Out-of-bounds read in SmartypantsRenderer.smartLeftAngle (CWE-125,
CVSS 7.5). Reachable transitively via LocalAGI's Email connector,
which renders inbound HTML email replies using html.CommonFlags
(includes Smartypants). An unmatched `<` in the inbound body could
panic the agent service.
Bump to v0.0.0-20260411013819-759bbc3e3207 (contains the fix). The
klauspost/compress entry loses its `// indirect` tag because
go mod tidy noticed pkg/utils/untar.go imports it directly.
Assisted-by: Claude:claude-opus-4-7 [Claude-Code]
Signed-off-by: Richard Palethorpe <io@richiejp.com>
* realtime: honor output_modalities to skip TTS in text-only mode
The emulated realtime pipeline previously ignored the OpenAI Realtime spec
field output_modalities and always synthesized TTS. Add resolveOutputModalities
+ modalitiesContainAudio helpers and gate the TTS / ResponseOutputAudio*
emission so a client requesting ["text"] gets only ResponseOutputText* events.
This lets thin clients (e.g. thing5-poc) cache TTS on the client side while
still using the realtime WS for VAD + STT + LLM + tool-call parsing.
Assisted-by: Claude:claude-opus-4-7
* realtime: plumb response-level output_modalities and echo on session
Follow-up to the previous commit:
- Resolve response.create's output_modalities at the gate so a per-response
override of an audio session is honored (the test asserted this contract
but the production call site was passing nil).
- Mirror OutputModalities in the RealtimeSession echo so session.update
round-trips the client-supplied value, matching MaxOutputTokens's pattern.
Assisted-by: Claude:claude-opus-4-7
* realtime: silence errcheck on deferred os.Remove of TTS file
CI's errcheck flagged the pre-existing `defer os.Remove(audioFilePath)`
inside the audio-emission block (now wrapped by the modality gate). Wrap
the call in a closure that explicitly discards the error — the canonical
Go pattern for "I want to defer a cleanup whose error I genuinely don't
care about."
Assisted-by: Claude:claude-opus-4-7 golangci-lint
---------
Co-authored-by: Ettore Di Giacinto <mudler@localai.io>
The Ollama /api/tags handler passes a nil filter to galleryop.ListModels.
When ModelsPath contains any non-skipped loose file the function then
calls filter(name, nil) and panics, which Echo surfaces to clients as
"Server disconnected without sending a response" - the exact failure
Home Assistant's Ollama integration reports against LocalAI.
Mirror the nil guard already present in
ModelConfigLoader.GetModelConfigsByFilter so every caller is safe, and
add a regression test that exercises the loose-file path with a nil
filter.
Assisted-by: claude:claude-opus-4-7 [Claude Code]
Signed-off-by: Ettore Di Giacinto <mudler@localai.io>
Co-authored-by: Ettore Di Giacinto <mudler@localai.io>
* fix(streaming): comply with OpenAI usage / stream_options spec (#8546)
LocalAI emitted `"usage":{"prompt_tokens":0,...}` on every streamed
chunk because `OpenAIResponse.Usage` was a value type without
`omitempty`. The official OpenAI Node SDK and its consumers
(continuedev/continue, Kilo Code, Roo Code, Zed, IntelliJ Continue)
filter on a truthy `result.usage` to detect the trailing usage chunk;
LocalAI's zero-but-non-null usage on every intermediate chunk made
that filter swallow every content chunk and surface an empty chat
response while the server log looked successful.
Changes:
- `core/schema/openai.go`: `Usage *OpenAIUsage \`json:"usage,omitempty"\``
so intermediate chunks no longer carry a `usage` key. Add
`OpenAIRequest.StreamOptions` with `include_usage` to mirror OpenAI's
request field.
- `core/http/endpoints/openai/chat.go` and `completion.go`: keep using
the `Usage` struct field as an in-process channel for the running
cumulative, but strip it before JSON marshalling. When the request
set `stream_options.include_usage: true`, emit a dedicated trailing
chunk with `"choices": []` and the populated usage (matching the
OpenAI spec and llama.cpp's server behavior).
- `chat_emit.go`: new `streamUsageTrailerJSON` helper; drop the
`usage` parameter from `buildNoActionFinalChunks` since chunks no
longer carry usage.
- Update `image.go`, `inpainting.go`, `edit.go` to wrap their Usage
values with `&` for the new pointer field.
- UI: send `stream_options:{include_usage:true}` from the React
(`useChat.js`) and legacy (`static/chat.js`) chat clients so the
token-count badge keeps populating now that the server is
spec-compliant.
Tests:
- New `chat_stream_usage_test.go` pins the spec invariants:
intermediate chunks have no `usage` key, the trailer JSON has
`"choices":[]` and a populated `usage`, and `OpenAIRequest` parses
`stream_options.include_usage`.
- Update `chat_emit_test.go` to reflect that finals no longer embed
usage.
Verified against the live LocalAI instance: before the fix Continue's
filter logic swallowed 16/16 token chunks; with the new shape it
yields 4/5 and routes usage through the dedicated trailer chunk.
Fixes#8546
Assisted-by: Claude:opus-4.7 [Claude Code]
Signed-off-by: Ettore Di Giacinto <mudler@localai.io>
* fix(streaming): silence errcheck on usage trailer Fprintf
The new spec-compliant `stream_options.include_usage` trailer writes
were flagged by errcheck since they're new code (golangci-lint runs
new-from-merge-base on master); the surrounding `fmt.Fprintf` data:
writes are grandfathered. Drop the return values explicitly to match
the linter's contract without adding a nolint shim.
Assisted-by: Claude:opus-4.7 [Claude Code]
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>
The llama.cpp backend already accepts a free-form options: array in the
model config that maps to common_params fields, but a coverage audit
against upstream pin 7f3f843c flagged 12 user-visible knobs that were
neither set via the typed proto fields nor reachable via options:.
Wire them up under the existing if/else chain in params_parse, before
the speculative section. Each new option follows the file's prevailing
patterns (try/catch around numeric parses, the same true/1/yes/on bool
form used elsewhere, hardware_concurrency() fallback for thread counts,
mirror of draft_override_tensor for override_tensor).
Top-level / batching / IO:
- n_ubatch (alias ubatch) -- physical batch size; was previously
force-aliased to n_batch at line 482, blocking embedding/rerank
workloads that need independent control
- threads_batch (alias n_threads_batch) -- main-model batch threads;
mirrors the existing draft_threads_batch
- direct_io (alias use_direct_io) -- O_DIRECT model loads
- verbosity -- llama.cpp log threshold (line 479 had this commented
out)
- override_tensor (alias tensor_buft_overrides) -- per-tensor buffer
overrides for the main model; mirrors draft_override_tensor
Embedding / multimodal:
- pooling_type (alias pooling) -- mean/cls/last/rank/none; previously
only auto-flipped to RANK for rerankers
- embd_normalize (alias embedding_normalize) -- and the embedding
handler now reads params_base.embd_normalize instead of a hardcoded
2 at the previous embd_normalize literal in Embedding()
- mmproj_use_gpu (alias mmproj_offload) -- mmproj on CPU vs GPU
- image_min_tokens / image_max_tokens -- per-image vision token budget
Reasoning surface (the audit-focus three; LocalAI's existing
ReasoningConfig.DisableReasoning only feeds the per-request
chat_template_kwargs.enable_thinking and does not touch any of these):
- reasoning_format -- none/auto/deepseek/deepseek-legacy parser
- enable_reasoning (alias reasoning_budget) -- -1/0/>0 thinking budget
- prefill_assistant -- trailing-assistant-message prefill toggle
All 14 referenced fields exist on both the upstream pin and the
turboquant fork's common.h, so no LOCALAI_LEGACY_LLAMA_CPP_SPEC guard
is needed.
