mirror of
https://github.com/mudler/LocalAI.git
synced 2026-07-10 00:08:42 -04:00
eb32cd9073
Realtime sessions previously lazy-loaded each pipeline sub-model (VAD,
transcription, LLM, TTS) on first use, so every cold session paid a
per-request model-load stall and load errors only surfaced mid-stream.
Warm the whole pipeline eagerly and blockingly at session start
(including the voice-gate speaker-recognition model, which an enforced
gate blocks each utterance on; compaction's summary_model stays lazy
since it only runs off the response path):
- Add backend.PreloadModel / PreloadModelByName as the single load path
for every modality (no transcription special-case; backend-omitted
configs are deprecated).
- The realtime session blocks on Model.Warmup and returns a
model_load_error to the client if any stage fails to load;
updateSession warms in the background. Opt out per pipeline with
pipeline.disable_warmup, exposed as a UI toggle via the
config-metadata registry.
Add a LocalAI-native POST /backend/load (and /v1/backend/load) that
pre-loads a model -- expanding realtime pipelines into their sub-models
-- as the inverse of /backend/shutdown. There is one preload engine
(backend.PreloadStages): the realtime Warmup methods, /backend/load and
the --load-to-memory startup flag all use it, so --load-to-memory now
also expands pipeline models and records load-failure traces. Pipeline
sub-model alias resolution is likewise shared
(ModelConfigLoader.LoadResolvedModelConfig). Surface the endpoint
everywhere an admin manages models:
- MCP admin tool load_model (httpapi + inproc clients, safety/catalog
prompts, catalog/dispatch tests).
- "Load into memory" action in the React models UI.
- Swagger regenerated; docs moved to the general backend-monitor page
since it is not realtime-specific.
Fix a Traces UI crash ("json: unsupported value: -Inf"): audio-snippet
RMS/peak now floor at a finite dBFS, and backend-trace data is sanitized
to drop non-finite floats before marshaling. The sanitizer is
copy-on-write -- it runs on every RecordBackendTrace, so containers are
only re-allocated on the paths that actually changed.
Migrate core/http/openresponses_test.go onto the prebuilt mock-backend
the rest of the http suite already uses -- it was the last spec still
pointing at a real HuggingFace model, so it 404'd wherever no vision
backend was built -- and fix its item_reference specs to send the
spec's "id" field instead of "item_id", which the handler never
accepted.
Assisted-by: Claude:claude-opus-4-8 Claude Code
Signed-off-by: Richard Palethorpe <io@richiejp.com>
2.0 KiB
2.0 KiB
Tool catalog
The MCP tools/list endpoint also exposes the full input schema for each of these. The list below is the canonical curated description.
Read-only
gallery_search— Search configured galleries for installable models.list_installed_models— List models currently installed on this LocalAI. Optionalcapabilityfilter (e.g.chat,embed,image).list_galleries— List configured model galleries.list_backends— List installed backends.list_known_backends— List backends available to install from configured backend galleries.get_job_status— Poll the status of an install/delete/upgrade job by id.get_model_config— Read the YAML/JSON config of an installed model.vram_estimate— Estimate VRAM use for a model under a given config.system_info— LocalAI version, paths, distributed flag, loaded models, installed backends.list_nodes— List federated worker nodes (only useful in distributed mode).
Mutating (require user confirmation per safety rule 1)
install_model— Install a model from a gallery. Returns a job id; poll withget_job_status.import_model_uri— Install a model from an arbitrary URI (HuggingFace, OCI, http(s), file://). May returnambiguous_backendwhen several backends apply; call again withbackend_preferenceto disambiguate.delete_model— Delete an installed model.install_backend— Install a backend.upgrade_backend— Upgrade an installed backend by name.edit_model_config— Patch (deep-merge) JSON into an installed model's config.reload_models— Reload all model configs from disk.load_model— Pre-load a model into memory so the first request pays no cold-start cost. For a realtime pipeline model, every sub-model (VAD, transcription, LLM, TTS, sound_detection, voice_recognition) is loaded. Inverse of stopping a model.toggle_model_state— Enable or disable a model (action:enableordisable).toggle_model_pinned— Pin or unpin a model (action:pinorunpin).