored-by: Ettore Di Giacinto <mudler@localai.io> |
||
|
LocalAI [bot]
|
79783120dd |
fix(config): gate parallel-slot default on per-device VRAM too (#10485) (#10507)
The first #10485 fix (#10494) made the Blackwell physical-batch boost per-device/context-aware, which neutralized the big compute-buffer OOM, but the reporter's 2x16 GiB consumer Blackwell still OOM'd. Tracing the post-fix log: the model now loads its weights, builds the main context and warms up fine, and dies only on the *last* allocation — the MTP draft context's 800 MiB KV cache on the tighter device. #10411 changed only two defaults: the physical batch (now gated) and a VRAM-scaled parallel-slot count. The KV cache is unified (n_ctx_seq == full context proves slots share the budget, so parallel doesn't multiply KV), but n_seq_max=4 still adds per-slot compute-graph / context-checkpoint / output scratch. On a device packed ~99% by a 27B model spanning both cards, that overhead is the few-hundred-MiB straw — which is why reverting #10411 (and only #10411) restores a working load. Gate the parallel-slot default on the same per-device headroom predicate as the batch boost: when a large context already fills a single card (largeContextForDevice), keep n_parallel=1. A user running one big-context model that barely fits across two consumer GPUs is not serving four concurrent tenants. Small contexts and large unified-memory devices (GB10) keep full concurrency. Applied on both the single-host path and the distributed router. Also make the auto-tuning visible and reversible (the debugging here needed DEBUG logs and a git bisect): - Log the effective performance-relevant runtime options at INFO once per model load ("effective runtime tuning …": context, n_batch, n_gpu_layers, parallel, flash_attention, f16) so an admin can see what will run and pin or override any value in the model YAML. - LOCALAI_DISABLE_HARDWARE_DEFAULTS=true skips the hardware auto-tuning entirely (mirrors LOCALAI_DISABLE_GUESSING) for stock llama.cpp behavior. Assisted-by: Claude:opus-4.8 [Claude Code] Signed-off-by: Ettore Di Giacinto <mudler@localai.io> Co-authored-by: Ettore Di Giacinto <mudler@localai.io> |
||
|
Ettore Di Giacinto
|
595e448714 |
docs(llama.cpp): note tensor split now works with quantized KV cache (#10135)
The split_mode: tensor description claimed tensor parallelism requires KV-cache quantization to be disabled. ggml-org/llama.cpp#23792 lifts that restriction by extending the meta backend to preserve shape information through KV-cache flatten/reshape, so cache_type_k/cache_type_v quantization can be combined with -sm tensor on builds that include it. Documentation only: no backend code, grpc-server.cpp comment, or llama.cpp pin changes. Assisted-by: Claude Code:claude-opus-4-8 Signed-off-by: Ettore Di Giacinto <mudler@localai.io> |
||
|
Zhao73
|
c61838dba6 |
docs: fix documentation typos (#10125)
Correct clear spelling mistakes in documentation without changing behavior. Confidence: high Scope-risk: narrow Tested: git diff --check; uvx codespell on changed files Not-tested: Full docs build not run; text-only changes Assisted-by: Codex:gpt-5 codespell |
||
|
LocalAI [bot]
|
4aad97971c |
chore: ⬆️ Update ggml-org/llama.cpp to 35c9b1f39ebe5a7bb83986d64415a079218be78d (#9998)
* ⬆️ Update ggml-org/llama.cpp Signed-off-by: github-actions[bot] <41898282+github-actions[bot]@users.noreply.github.com> * fix(llama-cpp): track upstream rename checkpoint_every_nt -> checkpoint_min_step Upstream llama.cpp renamed common_params::checkpoint_every_nt to checkpoint_min_step and changed its default from 8192 to 256. The semantics also shifted: it used to enforce a fixed checkpoint cadence during prefill, now it sets a minimum spacing between context checkpoints. Track the new field name in grpc-server.cpp and accept the old option names as backward- compatible aliases for users with existing configs. Signed-off-by: Ettore Di Giacinto <mudler@localai.io> Assisted-by: claude-code:claude-opus-4-7 --------- Signed-off-by: github-actions[bot] <41898282+github-actions[bot]@users.noreply.github.com> Signed-off-by: Ettore Di Giacinto <mudler@localai.io> Co-authored-by: mudler <2420543+mudler@users.noreply.github.com> Co-authored-by: Ettore Di Giacinto <mudler@localai.io> |
||
|
LocalAI [bot]
|
959de86761 |
feat(llama-cpp): make server-side prompt cache work by default (#9925)