Docs: extend model-configuration.md with new "Reasoning Models",
"Multimodal Backend Options", "Embedding & Reranking Backend Options",
and "Other Backend Tuning Options" subsections; also refresh the
Speculative Type Values table to show the new dash-separated canonical
names alongside the underscore aliases LocalAI still accepts.
Assisted-by: claude-code:claude-opus-4-7
Signed-off-by: Ettore Di Giacinto <mudler@localai.io>
Co-authored-by: Ettore Di Giacinto <mudler@localai.io>
* ⬆️ Update ggml-org/llama.cpp
Signed-off-by: github-actions[bot] <41898282+github-actions[bot]@users.noreply.github.com>
* fix(llama-cpp): adapt to upstream COMMON_SPECULATIVE_TYPE_DRAFT rename
ggml-org/llama.cpp#22964 ("spec: update CLI arguments for better
consistency") renamed the speculative type enum values:
COMMON_SPECULATIVE_TYPE_DRAFT -> COMMON_SPECULATIVE_TYPE_DRAFT_SIMPLE
COMMON_SPECULATIVE_TYPE_EAGLE3 -> COMMON_SPECULATIVE_TYPE_DRAFT_EAGLE3
and the registered name strings flipped from underscore- to dash-
separated form (e.g. ngram_simple -> ngram-simple), with the bare
draft/eagle3 aliases replaced by draft-simple/draft-eagle3.
This broke the build with the new LLAMA_VERSION on every variant
(vulkan/arm64, darwin and likely all the rest) at grpc-server.cpp:461.
Update the upstream branch of the speculative-type fallback to use the
new identifier (the LOCALAI_LEGACY_LLAMA_CPP_SPEC fork branch keeps the
old name), and normalize spec_type option tokens before passing them to
common_speculative_types_from_names so existing model configs that say
spec_type:draft / spec_type:ngram_simple keep working.
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>
* ci(image): wire singleton merges + `--` artifact separator
Closes the same singletons gap on the LocalAI server image workflow that
PR #9781 closed for backends. The user observed it as missing
:latest-gpu-nvidia-cuda-12 etc. on quay.io/go-skynet/local-ai — the
build matrix has six single-arch entries with no corresponding merge
step, so their per-arch digests push (push-by-digest=true) and never
get tagged:
- -gpu-hipblas (hipblas-jobs)
- -gpu-nvidia-cuda-12 (core-image-build)
- -gpu-nvidia-cuda-13 (core-image-build)
- -gpu-intel (core-image-build)
- -nvidia-l4t-arm64 (gh-runner)
- -nvidia-l4t-arm64-cuda-13 (gh-runner)
Only :latest, :v<X>, :latest-gpu-vulkan and :v<X>-gpu-vulkan were
actually being published before this commit (the two multiarch suffixes
that had merge jobs).
Changes:
1. image.yml: add six new merge jobs, one per single-arch entry. Each
`needs:` only its parent build job (matching the existing pattern
for core-image-merge / gpu-vulkan-image-merge).
2. image_build.yml: switch artifact name to
`digests-localai<suffix>--<platform-tag-or-"single">`. The `--`
separator anchors the merge-side glob so a singleton tag-suffix
doesn't over-match a longer suffix that shares its prefix
(-nvidia-l4t-arm64 vs -nvidia-l4t-arm64-cuda-13). Same convention
as backend_build.yml's fix.
3. image_merge.yml: update the download pattern to match.
Next master push or tag release should produce :latest-gpu-hipblas,
:latest-gpu-nvidia-cuda-12, :latest-gpu-nvidia-cuda-13, :latest-gpu-intel,
:latest-nvidia-l4t-arm64, :latest-nvidia-l4t-arm64-cuda-13 (and their
:v<X>-* equivalents) for the first time on the post-#9781 workflow.
Assisted-by: Claude:claude-opus-4-7
Signed-off-by: Ettore Di Giacinto <mudler@localai.io>
* ci(image): add !cancelled() guard to all 8 image merge jobs
Parity pass with backend.yml's merge jobs (8521af14). Without
!cancelled(), GHA's default `needs:` cascade skips the merge when ANY
matrix cell of the parent build job fails or is cancelled — so a single
flaky leg would suppress publication of every other tag-suffix's
manifest list. Same fix the backend got after v4.2.1 showed 2 failed
singlearch builds cascade-skip 199 singlearch merge entries.
Applied to all 8 image merges:
- core-image-merge
- gpu-vulkan-image-merge
- gpu-nvidia-cuda-12-image-merge (added in e5300f1a)
- gpu-nvidia-cuda-13-image-merge (added in e5300f1a)
- gpu-intel-image-merge (added in e5300f1a)
- gpu-hipblas-image-merge (added in e5300f1a)
- nvidia-l4t-arm64-image-merge (added in e5300f1a)
- nvidia-l4t-arm64-cuda-13-image-merge (added in e5300f1a)
Build jobs (hipblas-jobs, core-image-build, gh-runner) are
intentionally NOT changed — they have no upstream `needs:` to cascade-
skip from.
Assisted-by: Claude:claude-opus-4-7
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>
* fix(middleware): parse OpenAI-spec tool_choice in /v1/chat/completions
Follows up on #9526 (the 3-site setter fix) by addressing the remaining
clause in #9508 — string mode and OpenAI-spec specific-function shape both
silently failed in the /v1/chat/completions parsing path.
Signed-off-by: Ettore Di Giacinto <mudler@localai.io>
* fix(middleware): restore LF endings and cover tool_choice parsing with specs
The previous commit on this branch saved core/http/middleware/request.go
with CRLF line endings, ballooning the diff against master to 684 / 651
for what is in reality a ~50-line parsing change. Restore LF (matches
.editorconfig end_of_line = lf).
Add 11 Ginkgo specs under "SetModelAndConfig tool_choice parsing
(chat completions)" that parallel the existing MergeOpenResponsesConfig
specs from #9509. They drive the full middleware chain (SetModelAndConfig
+ SetOpenAIRequest) and assert:
* "required" -> ShouldUseFunctions=true, no specific name
* "none" -> ShouldUseFunctions=false (tools disabled per OpenAI spec)
* "auto" -> default, tools available, no specific name
* {type:function, function:{name:X}} (spec) -> X is forced
* {type:function, name:X} (legacy) -> X is forced
* nested wins when both forms are present
* malformed shapes (no type, wrong type, no name, empty name) are no-ops
Update the inline comment on the string case to describe the actual
mechanism: "none" reaches SetFunctionCallString("none") downstream and
is then honored by ShouldUseFunctions() returning false. Before this PR
json.Unmarshal([]byte("none"), &functions.Tool{}) failed silently, so
"none" was ignored - making "none" actually work is a real behavior fix
this PR brings.
Signed-off-by: Ettore Di Giacinto <mudler@localai.io>
Assisted-by: Claude:opus-4-7 [Claude Code]
* fix(middleware): preserve pre-#9559 support for JSON-string-encoded tool_choice
Some non-spec clients send tool_choice as a JSON-encoded string of an
object form, e.g. "{\"type\":\"function\",\"function\":{\"name\":\"X\"}}".
The pre-#9559 code accepted this by accident: its case string: branch
ran json.Unmarshal([]byte(content), &functions.Tool{}), which succeeded
for that double-encoded shape even though it failed for the legitimate
plain string modes "auto" / "none" / "required".
The first version of this PR routed every string straight to
SetFunctionCallString as a mode, which fixed the plain-string cases but
silently regressed the double-encoded one (funcs.Select("{...}") returns
nothing). Restore the fallback: when a string looks like a JSON object,
try parsing it as a tool_choice map first; fall through to mode-string
handling only when no usable name comes out.
Factor the map-name extraction into a small helper
(extractToolChoiceFunctionName) so the string-fallback and the regular
map case go through identical code, and accept both the OpenAI-spec
nested shape and the legacy/Anthropic flat shape from either entry
point.
Add 3 Ginkgo specs covering the double-encoded case (nested form, legacy
form, and the fall-through when the JSON has no usable name).
Signed-off-by: Ettore Di Giacinto <mudler@localai.io>
Assisted-by: Claude:opus-4-7 [Claude Code]
* test(middleware): silence errcheck on AfterEach os.RemoveAll
The new tool_choice parsing tests added a second AfterEach that calls
os.RemoveAll(modelDir) without checking the error; errcheck flagged it.
Suppress with the standard _ = idiom. The pre-existing AfterEach on the
earlier Describe still elides the check the same way it did before -
leaving that untouched to keep this commit minimal.
Assisted-by: Claude:opus-4-7 [Claude Code]
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>
* fix(agentpool): close truncate-then-read race in agent_jobs.json persistence
Three call sites wrote and read agent_jobs.json (and agent_tasks.json)
through three independent mutexes:
- AgentJobService.ExecuteJob spawns go saveJobs(job) -> fileJobPersister
holding p.mu
- AgentJobService.SaveJobsToFile holding service.fileMutex
- AgentJobService.LoadJobsFromFile on a separate service instance holding
a different service.fileMutex
Nothing serialized those mutexes, and both writers used os.WriteFile, which
opens O_TRUNC. A reader landing between the truncate and the write saw a
zero-byte file and surfaced as `unexpected end of JSON input` at offset 0.
The macOS tests-apple job started hitting this consistently once the path
filter was removed from .github/workflows/test.yml and the file-mode race
test ran on every push (run 25823124797 was the first observed failure).
Two changes close the window:
1. fileJobPersister.saveTasksToFile / saveJobsToFile now write to a
same-directory temp file and os.Rename to the final path. rename(2) is
atomic on POSIX, so concurrent readers see either the prior contents or
the new contents and never a zero-byte window. The helper Syncs before
close so a crash mid-write leaves either the old file intact or the temp
behind (cleaned up on next save).
2. AgentJobService.{Load,Save}{Tasks,Jobs}{FromFile,ToFile} are collapsed
to thin wrappers around fileJobPersister, removing the duplicate write
path and the redundant service.fileMutex / service.tasksFile /
service.jobsFile fields. Within a single service all task/job I/O now
serializes on the persister's mutex; the atomic rename handles the
cross-instance case the tests exercise.
Adds a regression test that hammers SaveJobsToFile and LoadJobsFromFile
concurrently for 500ms across two service instances on the same paths.
On master this reproduces `unexpected end of JSON input` on Linux within
~500ms; with the fix the suite ran -until-it-fails for 30s (54 attempts,
all green).
Assisted-by: Claude:claude-opus-4-7 [Claude Code]
Signed-off-by: Ettore Di Giacinto <mudler@localai.io>
* refactor(agentpool): route service flush/load through JobPersister interface
The first cut of the race fix made AgentJobService.{Save,Load}{Tasks,Jobs}*
type-assert s.persister to *fileJobPersister so they could reach the
unexported saveTasksToFile / saveJobsToFile helpers. That defeats the
JobPersister interface: the service is back to reasoning about a concrete
implementation instead of an abstraction.
Promote the bulk-flush operations to the interface as FlushTasks / FlushJobs:
- fileJobPersister.FlushTasks/FlushJobs call the existing private helpers
(atomic temp+rename writes from the prior commit).
- dbJobPersister.FlushTasks/FlushJobs are no-ops because SaveTask/SaveJob
are already write-through to the database.
The service's four file-named methods now talk only to the interface:
LoadTasks/LoadJobs read through s.persister.LoadTasks/LoadJobs, and the
Save side calls FlushTasks/FlushJobs. The "FromFile"/"ToFile" suffixes
stay for backward compat with user_services.go and the existing tests,
but they no longer claim a file-only contract.
Assisted-by: Claude:claude-opus-4-7 [Claude Code]
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>
Mirror of 8521af14 (which fixed backend_merge.yml) for image_merge.yml.
Today's master-push run 25823024353 failed the gpu-vulkan-image-merge job
with the exact same error pattern the backend merge had on v4.2.2:
ERROR: quay.io/go-skynet/local-ai@sha256:68b22611...: not found
Same root cause: image_build.yml pushes the per-arch manifest to
quay.io/go-skynet/local-ai with push-by-digest=true (no tag), then the
merge runs minutes-to-hours later, by which time quay's per-repo manifest
GC has reaped the untagged digest from local-ai. The blob still lives in
quay's storage but local-ai@<digest> no longer resolves.
Three matching edits:
1. image_build.yml: anchor each per-arch digest into ci-cache immediately
after the push, reusing .github/scripts/anchor-digest-in-cache.sh with
SOURCE_IMAGE=quay.io/go-skynet/local-ai and TAG_SUFFIX defaulting to
"-core" for the core image (matches the artifact-name convention).
2. image_merge.yml: change the quay merge source from local-ai@<digest>
to ci-cache@<digest>. Same correctness argument as backend_merge.yml —
the manifest content is alive in ci-cache; buildx imagetools create
republishes it into local-ai and writes the user-facing manifest list
pointing at it. End state in local-ai is self-contained.
3. image_merge.yml: add a sparse `actions/checkout@v6` (only
.github/scripts) so cleanup-keepalive-tags.sh is available, plus the
cleanup step itself with TAG_SUFFIX matching the anchor's "-core"
placeholder.
v4.2.3's image.yml run completed successfully (~50 min between push and
merge — beat quay's GC). This commit closes the race for future releases
and master pushes regardless of run length.