Aligns LocalAI's llama-cpp gRPC backend with upstream's auto-on prompt cache path so repeated system prompts (agents, OpenAI/Anthropic-compatible CLIs, coding assistants) skip prefill on subsequent calls without any YAML changes. Reported in #9921. Upstream's server enables `kv_unified=true` (and bumps `n_parallel` to 4) when slot count is auto, which unlocks `cache_idle_slots`. LocalAI hardcodes `n_parallel=1` and so far also hardcoded `kv_unified=false`, which silently force-disables idle-slot saving at server init. The host prompt cache was allocated but never written across requests. Changes in backend/cpp/llama-cpp/grpc-server.cpp: - params.kv_unified: false -> true (single-slot path now benefits from the prompt cache; users can opt out with `kv_unified:false`) - params.n_ctx_checkpoints: 8 -> 32 (match upstream default) - params.cache_idle_slots = true initialized explicitly (upstream default) - params.checkpoint_every_nt = 8192 initialized explicitly (upstream default) - New option parsers: cache_idle_slots / idle_slots_cache, checkpoint_every_nt / checkpoint_every_n_tokens Docs: - features/text-generation.md: fix misleading `cache_ram` description (it's the host-side prompt cache, not the KV cache), document the kv_unified + cache_ram + cache_idle_slots interaction, add rows for the two newly-exposed options, and add a worked example for the agent/CLI workload from the issue. - advanced/model-configuration.md: mark the legacy `prompt_cache_path` / `prompt_cache_all` / `prompt_cache_ro` YAML fields as unused by the llama-cpp gRPC backend (they target upstream's CLI completion tool and are not consumed by grpc-server.cpp) and point readers at the new prompt-cache explainer. Closes #9921 Assisted-by: claude:opus-4.7 Signed-off-by: Ettore Di Giacinto <mudler@localai.io> Co-authored-by: Ettore Di Giacinto <mudler@localai.io> |
||
|
Richard Palethorpe
|
c894d9c826 |
feat(sglang): wire engine_args, add cuda13 build, ship MTP gallery demos (#9686)
Bring the sglang Python backend up to feature parity with vllm by adding
the same engine_args:-map plumbing the vLLM backend already has. Any
ServerArgs field (~380 in sglang 0.5.11) becomes settable from a model
YAML, including the speculative-decoding flags needed for Multi-Token
Prediction. Validation matches the vllm backend's: keys are checked
against dataclasses.fields(ServerArgs), unknown keys raise ValueError
with a difflib close-match suggestion at LoadModel time, and the typed
ModelOptions fields keep their existing meaning with engine_args
overriding them.
Backend code:
* backend/python/sglang/backend.py: add _apply_engine_args, import
dataclasses/difflib/ServerArgs, call from LoadModel; rename Seed ->
sampling_seed (sglang 0.5.11 renamed the SamplingParams field).
* backend/python/sglang/test.py + test.sh + Makefile: six unit tests
exercising the helper directly (no engine load required).
Build / CI / backend gallery (cuda13 + l4t13 paths are now first-class):
* backend/python/sglang/install.sh: add --prerelease=allow because
sglang 0.5.11 hard-pins flash-attn-4 which only ships beta wheels;
add --index-strategy=unsafe-best-match for cublas12 so the cu128
torch index wins over default-PyPI's cu130; new pyproject.toml-driven
l4t13 install path so [tool.uv.sources] can pin torch/torchvision/
torchaudio/sglang to the jetson-ai-lab index without forcing every
transitive PyPI dep through the L4T mirror's flaky proxy (mirrors the
equivalent fix in backend/python/vllm/install.sh).
* backend/python/sglang/pyproject.toml (new): L4T project spec with
explicit-source jetson-ai-lab index. Replaces requirements-l4t13.txt
for the l4t13 BUILD_PROFILE; other profiles still go through the
requirements-*.txt pipeline via libbackend.sh's installRequirements.
* backend/python/sglang/requirements-l4t13.txt: removed; superseded
by pyproject.toml.
* backend/python/sglang/requirements-cublas{12,13}{,-after}.txt: pin
sglang>=0.5.11 (Gemma 4 floor); add cu130 torch index for cublas13
(new files) and cu128 torch index for cublas12 (default PyPI now
ships cu130 torch wheels by default and breaks cu12 hosts).
* backend/index.yaml: add cuda13-sglang and cuda13-sglang-development
capability mappings + image entries pointing at
quay.io/.../-gpu-nvidia-cuda-13-sglang.
* .github/workflows/backend.yml: new cublas13 sglang matrix entry,
mirroring vllm's cuda13 build.
Model gallery + docs:
* gallery/sglang.yaml: base sglang config template, mirrors vllm.yaml.
* gallery/sglang-gemma-4-{e2b,e4b}-mtp.yaml: Gemma 4 MTP demos
transcribed verbatim from the SGLang Gemma 4 cookbook MTP commands.
* gallery/sglang-mimo-7b-mtp.yaml: MiMo-7B-RL with built-in MTP heads
+ online fp8 weight quantization, verified end-to-end on a 16 GB
RTX 5070 Ti at ~88 tok/s. Uses mem_fraction_static: 0.7 because the
MTP draft worker's vocab embedding is loaded unquantised and OOMs
the static reservation at sglang's 0.85 default.
* gallery/index.yaml: three new entries (gemma-4-e2b-it:sglang-mtp,
gemma-4-e4b-it:sglang-mtp, mimo-7b-mtp:sglang).
* docs/content/features/text-generation.md: new SGLang section with
setup, engine_args reference, MTP demos, version requirements.
* .agents/sglang-backend.md (new): agent one-pager covering the flat
ServerArgs structure, the typed-vs-engine_args precedence, the
speculative-decoding cheatsheet, and the mem_fraction_static gotcha
documented above.