Assisted-by: Claude:claude-opus-4-7
Signed-off-by: Ettore Di Giacinto <mudler@localai.io>
* fix(http): honor X-Forwarded-Prefix when proxy strips the prefix
Closes#9145.
Two related issues kept the React UI from loading when a reverse proxy
rewrites a sub-path with prefix-stripping (e.g. Caddy `handle_path`):
1. `BaseURL` only computed a prefix from the path StripPathPrefix had
removed, so when the proxy strips the prefix before forwarding, the
request arrives without it and the base URL was returned without a
prefix. Extract a `BasePathPrefix` helper and add an
`X-Forwarded-Prefix` header fallback so the prefix is recovered.
2. `<base href>` only changes how relative URLs resolve; the build
emits path-absolute references like `/assets/...` and
`/favicon.svg`, which still resolve against the origin and bypass
the proxy prefix. Rewrite those references in the served
`index.html` so the browser requests them through the proxy.
Adds unit coverage for `BaseURL` with a pre-stripped path and an
end-to-end test for the proxy-stripped scenario.
Assisted-by: Claude:claude-opus-4-7
* fix(http): gate X-Forwarded-Prefix through SafeForwardedPrefix in BasePathPrefix
BasePathPrefix consumed X-Forwarded-Prefix directly, so a value the
codebase elsewhere rejects (e.g. "//evil.com") slipped through and was
interpolated into the SPA index.html — both into the path-absolute asset
URL rewrite in serveIndex (turning "/assets/..." into "//evil.com/assets/...",
a protocol-relative URL that loads JS from a foreign origin) and into
<base href>. Route the header through the existing SafeForwardedPrefix
validator that StripPathPrefix and prefixRedirect already use, and
HTML-escape the prefix before injecting it into the asset rewrite as
defense in depth against attribute breakout.
Tests cover //evil.com, backslashes, control chars, CR/LF and a missing
leading slash; the integration test asserts an unsafe prefix can't poison
asset URLs.
Signed-off-by: Ettore Di Giacinto <mudler@localai.io>
Assisted-by: claude-code:claude-opus-4-7-1m [Read] [Edit] [Bash]
---------
Signed-off-by: Ettore Di Giacinto <mudler@localai.io>
Co-authored-by: Ettore Di Giacinto <mudler@localai.io>
* fix(distributed): cascade-clean stale node_models on drain and filter routing by healthy status
Stale node_models rows (state="loaded") were surviving past the healthy
state of their owning node, causing /embeddings (and other inference
paths) to dispatch to a backend whose process was gone or drained. The
downstream symptom in a live cluster was pgvector rejecting inserts
with "vector cannot have more than 16000 dimensions (SQLSTATE 54000)"
because the misbehaving backend silently returned a malformed
(oversized) tensor; the Models page showed the model as "running"
without an associated node, like a stale entry, even though the node
was no longer visible in the Nodes view.
Two changes here, plus a third in a follow-up commit:
- MarkDraining now cascade-deletes node_models rows for the affected
node, mirroring MarkOffline. Drains are explicit operator actions —
the box has been intentionally taken out of rotation — so clearing
the rows stops the Models UI from misreporting and prevents the
routing layer from picking those rows if scheduling logic is ever
relaxed. In-flight requests already hold their gRPC client through
Route() and finish normally; the only observable effect is a
non-fatal IncrementInFlight warning, acceptable for a drain.
MarkUnhealthy is deliberately left status-only: it fires from
managers_distributed / reconciler on a single nats.ErrNoResponders
with no retry, so a transient NATS hiccup must not nuke every loaded
model and force a full reload on recovery.
- FindAndLockNodeWithModel's inner JOIN now filters on
backend_nodes.status = healthy in addition to node_models.state =
loaded. The previous version relied on the second node-fetch step to
reject non-healthy nodes, but a concurrent reader could still pick
the same stale row in the same window. Belt-and-braces.
- DistributedConfig.PerModelHealthCheck renamed to
DisablePerModelHealthCheck and inverted at the call site so
per-model gRPC probing is on by default. The probe (now made
consecutive-miss aware in a follow-up commit) independently health-
checks each model's gRPC address and removes stale node_models rows
when the backend has crashed even though the worker's node-level
heartbeat is still arriving.
Migration: the field had no CLI flag, env var binding, or YAML key
in tree (only the bare struct field), so there is no user-facing
migration. Anything constructing DistributedConfig in code needs to
drop the assignment (default now does the right thing) or invert it.
Assisted-by: Claude:claude-opus-4-7 go-vet go-test golangci-lint
Signed-off-by: Ettore Di Giacinto <mudler@localai.io>
* fix(distributed): require consecutive misses before per-model probe removes a row
The per-model gRPC probe used to remove a node_models row on a single
failed health check. With the per-model probe now on by default, that
made any 5-second gRPC blip (network jitter, a long-running request
hogging the worker's gRPC server thread, brief GC pause) trigger a
full reload of the affected model — too eager for production.
Require perModelMissThreshold (3) consecutive failed probes before
removal. At the default 15s tick a model must be unreachable for ~45s
before reap; a single successful probe in between resets the streak.
Per-(node, model, replica) state tracked under a mutex on the monitor.
If the removal call itself fails, the miss counter is left in place
so the next tick retries rather than starting the streak over.
Tests:
- removes stale model via per-model health check after consecutive
failures (replaces the single-shot expectation)
- preserves model row when an intermittent failure is followed by a
success (covers the reset-on-success path and verifies the counter
reset by failing twice more without crossing threshold)
- newTestHealthMonitor initializes the misses map so direct-construct
test helpers don't nil-map-panic in the probe path
Assisted-by: Claude:claude-opus-4-7 go-vet go-test golangci-lint
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>
* feat(liquid-audio): add LFM2.5-Audio any-to-any backend + realtime_audio usecase
Wires LiquidAI's LFM2.5-Audio-1.5B as a self-contained Realtime API model:
single engine handles VAD, transcription, LLM, and TTS in one bidirectional
stream — drop-in alternative to a VAD+STT+LLM+TTS pipeline.
Backend
- backend/python/liquid-audio/ — new Python gRPC backend wrapping the
`liquid-audio` package. Modes: chat / asr / tts / s2s, voice presets,
Load/Predict/PredictStream/AudioTranscription/TTS/VAD/AudioToAudioStream/
Free and StartFineTune/FineTuneProgress/StopFineTune. Runtime monkey-patch
on `liquid_audio.utils.snapshot_download` so absolute local paths from
LocalAI's gallery resolve without a HF round-trip. soundfile in place of
torchaudio.load/save (torchcodec drags NVIDIA NPP we don't bundle).
- backend/backend.proto + pkg/grpc/{backend,client,server,base,embed,
interface}.go — new AudioToAudioStream RPC mirroring AudioTransformStream
(config/frame/control oneof in; typed event+pcm+meta out).