* AGENTS.md: index entry for the new agent doc.
Known limitation: the two Gemma 4 MTP gallery entries ship a recipe
that doesn't yet run on stock libraries. The drafter checkpoints
(google/gemma-4-{E2B,E4B}-it-assistant) declare
model_type: gemma4_assistant / Gemma4AssistantForCausalLM, which
neither transformers (<=5.6.0, including the SGLang cookbook's pinned
commit 91b1ab1f... and main HEAD) nor sglang's own model registry
(<=0.5.11) registers as of 2026-05-06. They will start working when
HF or sglang upstream registers the architecture -- no LocalAI
changes needed. The MiMo MTP demo and the non-MTP Gemma 4 paths work
today on this build (verified on RTX 5070 Ti, 16 GB).
Assisted-by: Claude:claude-opus-4-7 [Read] [Edit] [Bash] [WebFetch] [WebSearch]
Signed-off-by: Richard Palethorpe <io@richiejp.com>
|
||
|
LocalAI [bot]
|
a8d7d37a3c |
fix: unbreak master CI (docs, kokoros, vibevoice-cpp ABI) (#9682)
* fix(docs): correct broken Hugo relrefs The Hugo build has been failing on master since the relevant pages landed: - text-generation.md:720 referenced `/docs/features/distributed-mode`, but Hugo `relref` paths are relative to the content root, not the rendered URL. Drop the `/docs/` prefix so the lookup matches the existing `features/...` form used elsewhere in the file. - audio-transform.md:144 referenced `tts.md`; the actual page is `text-to-audio.md`. Assisted-by: Claude:claude-opus-4-7[1m] Signed-off-by: Ettore Di Giacinto <mudler@localai.io> * fix(kokoros): stub Diarize and AudioTransform Backend trait methods The recent backend.proto additions (Diarize, AudioTransform, AudioTransformStream) extended the gRPC Backend trait, breaking kokoros-grpc compilation with E0046 because the Rust implementation hadn't picked up the new methods. Add Unimplemented stubs matching the existing pattern for non-applicable RPCs in this TTS-only backend. Assisted-by: Claude:claude-opus-4-7[1m] Signed-off-by: Ettore Di Giacinto <mudler@localai.io> * fix(vibevoice-cpp): track upstream ABI + wire 1.5B voice cloning Two recent commits in mudler/vibevoice.cpp reshaped the vv_capi_tts signature without a corresponding bump on the LocalAI side: 3bd759c "1.5b: unify into a single tts entry point" inserted a ref_audio_path parameter between voice_path and dst_wav_path. ad856bd "1.5b: multi-speaker dialog support" promoted that to a (const char* const* ref_audio_paths, int n_ref_audio_paths) pair for per-speaker conditioning. Because purego resolves symbols by name and not by signature, the build kept linking; at runtime the misaligned arguments turned the TTS->ASR closed-loop test into a SIGSEGV inside cgo. Track HEAD explicitly and bring the bridge in line with it: * Update the CppTTS purego binding to the 9-arg form. purego marshals []*byte as a **char by handing the C side the underlying array address; nil/empty maps to NULL, which matches the C contract for "no reference audio" on the realtime-0.5B path. * Add a `ref_audio` gallery option (comma-separated, repeatable) that the 1.5B path consumes for runtime voice cloning. Multiple entries are interpreted as one WAV per speaker (Speaker 0..n-1). * TTSRequest.Voice now routes by extension/shape: `.wav` or a comma-separated list goes to ref_audio_paths; anything else stays on voice_path (realtime-0.5B's pre-baked voice gguf). * Pin VIBEVOICE_CPP_VERSION to ad856bd and wire the Makefile into the existing bump_deps matrix so future upstream rolls land as reviewable PRs instead of a silent CI break. Assisted-by: Claude:claude-opus-4-7[1m] Signed-off-by: Ettore Di Giacinto <mudler@localai.io> * refactor(vibevoice-cpp): use ModelOptions.AudioPath for 1.5B ref audio Use the existing audio_path field from ModelOptions (already plumbed through config_file's `audio_path:` YAML and consumed by other audio backends like kokoros) instead of inventing a custom `ref_audio:` Options[] string. Multi-speaker setups stay on a single comma- separated value. No behavior change beyond the gallery key name; per-call routing via TTSRequest.Voice is unchanged. Assisted-by: Claude:claude-opus-4-7[1m] Signed-off-by: Ettore Di Giacinto <mudler@localai.io> --------- Signed-off-by: Ettore Di Giacinto <mudler@localai.io> Co-authored-by: Ettore Di Giacinto <mudler@localai.io> |
||
|
Richard Palethorpe
|
8e43842175 |
feat(vllm, distributed): tensor parallel distributed workers (#9612)
* feat(vllm): build vllm from source for Intel XPU
Upstream publishes no XPU wheels for vllm. The Intel profile was
silently picking up a non-XPU wheel that imported but errored at
engine init, and several runtime deps (pillow, charset-normalizer,
chardet) were missing on Intel -- backend.py crashed at import time
before the gRPC server came up.
Switch the Intel profile to upstream's documented from-source
procedure (docs/getting_started/installation/gpu.xpu.inc.md in
vllm-project/vllm):
- Bump portable Python to 3.12 -- vllm-xpu-kernels ships only a
cp312 wheel.
- Source /opt/intel/oneapi/setvars.sh so vllm's CMake build sees
the dpcpp/sycl compiler from the oneapi-basekit base image.
- Hide requirements-intel-after.txt during installRequirements
(it used to 'pip install vllm'); install vllm's deps from a
fresh git clone of vllm via 'uv pip install -r
requirements/xpu.txt', swap stock triton for
triton-xpu==3.7.0, then 'VLLM_TARGET_DEVICE=xpu uv pip install
--no-deps .'.
- requirements-intel.txt trimmed to LocalAI's direct deps
(accelerate / transformers / bitsandbytes); torch-xpu, vllm,
vllm_xpu_kernels and the rest come from upstream's xpu.txt
during the source build.
- requirements.txt: add pillow + charset-normalizer + chardet --
used by backend.py and missing on the Intel install profile.
- run.sh: 'set -x' so backend startup is visible in container
logs (the gRPC startup error path was previously opaque).
Also adds a one-line docs example for engine_args.attention_backend
under the vLLM section, since older XE-HPG GPUs (e.g. Arc A770)
need TRITON_ATTN to bypass the cutlass path in vllm_xpu_kernels.
Tested end-to-end on an Intel Arc A770 with Qwen2.5-0.5B-Instruct
via LocalAI's /v1/chat/completions.
Assisted-by: Claude:claude-opus-4-7 [Claude Code]
Signed-off-by: Richard Palethorpe <io@richiejp.com>
* feat(vllm): add multi-node data-parallel follower worker
vLLM v1's multi-node story is one process per node sharing a DP
coordinator over ZMQ -- the head runs the API server with
data_parallel_size > 1 and followers run `vllm serve --headless ...`
with matching topology. Today LocalAI can already configure DP on the
head via the engine_args YAML map, but there's no way to bring up the
follower nodes -- so the head sits waiting for ranks that never
handshake.
Add `local-ai p2p-worker vllm`, mirroring MLXDistributed's structural
precedent (operator-launched, static config, no NATS placement). The
worker:
- Optionally self-registers with the frontend as an agent-type node
tagged `node.role=vllm-follower` so it's visible in the admin UI
and operators can scope ordinary models away via inverse
selectors.
- Resolves the platform-specific vllm backend via the gallery's
"vllm" meta-entry (cuda*, intel-vllm, rocm-vllm, ...).
- Runs vLLM as a child process so the heartbeat goroutine survives
until vLLM exits; forwards SIGINT/SIGTERM so vLLM can clean up its
ZMQ sockets before we tear down.
- Validates --headless + --start-rank 0 is rejected (rank 0 is the
head and must serve the API).
Backend run.sh dispatches `serve` as the first arg to vllm's own CLI
instead of LocalAI's backend.py gRPC server -- the follower speaks
ZMQ directly to the head, there is no LocalAI gRPC on the follower
side. Single-node usage is unchanged.
Generalises the gallery resolution helper into findBackendPath()
shared by MLX and vLLM workers; extracts ParseNodeLabels for the
comma-separated label parsing both use.
Ships with two compose recipes (`docker-compose.vllm-multinode.yaml`
for NVIDIA, `docker-compose.vllm-multinode.intel.yaml` for Intel
XPU/xccl) plus `tests/e2e/vllm-multinode/smoke.sh`. Both vendors are
supported (NCCL for CUDA/ROCm, xccl for XPU) but mixed-vendor DP is
not -- PyTorch's process group requires every rank to use the same
collective backend, and NCCL/xccl/gloo don't interoperate.
Out of scope (deferred): SmartRouter-driven placement of follower
ranks via NATS backend.install events, follower log streaming through
/api/backend-logs, tensor-parallel across nodes, disaggregated
prefill via KVTransferConfig.
Assisted-by: Claude:claude-opus-4-7 [Claude Code]
Signed-off-by: Richard Palethorpe <io@richiejp.com>
* test(vllm): CPU-only end-to-end test for multi-node DP
Adds tests/e2e/vllm-multinode/, a Ginkgo + testcontainers-go suite
that brings up a head + headless follower from the locally-built
local-ai:tests image, bind-mounts the cpu-vllm backend extracted by
make extract-backend-vllm so it's seen as a system backend (no gallery
fetch, no registry server), and asserts a chat completion across both
DP ranks. New `make test-e2e-vllm-multinode` target wires the docker
build, backend extract, and ginkgo run together; BuildKit caches both
images so re-runs only rebuild what changed. Tagged Label("VLLMMultinode")
so the existing distributed suite isn't pulled along.
Two pre-existing bugs surfaced by the test:
1. extract-backend-% (Makefile) failed for every backend, because all
backend images end with `FROM scratch` and `docker create` rejects
an image with no CMD/ENTRYPOINT. Fixed by passing
--entrypoint=/run.sh -- the container is never started, only
docker-cp'd, so the path doesn't have to exist; we just need
anything that satisfies the daemon's create-time validation.