- core/services/nodes/{health_mock,inflight}_test.go — add stubs for the
new RPC to the test fakes.
Config + capabilities
- core/config/backend_capabilities.go — UsecaseRealtimeAudio, MethodAudio
ToAudioStream, UsecaseInfoMap entry, liquid-audio BackendCapability row.
- core/config/model_config.go — FLAG_REALTIME_AUDIO bitmask, ModalityGroups
membership in both speech-input and audio-output groups so a lone flag
still reads as multimodal, GetAllModelConfigUsecases entry, GuessUsecases
branch.
Realtime endpoint
- core/http/endpoints/openai/realtime.go — extract prepareRealtimeConfig()
so the gate is unit-testable; accept realtime_audio models and self-fill
empty pipeline slots with the model's own name (user-pinned slots win).
- core/http/endpoints/openai/realtime_gate_test.go — six specs covering nil
cfg, empty pipeline, legacy pipeline, self-contained realtime_audio,
user-pinned VAD slot, and partial legacy pipeline.
UI + endpoints
- core/http/routes/ui.go — /api/pipeline-models accepts either a legacy
VAD+STT+LLM+TTS pipeline or a realtime_audio model; surfaces a
self_contained flag so the Talk page can collapse the four cards.
- core/http/routes/ui_api.go — realtime_audio in usecaseFilters.
- core/http/routes/ui_pipeline_models_test.go — covers both code paths.
- core/http/react-ui/src/pages/Talk.jsx — self-contained badge instead of
the four-slot grid; rename Edit Pipeline → Edit Model Config; less
pipeline-specific wording.
- core/http/react-ui/src/pages/Models.jsx + locales/en/models.json — new
realtime_audio filter button + i18n.
- core/http/react-ui/src/utils/capabilities.js — CAP_REALTIME_AUDIO.
- core/http/react-ui/src/pages/FineTune.jsx — voice + validation-dataset
fields, surfaced when backend === liquid-audio, plumbed via
extra_options on submit/export/import.
Gallery + importer
- gallery/liquid-audio.yaml — config template with known_usecases:
[realtime_audio, chat, tts, transcript, vad].
- gallery/index.yaml — four model entries (realtime/chat/asr/tts) keyed by
mode option. Fixed pre-existing `transcribe` typo on the asr entry
(loader silently dropped the unknown string → entry never surfaced as a
transcript model).
- gallery/lfm.yaml — function block for the LFM2 Pythonic tool-call format
`<|tool_call_start|>[name(k="v")]<|tool_call_end|>` matching
common_chat_params_init_lfm2 in vendored llama.cpp.
- core/gallery/importers/{liquid-audio,liquid-audio_test}.go — detector
matches LFM2-Audio HF repos (excludes -gguf mirrors); mode/voice
preferences plumbed through to options.
- core/gallery/importers/importers.go — register LiquidAudioImporter
before LlamaCPPImporter.
- pkg/functions/parse_lfm2_test.go — seven specs for the response/argument
regex pair on the LFM2 pythonic format.
Build matrix
- .github/backend-matrix.yml — seven liquid-audio targets (cuda12, cuda13,
l4t-cuda-13, hipblas, intel, cpu amd64, cpu arm64). Jetpack r36 cuda-12
is skipped (Ubuntu 22.04 / Python 3.10 incompatible with liquid-audio's
3.12 floor).
- backend/index.yaml — anchor + 13 image entries.
- Makefile — .NOTPARALLEL, prepare-test-extra, test-extra,
docker-build-liquid-audio.
Docs
- .agents/plans/liquid-audio-integration.md — phased plan; PR-D (real
any-to-any wiring via AudioToAudioStream), PR-E (mid-audio tool-call
detector), PR-G (GGUF entries once upstream llama.cpp PR #18641 lands)
remain.
- .agents/api-endpoints-and-auth.md — expand the capability-surface
checklist with every place a new FLAG_* needs to be registered.
Assisted-by: claude-code:claude-opus-4-7-1m [Claude Code]
Signed-off-by: Richard Palethorpe <io@richiejp.com>
* feat(realtime): function calling + history cap for any-to-any models
Three pieces, all on the realtime_audio path that just landed:
1. liquid-audio backend (backend/python/liquid-audio/backend.py):
- _build_chat_state grows a `tools_prelude` arg.
- new _render_tools_prelude parses request.Tools (the OpenAI Chat
Completions function array realtime.go already serialises) and
emits an LFM2 `<|tool_list_start|>…<|tool_list_end|>` system turn
ahead of the user history. Mirrors gallery/lfm.yaml's `function:`
template so the model sees the same prompt shape whether served
via llama-cpp or here. Without this the backend silently dropped
tools — function calling was wired end-to-end on the Go side but
the model never saw a tool list.
2. Realtime history cap (core/http/endpoints/openai/realtime.go):
- Session grows MaxHistoryItems int; default picked by new
defaultMaxHistoryItems(cfg) — 6 for realtime_audio models (LFM2.5
1.5B degrades quickly past a handful of turns), 0/unlimited for
legacy pipelines composing larger LLMs.
- triggerResponse runs conv.Items through trimRealtimeItems before
building conversationHistory. Helper walks the cut left if it
would orphan a function_call_output, so tool result + call pairs
stay intact.
- realtime_gate_test.go: specs for defaultMaxHistoryItems and
trimRealtimeItems (zero cap, under cap, over cap, tool-call pair
preservation).
3. Talk page (core/http/react-ui/src/pages/Talk.jsx):
- Reuses the chat page's MCP plumbing — useMCPClient hook,
ClientMCPDropdown component, same auto-connect/disconnect effect
pattern. No bespoke tool registry, no new REST endpoints; tools
come from whichever MCP servers the user toggles on, exactly as
on the chat page.
- sendSessionUpdate now passes session.tools=getToolsForLLM(); the
update re-fires when the active server set changes mid-session.
- New response.function_call_arguments.done handler executes via
the hook's executeTool (which round-trips through the MCP client
SDK), then replies with conversation.item.create
{type:function_call_output} + response.create so the model
completes its turn with the tool output. Mirrors chat's
client-side agentic loop, translated to the realtime wire shape.
UI changes require a LocalAI image rebuild (Dockerfile:308-313 bakes
react-ui/dist into the runtime image). Backend.py changes can be
swapped live in /backends/<id>/backend.py + /backend/shutdown.
Assisted-by: claude-code:claude-opus-4-7-1m [Claude Code]
Signed-off-by: Richard Palethorpe <io@richiejp.com>
* feat(realtime): LocalAI Assistant ("Manage Mode") for the Talk page
Mirrors the chat-page metadata.localai_assistant flow so users can ask the
realtime model what's loaded / installed / configured. Tools are run
server-side via the same in-process MCP holder that powers the chat
modality — no transport switch, no proxy, no new wire protocol.
Wire:
- core/http/endpoints/openai/realtime.go:
- RealtimeSessionOptions{LocalAIAssistant,IsAdmin}; isCurrentUserAdmin
helper mirrors chat.go's requireAssistantAccess (no-op when auth
disabled, else requires auth.RoleAdmin).