2. backend/python/vllm/run.sh's `serve` shortcut for the multi-node DP
follower exec'd ${EDIR}/venv/bin/vllm directly, but uv bakes an
absolute build-time shebang (`#!/vllm/venv/bin/python3`) that no
longer resolves once the backend is relocated to BackendsPath.
_makeVenvPortable's shebang rewriter only matches paths that
already point at ${EDIR}, so the original shebang slips through
unchanged. Fixed by exec-ing ${EDIR}/venv/bin/python with the script
as an argument -- Python ignores the script's shebang in that case.
The test fixture caps memory aggressively (max_model_len=512,
VLLM_CPU_KVCACHE_SPACE=1, TORCH_COMPILE_DISABLE=1) so two CPU engines
fit on a 32 GB box. TORCH_COMPILE_DISABLE is currently mandatory for
cpu-vllm: torch._inductor's CPU-ISA probe runs even with
enforce_eager=True and needs g++ on PATH, which the LocalAI runtime
image doesn't ship -- to be addressed in a follow-up that bundles a
toolchain in the cpu-vllm backend.
Assisted-by: Claude:claude-opus-4-7 [Claude Code]
Signed-off-by: Richard Palethorpe <io@richiejp.com>
* feat(vllm): bundle a g++ toolchain in the cpu-vllm backend image
torch._inductor's CPU-ISA probe (`cpu_model_runner.py:65 "Warming up
model for the compilation"`) shells out to `g++` at vllm engine
startup, regardless of `enforce_eager=True` -- the eager flag only
disables CUDA graphs, not inductor's first-batch warmup. The LocalAI
CPU runtime image (Dockerfile, unconditional apt list) does not ship
build-essential, and the cpu-vllm backend image is `FROM scratch`,
so any non-trivial inference on cpu-vllm crashes with:
torch._inductor.exc.InductorError:
InvalidCxxCompiler: No working C++ compiler found in
torch._inductor.config.cpp.cxx: (None, 'g++')
Bundling the toolchain in the CPU runtime image would bloat every
non-vllm-CPU deployment and force a single GCC version on backends
that may want clang or a different version. So this lives in the
backend, gated to BUILD_TYPE=='' (the CPU profile).
`package.sh` snapshots g++ + binutils + cc1plus + libstdc++ + libc6
(runtime + dev) + the math libs cc1plus links (libisl/libmpc/libmpfr/
libjansson) into ${BACKEND}/toolchain/, mirroring /usr/... layout. The
unversioned binaries on Debian/Ubuntu are symlink chains pointing into
multiarch packages (`g++` -> `g++-13` -> `x86_64-linux-gnu-g++-13`,
the latter in `g++-13-x86-64-linux-gnu`), so the package list resolves
both the version and the arch-triplet variant. Symlinks /lib ->
usr/lib and /lib64 -> usr/lib64 are recreated under the toolchain
root because Ubuntu's UsrMerge keeps them at /, and ld scripts
(`libc.so`, `libm.so`) hardcode `/lib/...` paths that --sysroot
re-roots into the toolchain.
The unversioned `g++`/`gcc`/`cpp` symlinks are replaced with wrapper
shell scripts that resolve their own location at runtime and pass
`--sysroot=<toolchain>` and `-B <toolchain>/usr/lib/gcc/<triplet>/<ver>/`
to the underlying versioned binary. That's how torch's bare `g++ foo.cpp
-o foo` invocation finds cc1plus (-B), system headers (--sysroot), and
the bundled libstdc++ (--sysroot, --sysroot is recursive into linker).
`run.sh` adds the toolchain bin dir to PATH and the toolchain's
shared-lib dir to LD_LIBRARY_PATH -- everything else (header search,
linker search, executable search) is encapsulated in the wrappers.
No-op for non-CPU builds, the dir doesn't exist there.
The cpu-vllm image grows by ~217 MB. Tradeoff is acceptable -- cpu-vllm
is already a niche profile (few users compared to GPU vllm) and the
alternative is a backend that crashes at first inference unless the
operator manually sets TORCH_COMPILE_DISABLE=1, which silently disables
all torch.compile optimizations.
Drops `TORCH_COMPILE_DISABLE=1` from tests/e2e/vllm-multinode -- the
smoke now exercises the real compile path through the bundled toolchain.
Test runtime is +20s for the warmup compile, still <90s end to end.
Assisted-by: Claude:claude-opus-4-7 [Claude Code]
Signed-off-by: Richard Palethorpe <io@richiejp.com>
* fix(vllm): scope jetson-ai-lab index to L4T-specific wheels via pyproject.toml
The L4T arm64 build resolves dependencies through pypi.jetson-ai-lab.io,
which hosts the L4T-specific torch / vllm / flash-attn wheels but also
transparently proxies the rest of PyPI through `/+f/<sha>/<filename>`
URLs. With `--extra-index-url` + `--index-strategy=unsafe-best-match`
uv would pick those proxy URLs for ordinary PyPI packages —
anthropic/openai/propcache/annotated-types — and fail when the proxy
503s. Master is hitting the same bug on its own l4t-vllm matrix entry.
Switch the l4t13 install path to a pyproject.toml that marks the
jetson-ai-lab index `explicit = true` and pins only torch, torchvision,
torchaudio, flash-attn, and vllm to it via [tool.uv.sources]. uv won't
consult the L4T mirror for anything else, so transitive deps fall back
to PyPI as the default index — no exposure to the proxy 503s.
`uv pip install -r requirements.txt` ignores [tool.uv.sources], so the
l4t13 branch in install.sh now invokes `uv pip install --requirement
pyproject.toml` directly, replacing the old requirements-l4t13*.txt
files. Other BUILD_PROFILEs continue using libbackend.sh's
installRequirements and never read pyproject.toml.
Local resolution test (x86_64, dry-run) confirms uv hits the L4T
index for torch and falls through to PyPI for everything else.
Assisted-by: claude-code:claude-opus-4-7-1m [Read] [Edit] [Bash] [Write]
Signed-off-by: Richard Palethorpe <io@richiejp.com>
---------
Signed-off-by: Richard Palethorpe <io@richiejp.com>
|
||
|
Richard Palethorpe
|
4916f8c880 |
feat(vllm): expose AsyncEngineArgs via generic engine_args YAML map (#9563)
* feat(vllm): expose AsyncEngineArgs via generic engine_args YAML map
LocalAI's vLLM backend wraps a small typed subset of vLLM's
AsyncEngineArgs (quantization, tensor_parallel_size, dtype, etc.).
Anything outside that subset -- pipeline/data/expert parallelism,
speculative_config, kv_transfer_config, all2all_backend, prefix
caching, chunked prefill, etc. -- requires a new protobuf field, a
Go struct field, an options.go line, and a backend.py mapping per
feature. That cadence is the bottleneck on shipping vLLM's
production feature set.
Add a generic `engine_args:` map on the model YAML that is
JSON-serialised into a new ModelOptions.EngineArgs proto field and
applied verbatim to AsyncEngineArgs at LoadModel time. Validation
is done by the Python backend via dataclasses.fields(); unknown
keys fail with the closest valid name as a hint.
dataclasses.replace() is used so vLLM's __post_init__ re-runs and
auto-converts dict values into nested config dataclasses
(CompilationConfig, AttentionConfig, ...). speculative_config and
kv_transfer_config flow through as dicts; vLLM converts them at
engine init.
Operators can now write:
engine_args:
data_parallel_size: 8
enable_expert_parallel: true
all2all_backend: deepep_low_latency
speculative_config:
method: deepseek_mtp
num_speculative_tokens: 3
kv_cache_dtype: fp8
without further proto/Go/Python plumbing per field.
Production defaults seeded by hooks_vllm.go: enable_prefix_caching
and enable_chunked_prefill default to true unless explicitly set.
Existing typed YAML fields (gpu_memory_utilization,
tensor_parallel_size, etc.) remain for back-compat; engine_args
overrides them when both are set.
Assisted-by: Claude:claude-opus-4-7 [Claude Code]
Signed-off-by: Richard Palethorpe <io@richiejp.com>
* chore(vllm): pin cublas13 to vLLM 0.20.0 cu130 wheel
vLLM's PyPI wheel is built against CUDA 12 (libcudart.so.12) and won't
load on a cu130 host. Switch the cublas13 build to vLLM's per-tag cu130
simple-index (https://wheels.vllm.ai/0.20.0/cu130/) and pin
vllm==0.20.0. The cu130-flavoured wheel ships libcudart.so.13 and
includes the DFlash speculative-decoding method that landed in 0.20.0.
cublas13 install gets --index-strategy=unsafe-best-match so uv consults
both the cu130 index and PyPI when resolving — PyPI also publishes
vllm==0.20.0, but with cu12 binaries that error at import time.
Verified: Qwen3.5-4B + z-lab/Qwen3.5-4B-DFlash loads and serves chat
completions on RTX 5070 Ti (sm_120, cu130).
Assisted-by: Claude:claude-opus-4-7 [Claude Code]
Signed-off-by: Richard Palethorpe <io@richiejp.com>
* ci(vllm): bot job to bump cublas13 vLLM wheel pin
vLLM's cu130 wheel index URL is itself version-locked
(wheels.vllm.ai/<TAG>/cu130/, no /latest/ alias upstream), so a vLLM
bump means rewriting two values atomically — the URL segment and the
version constraint. bump_deps.sh handles git-sha-in-Makefile only;
add a sibling bump_vllm_wheel.sh and a matching workflow job that
mirrors the existing matrix's PR-creation pattern.
The bumper queries /releases/latest (which excludes prereleases),
strips the leading 'v', and seds both lines unconditionally. When the
file is already on the latest tag the rewrite is a no-op and
peter-evans/create-pull-request opens no PR.
Assisted-by: Claude:claude-opus-4-7 [Claude Code]
Signed-off-by: Richard Palethorpe <io@richiejp.com>
* docs(vllm): document engine_args and speculative decoding
The new engine_args: map plumbs arbitrary AsyncEngineArgs through to
vLLM, but the public docs only covered the basic typed fields. Add a
short subsection in the vLLM section explaining the typed/generic
split and showing a worked DFlash speculative-decoding config, with
pointers to vLLM's SpeculativeConfig reference and z-lab's drafter
collection.