- Session grows AssistantExecutor mcpTools.ToolExecutor.
- runRealtimeSession, when opts.LocalAIAssistant is set: gate on admin,
fail closed if DisableLocalAIAssistant or the holder has no tools,
DiscoverTools and inject into session.Tools, prepend
holder.SystemPrompt() to instructions.
- Tool-call dispatch loop: when AssistantExecutor.IsTool(name), run
ExecuteTool inproc, append a FunctionCallOutput to conv.Items, skip
the function_call_arguments client emit (the client can't execute
these — it doesn't know about them). After the loop, if any
assistant tool ran, trigger another response so the model speaks the
result. Mirrors chat's agentic loop, driven server-side rather than
via client round-trip.
- core/http/endpoints/openai/realtime_webrtc.go: RealtimeCallRequest
gains `localai_assistant` (JSON omitempty). Handshake calls
isCurrentUserAdmin and builds RealtimeSessionOptions.
- core/http/react-ui/src/pages/Talk.jsx: admin-only "Manage Mode"
checkbox under the Tools dropdown; passes localai_assistant: true to
realtimeApi.call's body, captured in the connect callback's deps.
Mirroring chat's pattern means the in-process MCP tools surface "just
works" for the Talk page without exposing a Streamable-HTTP MCP endpoint
(which was the alternative). Clients with their own MCP servers can
still use the existing ClientMCPDropdown path in parallel; the realtime
handler distinguishes them by AssistantExecutor.IsTool() at dispatch
time.
Assisted-by: claude-code:claude-opus-4-7-1m [Claude Code]
Signed-off-by: Richard Palethorpe <io@richiejp.com>
* feat(realtime): render Manage Mode tool calls in the Talk transcript
Previously the realtime endpoint only emitted response.output_item.added
for the FunctionCall item, and Talk.jsx's switch ignored the event — so
server-side tool runs were invisible in the UI. The model would speak
the result but the user had no way to see what tool was actually
called.
realtime.go: after executing an assistant tool inproc, emit a second
output_item.added/.done pair for the FunctionCallOutput item. Mirrors
the way the chat page displays tool_call + tool_result blocks.
Talk.jsx: handle both response.output_item.added and .done. Render
FunctionCall (with arguments) and FunctionCallOutput (pretty-printed
JSON when possible) as two transcript entries — `tool_call` with the
wrench icon, `tool_result` with the clipboard icon, both in mono-space
secondary-colour. Resets streamingRef after the result so the next
assistant text delta starts a fresh transcript entry instead of
appending to the previous turn.
Assisted-by: claude-code:claude-opus-4-7-1m [Claude Code]
Signed-off-by: Richard Palethorpe <io@richiejp.com>
* refactor(realtime): bound the Manage Mode tool-loop + preserve assistant tools
Fallout from a review pass on the Manage Mode patches:
- Bound the server-side agentic loop. triggerResponse used to recurse on
executedAssistantTool with no cap — a model that kept calling tools
would blow the goroutine stack. New maxAssistantToolTurns = 10 (mirrors
useChat.js's maxToolTurns). Public triggerResponse is now a thin shim
over triggerResponseAtTurn(toolTurn int); recursion increments the
counter and stops at the cap with an xlog.Warn.
- Preserve Manage Mode tools across client session.update. The handler
used to blindly overwrite session.Tools, so toggling a client MCP
server mid-session silently wiped the in-process admin tools. Session
now caches the original AssistantTools slice at session creation and
the session.update handler merges them back in (client names win on
collision — the client is explicit).
- strconv.ParseBool for the localai_assistant query param instead of
hand-rolled "1" || "true". Mirrors LocalAIAssistantFromMetadata.
- Talk.jsx: render both tool_call and tool_result on
response.output_item.done instead of splitting them across .added and
.done. The server's event pairing (added → done) stays correct; the
UI just doesn't need to inspect both phases of the same item. One
switch case instead of two, no behavioural change.
Out of scope (noted for follow-ups): extract a shared assistant-tools
helper between chat.go and realtime.go (duplication is small enough
that two parallel implementations stay readable for now), and an i18n
key for the Manage Mode helper text (Talk.jsx doesn't use i18n
anywhere else yet).
Assisted-by: claude-code:claude-opus-4-7-1m [Claude Code]
Signed-off-by: Richard Palethorpe <io@richiejp.com>
* ci(test-extra): wire liquid-audio backend smoke test
The backend ships test.py + a `make test` target and is listed in
backend-matrix.yml, so scripts/changed-backends.js already writes a
`liquid-audio=true|false` output when files under backend/python/liquid-audio/
change. The workflow just wasn't reading it.
- Expose the `liquid-audio` output on the detect-changes job
- Add a tests-liquid-audio job that runs `make` + `make test` in
backend/python/liquid-audio, gated on the per-backend detect flag
The smoke covers Health() and LoadModel(mode:finetune); fine-tune mode
short-circuits before any HuggingFace download (backend.py:192), so the
job needs neither weights nor a GPU. The full-inference path remains
gated on LIQUID_AUDIO_MODEL_ID, which CI doesn't set.
The four new Go test files (core/gallery/importers/liquid-audio_test.go,
core/http/endpoints/openai/realtime_gate_test.go,
core/http/routes/ui_pipeline_models_test.go, pkg/functions/parse_lfm2_test.go)
are already picked up by the existing test.yml workflow via `make test` →
`ginkgo -r ./pkg/... ./core/...`; their packages all carry RunSpecs entries.
Assisted-by: Claude:claude-opus-4-7
Signed-off-by: Richard Palethorpe <io@richiejp.com>
---------
Signed-off-by: Richard Palethorpe <io@richiejp.com>
Follow-up to PR #9781. v4.2.2 (run 25745181433) showed the keepalive
anchor in ci-cache wasn't enough on its own: 19 of 37 multiarch merges
still failed with "manifest not found" for the same digests we'd just
anchored.
Quay's manifest GC is per-repository. The anchor tag in ci-cache
protects the manifest copy that lives in ci-cache, but the same digest
in local-ai-backends is independently tracked and gets reaped because
nothing in local-ai-backends references it (push-by-digest=true leaves
it untagged). The merge then asks
`local-ai-backends@sha256:<digest>` and quay correctly says "not found"
in that repo, even though `ci-cache@sha256:<digest>` is alive and well.
Empirical confirmation against a live failed digest from v4.2.2:
$ docker buildx imagetools inspect quay.io/go-skynet/ci-cache@sha256:05377fe6...
Name: quay.io/go-skynet/ci-cache@sha256:05377fe6...
MediaType: application/vnd.docker.distribution.manifest.v2+json
$ docker buildx imagetools inspect quay.io/go-skynet/local-ai-backends@sha256:05377fe6...