Assisted-by: Claude:claude-opus-4-7 [Claude Code]
Signed-off-by: Richard Palethorpe <io@richiejp.com>
---------
Signed-off-by: Richard Palethorpe <io@richiejp.com>
Co-authored-by: Ettore Di Giacinto <mudler@users.noreply.github.com>
|
||
|
Ettore Di Giacinto
|
21eace40ec |
feat(llama-cpp): expose split_mode option for multi-GPU placement (#9560)
Adds split_mode (alias sm) to the llama.cpp backend options allowlist, accepting none|layer|row|tensor. The tensor value targets the experimental backend-agnostic tensor parallelism from ggml-org/llama.cpp#19378 and requires a llama.cpp build that includes that PR, FlashAttention enabled, KV-cache quantization disabled, and a manually set context size. Assisted-by: Claude:claude-opus-4-7 Signed-off-by: Ettore Di Giacinto <mudler@localai.io> |
||
|
Ettore Di Giacinto
|
95efb8a562 |
feat(backend): add turboquant llama.cpp-fork backend (#9355)
* feat(backend): add turboquant llama.cpp-fork backend
turboquant is a llama.cpp fork (TheTom/llama-cpp-turboquant, branch
feature/turboquant-kv-cache) that adds a TurboQuant KV-cache scheme.
It ships as a first-class backend reusing backend/cpp/llama-cpp sources
via a thin wrapper Makefile: each variant target copies ../llama-cpp
into a sibling build dir and invokes llama-cpp's build-llama-cpp-grpc-server
with LLAMA_REPO/LLAMA_VERSION overridden to point at the fork. No
duplication of grpc-server.cpp — upstream fixes flow through automatically.
Wires up the full matrix (CPU, CUDA 12/13, L4T, L4T-CUDA13, ROCm, SYCL
f32/f16, Vulkan) in backend.yml and the gallery entries in index.yaml,
adds a tests-turboquant-grpc e2e job driven by BACKEND_TEST_CACHE_TYPE_K/V=q8_0
to exercise the KV-cache config path (backend_test.go gains dedicated env
vars wired into ModelOptions.CacheTypeKey/Value — a generic improvement
usable by any llama.cpp-family backend), and registers a nightly auto-bump
PR in bump_deps.yaml tracking feature/turboquant-kv-cache.
scripts/changed-backends.js gets a special-case so edits to
backend/cpp/llama-cpp/ also retrigger the turboquant CI pipeline, since
the wrapper reuses those sources.
* feat(turboquant): carry upstream patches against fork API drift
turboquant branched from llama.cpp before upstream commit 66060008
("server: respect the ignore eos flag", #21203) which added the
`logit_bias_eog` field to `server_context_meta` and a matching
parameter to `server_task::params_from_json_cmpl`. The shared
backend/cpp/llama-cpp/grpc-server.cpp depends on that field, so
building it against the fork unmodified fails.
Cherry-pick that commit as a patch file under
backend/cpp/turboquant/patches/ and apply it to the cloned fork
sources via a new apply-patches.sh hook called from the wrapper
Makefile. Simplifies the build flow too: instead of hopping through
llama-cpp's build-llama-cpp-grpc-server indirection, the wrapper now
drives the copied Makefile directly (clone -> patch -> build).
Drop the corresponding patch whenever the fork catches up with
upstream — the build fails fast if a patch stops applying, which
is the signal to retire it.
* docs: add turboquant backend section + clarify cache_type_k/v
Document the new turboquant (llama.cpp fork with TurboQuant KV-cache)
backend alongside the existing llama-cpp / ik-llama-cpp sections in
features/text-generation.md: when to pick it, how to install it from
the gallery, and a YAML example showing backend: turboquant together
with cache_type_k / cache_type_v.
Also expand the cache_type_k / cache_type_v table rows in
advanced/model-configuration.md to spell out the accepted llama.cpp
quantization values and note that these fields apply to all
llama.cpp-family backends, not just vLLM.
* feat(turboquant): patch ggml-rpc GGML_OP_COUNT assertion
The fork adds new GGML ops bringing GGML_OP_COUNT to 97, but
ggml/include/ggml-rpc.h static-asserts it equals 96, breaking
the GGML_RPC=ON build paths (turboquant-grpc / turboquant-rpc-server).
Carry a one-line patch that updates the expected count so the
assertion holds. Drop this patch whenever the fork fixes it upstream.
* feat(turboquant): allow turbo* KV-cache types and exercise them in e2e
The shared backend/cpp/llama-cpp/grpc-server.cpp carries its own
allow-list of accepted KV-cache types (kv_cache_types[]) and rejects
anything outside it before the value reaches llama.cpp's parser. That
list only contains the standard llama.cpp types — turbo2/turbo3/turbo4
would throw "Unsupported cache type" at LoadModel time, meaning
nothing the LocalAI gRPC layer accepted was actually fork-specific.
Add a build-time augmentation step (patch-grpc-server.sh, called from
the turboquant wrapper Makefile) that inserts GGML_TYPE_TURBO2_0/3_0/4_0
into the allow-list of the *copied* grpc-server.cpp under
turboquant-<flavor>-build/. The original file under backend/cpp/llama-cpp/
is never touched, so the stock llama-cpp build keeps compiling against
vanilla upstream which has no notion of those enum values.
Switch test-extra-backend-turboquant to set
BACKEND_TEST_CACHE_TYPE_K=turbo3 / _V=turbo3 so the e2e gRPC suite
actually runs the fork's TurboQuant KV-cache code paths (turbo3 also
auto-enables flash_attention in the fork). Picking q8_0 here would
only re-test the standard llama.cpp path that the upstream llama-cpp
backend already covers.
Refresh the docs (text-generation.md + model-configuration.md) to
list turbo2/turbo3/turbo4 explicitly and call out that you only get
the TurboQuant code path with this backend + a turbo* cache type.
* fix(turboquant): rewrite patch-grpc-server.sh in awk, not python3
The builder image (ubuntu:24.04 stage-2 in Dockerfile.turboquant)
does not install python3, so the python-based augmentation step
errored with `python3: command not found` at make time. Switch to
awk, which ships in coreutils and is already available everywhere
the rest of the wrapper Makefile runs.
* Apply suggestion from @mudler
Signed-off-by: Ettore Di Giacinto <mudler@users.noreply.github.com>
---------
Signed-off-by: Ettore Di Giacinto <mudler@users.noreply.github.com>
|
||
|
Ettore Di Giacinto
|
9ca03cf9cc |
feat(backends): add ik-llama-cpp (#9326)
* feat(backends): add ik-llama-cpp Signed-off-by: Ettore Di Giacinto <mudler@localai.io> * chore: add grpc e2e suite, hook to CI, update README Signed-off-by: Ettore Di Giacinto <mudler@localai.io> * Apply suggestion from @mudler Signed-off-by: Ettore Di Giacinto <mudler@users.noreply.github.com> * Apply suggestion from @mudler Signed-off-by: Ettore Di Giacinto <mudler@users.noreply.github.com> --------- Signed-off-by: Ettore Di Giacinto <mudler@localai.io> Signed-off-by: Ettore Di Giacinto <mudler@users.noreply.github.com> |
||
|
Ettore Di Giacinto
|
7e0b73deaa |
fix(docs): fix broken references to distributed mode
Signed-off-by: Ettore Di Giacinto <mudler@localai.io> |
||
|
Ettore Di Giacinto
|
580517f9db |
feat: pass-by metadata to predict options (#8795)
Signed-off-by: Ettore Di Giacinto <mudler@localai.io> |
||
|
Ettore Di Giacinto
|
05904c77f5 |
chore(exllama): drop backend now almost deprecated (#8186)
exllama2 development has stalled and only old architectures are supported. exllamav3 is still in development, meanwhile cleaning up exllama2 from the gallery. Signed-off-by: Ettore Di Giacinto <mudler@localai.io> |
||
|
Ettore Di Giacinto
|
4bf2f8bbd8 |
chore(docs): update docs with Anthropic API and openresponses
Signed-off-by: Ettore Di Giacinto <mudler@localai.io> |
||
|
Ettore Di Giacinto
|
2387b266d8 |
chore(llama.cpp): Add Missing llama.cpp Options to gRPC Server (#7584)
Signed-off-by: Ettore Di Giacinto <mudler@localai.io> |
||
|
Ettore Di Giacinto
|
2cc4809b0d |
feat: docs revamp (#7313)
* docs Signed-off-by: Ettore Di Giacinto <mudler@localai.io> * Small enhancements Signed-off-by: Ettore Di Giacinto <mudler@localai.io> * Enhancements * Default to zen-dark Signed-off-by: Ettore Di Giacinto <mudler@localai.io> * fixups Signed-off-by: Ettore Di Giacinto <mudler@localai.io> --------- Signed-off-by: Ettore Di Giacinto <mudler@localai.io> |
||
|
Ettore Di Giacinto
|
6ca4d38a01 |
docs/examples: enhancements (#1572)
* docs: re-order sections * fix references * Add mixtral-instruct, tinyllama-chat, dolphin-2.5-mixtral-8x7b * Fix link * Minor corrections * fix: models is a StringSlice, not a String Signed-off-by: Ettore Di Giacinto <mudler@localai.io> * WIP: switch docs theme * content * Fix GH link * enhancements * enhancements * Fixed how to link Signed-off-by: lunamidori5 <118759930+lunamidori5@users.noreply.github.com> * fixups * logo fix * more fixups * final touches --------- Signed-off-by: Ettore Di Giacinto <mudler@localai.io> Signed-off-by: lunamidori5 <118759930+lunamidori5@users.noreply.github.com> Co-authored-by: lunamidori5 <118759930+lunamidori5@users.noreply.github.com> |
||
|
Ettore Di Giacinto
|
c5c77d2b0d |
docs: Initial import from localai-website (#1312)
Signed-off-by: Ettore Di Giacinto <mudler@localai.io> |