ERROR: ... not found
Switch the source of the quay merge step to ci-cache. The blobs the
manifest references are already accessible from local-ai-backends
(verified via direct registry HEAD: HTTP 200 from both repos — the
original push cross-mounted blobs at content-addressable storage time
and they outlive the per-repo manifest GC). buildx imagetools create
republishes the manifest into local-ai-backends, then writes the
user-facing manifest list pointing at it. End state is self-contained:
the published manifest list references child manifests by digest only,
no embedded reference to ci-cache.
Dockerhub merge step is unchanged. Dockerhub's GC isn't aggressive
enough to reap untagged manifests at the timescales we operate on
(verified: localai/localai-backends@<same digest> still resolves cleanly
after >24h).
Assisted-by: Claude:claude-opus-4-7
Signed-off-by: Ettore Di Giacinto <mudler@localai.io>
* chore(llama.cpp): bump to 1ec7ba0c14f33f17e980daeeda5f35b225d41994
Picks up the upstream `spec : parallel drafting support` change
(ggml-org/llama.cpp#22838) which reshapes the speculative-decoding API
and `server_context_impl`.
Adapt the grpc-server wrapper accordingly:
* `common_params_speculative::type` (single enum) became `types`
(`std::vector<common_speculative_type>`). Update both the
"default to draft when a draft model is set" branch and the
`spec_type`/`speculative_type` option parser. The parser now also
tolerates comma-separated lists, mirroring the upstream
`common_speculative_types_from_names` semantics.
* `common_params_speculative_draft::n_ctx` is gone (draft now shares
the target context size). Keep the `draft_ctx_size` option name for
backward compatibility and ignore the value rather than failing.
* `server_context_impl::model` was renamed to `model_tgt`; update the
two reranker / model-metadata call sites.
Replaces #9763. Builds cleanly under the linux/amd64 cpu-llama-cpp
target locally.
Signed-off-by: Ettore Di Giacinto <mudler@localai.io>
* feat(llama-cpp): expose new speculative-decoding option keys
Upstream `spec : parallel drafting support` (ggml-org/llama.cpp#22838)
adds the `ngram_mod`, `ngram_map_k`, and `ngram_map_k4v` speculative
families and beefs up the draft-model knobs. The previous bump only
adapted the API; this exposes the new fields through the grpc-server
options dictionary so model configs can drive them.
New `options:` keys (all under `backend: llama-cpp`):
ngram_mod (`ngram_mod` type):
spec_ngram_mod_n_min / spec_ngram_mod_n_max / spec_ngram_mod_n_match
ngram_map_k (`ngram_map_k` type):
spec_ngram_map_k_size_n / spec_ngram_map_k_size_m / spec_ngram_map_k_min_hits
ngram_map_k4v (`ngram_map_k4v` type):
spec_ngram_map_k4v_size_n / spec_ngram_map_k4v_size_m /
spec_ngram_map_k4v_min_hits
ngram lookup caches (`ngram_cache` type):
spec_lookup_cache_static / lookup_cache_static
spec_lookup_cache_dynamic / lookup_cache_dynamic
Draft-model tuning (active when `spec_type` is `draft`):
draft_cache_type_k / spec_draft_cache_type_k
draft_cache_type_v / spec_draft_cache_type_v
draft_threads / spec_draft_threads
draft_threads_batch / spec_draft_threads_batch
draft_cpu_moe / spec_draft_cpu_moe (bool flag)
draft_n_cpu_moe / spec_draft_n_cpu_moe (first N MoE layers on CPU)
draft_override_tensor / spec_draft_override_tensor
(comma-separated <tensor regex>=<buffer type>; re-implements upstream's
static parse_tensor_buffer_overrides since it isn't exported)
`spec_type` already accepted comma-separated lists after the previous
commit, matching upstream's `common_speculative_types_from_names`.
Docs: refresh `docs/content/advanced/model-configuration.md` with
per-family tables and a note about multi-type chaining.
Builds locally with `make docker-build-llama-cpp` (linux/amd64
cpu-llama-cpp AVX variant).
Signed-off-by: Ettore Di Giacinto <mudler@localai.io>
* fix(turboquant): bridge new llama.cpp spec API to the legacy fork layout
The previous commits in this series adapted backend/cpp/llama-cpp/grpc-server.cpp
to the post-#22838 (parallel drafting) llama.cpp API. The turboquant build
reuses the same grpc-server.cpp through backend/cpp/turboquant/Makefile,
which copies it into turboquant-<flavor>-build/ and runs patch-grpc-server.sh
on the copy. The fork branched before the API refactor, so it errors out on:
* `ctx_server.impl->model_tgt` (fork still has `model`)
* `params.speculative.{ngram_mod,ngram_map_k,ngram_map_k4v,ngram_cache}.*`
(none of these sub-structs exist in the fork)
* `params.speculative.draft.{cache_type_k/v, cpuparams[, _batch].n_threads,
tensor_buft_overrides}` (fork uses the pre-#22397 flat layout)
* `params.speculative.types` vector / `common_speculative_types_from_names`
(fork has a scalar `type` and only the singular helper)
Approach:
1. backend/cpp/llama-cpp/grpc-server.cpp: introduce a single feature switch
`LOCALAI_LEGACY_LLAMA_CPP_SPEC`. When defined, the two `speculative.type[s]`
discriminations (the "default to draft when a draft model is set" branch
and the `spec_type` / `speculative_type` option parser) fall back to the
singular scalar form, and the entire new-option block (ngram_mod / map_k
/ map_k4v / ngram_cache / draft.{cache_type_*, cpuparams*,
tensor_buft_overrides}) is preprocessed out. The macro is *not* defined
in the source tree — stock llama-cpp builds get the full new API.
2. backend/cpp/turboquant/patch-grpc-server.sh: two new patch steps applied
to the per-flavor build copy at turboquant-<flavor>-build/grpc-server.cpp:
- substitute `ctx_server.impl->model_tgt` -> `ctx_server.impl->model`
- inject `#define LOCALAI_LEGACY_LLAMA_CPP_SPEC 1` before the first
`#include`, so the guarded blocks above drop out for the fork build.
Both patches are idempotent and follow the existing sed/awk pattern in
this script (KV cache types, `get_media_marker`, flat speculative
renames). Stock llama-cpp's `grpc-server.cpp` is never touched.
Drop both legacy patches once the turboquant fork rebases past
ggml-org/llama.cpp#22397 / #22838.
Signed-off-by: Ettore Di Giacinto <mudler@localai.io>
* fix(turboquant): close draft_ctx_size brace inside legacy guard
The previous turboquant fix wrapped the new option-handler blocks in
`#ifndef LOCALAI_LEGACY_LLAMA_CPP_SPEC ... #endif` but placed the guard
in the middle of an `else if` chain — the `} else if` openings of the
new blocks were responsible for closing the previous block's brace.
With the macro defined the new blocks vanish, draft_ctx_size's `{`
loses its closer, the for-loop's `}` is consumed instead, and the
file ends with a stray opening brace — clang reports it as
`function-definition is not allowed here before '{'` on the next
top-level `int main(...)` and `expected '}' at end of input`.
Move the chain split inside the draft_ctx_size branch:
} else if (... "draft_ctx_size") {
// ...
#ifdef LOCALAI_LEGACY_LLAMA_CPP_SPEC
} // legacy: chain ends here
#else
} else if (... "spec_ngram_mod_n_min") { // modern: chain continues
...
} else if (... "draft_override_tensor") {
...
} // closes last branch
#endif
} // closes for-loop
Brace count is now balanced under both preprocessor branches (verified
with `tr -cd '{' | wc -c` against the patched and unpatched outputs).
Local `make docker-build-turboquant` builds the linux/amd64 cpu-llama-cpp
`turboquant-avx` variant cleanly.
Signed-off-by: Ettore Di Giacinto <mudler@localai.io>
* fix(ci): forward AMDGPU_TARGETS into Dockerfile.turboquant builder-prebuilt
Dockerfile.turboquant's `builder-prebuilt` stage was missing the
`ARG AMDGPU_TARGETS` / `ENV AMDGPU_TARGETS=${AMDGPU_TARGETS}` pair that
`builder-fromsource` already has (and that `Dockerfile.llama-cpp`
mirrors across both stages). When CI uses the prebuilt base image
(quay.io/go-skynet/ci-cache:base-grpc-*, the common path) the build-arg
passed by the workflow never reaches the env inside the compile stage.
backend/cpp/llama-cpp/Makefile:38 (introduced by #9626) errors out on
hipblas builds when AMDGPU_TARGETS is empty, and the turboquant
Makefile reuses backend/cpp/llama-cpp via a sibling build dir, so the
same check fires from turboquant-fallback under BUILD_TYPE=hipblas:
Makefile:38: *** AMDGPU_TARGETS is empty — set it to a comma-separated
list of gfx targets e.g. gfx1100,gfx1101. Stop.
make: *** [Makefile:66: turboquant-fallback] Error 2
The bug is latent on master because the docker layer cache stays warm
across builds — the compile step rarely re-runs from scratch. The
llama.cpp bump in this PR invalidates the cache, so the missing env var
becomes load-bearing and the hipblas turboquant CI job fails.
Mirror the existing pattern from Dockerfile.llama-cpp.
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>
* ci: close the GC race + cascade-skip + darwin grpc gaps from v4.2.1
v4.2.1's backend.yml run (#25701862853) exposed three independent issues
on top of the singletons fix shipped in ea001995. Address all three plus
two related cleanups:
1. quay GC race in backend-merge-jobs-multiarch (12/37 merges failed with
"manifest not found"). Even after PR #9746 split multi/single-arch
merges, the multiarch matrix itself takes ~2h to drain at
max-parallel: 8, and the earliest per-arch digests (push-by-digest,
no tag) get reaped by quay's GC before the merge runs. The split
bounded the race for multiarch; it doesn't eliminate it. Anchor each
per-arch digest immediately to a tag in the internal ci-cache image
(`keepalive-<run_id><tag-suffix>-<platform-tag>`). Quay won't GC
tagged manifests. backend_merge.yml deletes the keepalive tags via
quay REST API after publishing the user-facing manifest list.
Cleanup is best-effort: if the quay token is not OAuth-scoped the
merge does NOT fail, the orphan tags just persist.
2. cascade-skip on backend-merge-jobs-singlearch. v4.2.1 had 2 failed
and 2 cancelled singlearch builds (out of 199); GHA's default
`needs:` semantics cascade-skipped the entire singlearch merge
matrix, so zero singleton tags were applied even though 197
singletons built successfully. Wrap the merge `if:` in
`!cancelled() && ...` for both multi and single arch in backend.yml
and backend_pr.yml so partial build failures publish the successful
tag-suffixes.
3. Darwin llama-cpp grpc-server build fails with `find_package(absl)`
not found. Same shape as the ccache/blake3/fmt/hiredis/xxhash/zstd
fix already in `Dependencies`: a brew cache hit restores
`/opt/homebrew/Cellar/grpc` so `brew install grpc` no-ops, but
abseil isn't in our Cellar cache list and never gets installed
alongside, leaving grpc's CMake unable to resolve it. Mirror the
`brew reinstall ccache` line with `brew reinstall grpc` to
re-validate grpc's full transitive dep closure on every cache-hit
run.
4. Move the four heaviest CUDA cpp builds back to bigger-runner. v4.2.1
wall-clock: -gpu-nvidia-cuda-12-llama-cpp 5h36m,
-gpu-nvidia-cuda-12-turboquant 6h05m,
-gpu-nvidia-cuda-13-llama-cpp 5h37m,
-gpu-nvidia-cuda-13-turboquant 6h05m. The cuda-12 turboquant and
cuda-13 turboquant entries are over GHA's 6h job timeout. Phase 5.3
of the free-tier migration (PR #9730) had explicitly flagged this
batch as 'highest-risk' with a per-entry revert path. All other
matrix entries (vulkan-llama-cpp ~47m, ROCm hipblas-llama-cpp ~2h,
intel sycl-f32 ~1h49m) stay on free-tier ubuntu-latest.
Verified locally: all six edited workflow YAMLs parse cleanly. Real
verification has to come from the next tag release run.
Assisted-by: Claude:claude-opus-4-7
Signed-off-by: Ettore Di Giacinto <mudler@localai.io>
* ci: extract keepalive anchor + cleanup into .github/scripts/
The two inline shell blocks from the previous commit are long enough to
hurt readability of the workflow YAML and benefit from their own files
with self-contained docs. Move them to .github/scripts/:
anchor-digest-in-cache.sh backend_build.yml's keepalive anchor
cleanup-keepalive-tags.sh backend_merge.yml's best-effort cleanup
Workflow steps reduce to a single `run:` invocation each, with all the
parameter plumbing handled by env vars on the step. backend_merge.yml
also gains a sparse `actions/checkout@v6` step (sparse to .github/scripts
only) so the cleanup script is available on the runner — backend_build
already checks out for the docker build.
Net workflow diff: -36 lines across the two files. Script logic and
behavior are byte-identical to the inline version.
Assisted-by: Claude:claude-opus-4-7
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>
Ollama's embedding endpoint accepts both `input` and `prompt` as the
input string value (see ollama/ollama docs/api.md#generate-embeddings).
LocalAI only accepted `input`, which broke client libraries that send
the `prompt` form.
Add `Prompt` to OllamaEmbedRequest and have GetInputStrings fall back
to it when Input is unset. Input still wins when both are provided.
Fixes#9767.
Assisted-by: Claude:claude-opus-4-7 [Claude Code]
Signed-off-by: Ettore Di Giacinto <mudler@localai.io>
Co-authored-by: Ettore Di Giacinto <mudler@localai.io>
* fix: parse vulkan VRAM from text
Assisted-by: opencode:gpt-5.5
Signed-off-by: Andreas Egli <github@kharan.ch>
* fix: replace string.split with streaming iteration
Assisted-by: Opencode:Gemma4
Signed-off-by: Andreas Egli <github@kharan.ch>
---------
Signed-off-by: Andreas Egli <github@kharan.ch>
