Files
LocalAI/core/trace/backend_trace.go
Richard Palethorpe 5d0c43ec6e feat(realtime): Semantic VAD EOU token (#10444)
* feat(realtime): EOU-driven semantic_vad turn detection

Add a `semantic_vad` turn-detection mode to the realtime API that feeds
the transcription model live and decides "the user finished speaking"
from the `<EOU>` end-of-utterance token rather than from silence alone.
When EOU fires the turn commits immediately (~0.3s); otherwise it falls
back to an eagerness-scaled silence threshold (low/med/high = 8/4/2s).

Plumbing, bottom to top:

- proto: `AudioTranscriptionLive` bidirectional RPC (config-first oneof,
  mono float PCM @16k, ready-ack / Unimplemented degrade signal) plus
  `TranscriptResult.eou` for the unary retranscribe gate.
- pkg/grpc: client/server/base/embed scaffolding for the bidi stream,
  modeled on AudioTransformStream; release stream conns on terminal Recv.
- parakeet-cpp: live transcription RPC with per-C-call engine locking
  (one live stream per turn, finalize+free at commit); bump parakeet.cpp
  to ABI v5 — incremental StreamingMel (no more quadratic per-feed mel
  recompute that delayed EOU on long turns) and the <EOU>/<EOB> split;
  strip the literal <EOU>/<EOB> from offline text and set Eou.
- core/backend: LiveTranscriptionSession wrapper + pipeline
  `turn_detection:` config block (type/eagerness/retranscribe).
- realtime: semantic_vad integration — live input captions streamed as
  transcription deltas while the user speaks, EOU-immediate commit with
  eagerness fallback, optional retranscribe gate (batch re-decode must
  also end in <EOU> to confirm), clause synthesis off the LLM token
  callback, and per-turn live-transcription / model_load telemetry.
- UI: show the realtime pipeline components as a vertical list.

Docs and tests included; opt-in via the pipeline YAML or per-session
`session.update`. Non-streaming STT backends degrade to silence-only.

Assisted-by: Claude Code:claude-opus-4-8 [Read] [Edit] [Write] [Bash]
Assisted-by: Claude Code:claude-fable-5 [Read] [Edit] [Bash]
Signed-off-by: Richard Palethorpe <io@richiejp.com>

* feat(realtime): explicit formally-verified state machines + parakeet streaming driver

The realtime API had several implicit state machines whose state was inferred
from scattered booleans, channels, and five separate mutexes, leaving
illegal/inconsistent states reachable. Make them explicit and keep the
implementation in step with a formal design; rework the parakeet streaming
backend along the same lines.

Realtime state machines (M1-M5). Each is a sealed sum-type State/Event/Effect
with a total, pure Next(state,event)->(state,[]effect) behind a single-writer
Coordinator:

  M1 conncoord    connection lifecycle: VAD toggle + once-only teardown
                  (replaces vadServerStarted + a `done` channel closed from
                  two sites).
  M2 turncoord    turn detection: collapses speechStarted and the live-stream
                  "turn open" flag into one state, so discardTurn can no longer
                  desync them and suppress the next onset.
  M3 respcoord    response coordination: serializes the dual-writer
                  start/cancel so at most one response is live; one
                  response.done per response.create.
  M4 compactcoord conversation compaction: single-flight (replaces the
                  `compacting atomic.Bool` CAS).
  M5 ttscoord     TTS pipeline: open->closing->closed, idempotent wait(),
                  rejects enqueue-after-close (was a silent drop).

The Coordinator/Sink/Next plumbing — only the sealed types and Next differed
per machine — is extracted once into core/http/endpoints/openai/coordinator as
a generic Coordinator[S,E,F]; each machine keeps its public API via type
aliases, so no sink, call-site, or test moved.

Hierarchy. session_lifecycle.fizz models M1 as the parent region with its
children (M2/M3/M4) as one statechart and asserts ChildrenDieWithParent (conn
torn => all children terminal, none start after teardown). respcoord and
compactcoord gain an absorbing Terminated state + Shutdown event; conncoord's
teardown drives the children terminal. This closes a compaction teardown gap: a
fire-and-forget compaction could outlive a torn session — compactionSink now
takes a session-scoped cancellable context + WaitGroup and joins the in-flight
summarize+evict on shutdown.

Formal verification. formal-verification/ holds one authoritative FizzBee spec
per machine plus the composition spec, each with an always-assertion and a
documented one-line edit that makes the checker fail (verified non-vacuous).
scripts/realtime-conformance.sh is fail-closed: all Go conformance suites under
-race AND a model-check of every .fizz spec; a missing FizzBee is a hard error
(only the loud REALTIME_CONFORMANCE_SKIP_FIZZBEE=1 bypasses it, never in CI).
FizzBee is pinned by sha256 and installed via scripts/install-fizzbee.sh into
.tools/ (gitignored). Wired as make test-realtime-conformance, a CI workflow,
and a pre-commit path filter. Go conformance tests are Ginkgo/Gomega (per the
repo's forbidigo lint): transition tables + fixed-seed property walks +
concurrent/-race specs, no rapid dependency. Design map:
docs/design/realtime-state-machines.md.

Parakeet streaming backend. The same treatment applied to the parakeet-cpp
streaming paths:
- AudioTranscriptionStream returns codes.Unimplemented for non-streaming models
  instead of decoding offline and emitting it as one delta + final. A client
  that asked for streaming learns the model cannot stream rather than receiving
  a batch result shaped like a stream. New grpcerrors.StreamTranscriptionUnsupported
  carries that signal; the HTTP /v1/audio/transcriptions stream path surfaces it
  as an SSE error event. Mirrors AudioTranscriptionLive, which already did this.
- utteranceBoundary (boundary.go): a single definition of the end-of-utterance
  latch, replacing three open-coded finalEou toggles. Modelled as a two-valued
  type so illegal states are unrepresentable.
- Shared decode driver (driver.go): streamFeedResult (one per-feed event) +
  feedChunk (hides the ABI v4 JSON vs text-only split) + feedSlices + flushTail.
  The feed loop is written once.
- AudioTranscriptionLive becomes a bidi adapter: it streams the per-feed
  {delta,eou,eob,words} the realtime turn detector consumes and a terminal
  FinalResult carrying only Text. Segments/duration/eou are offline-only and no
  longer produced (nor read) on the live path; liveTraceState drops the terminal
  eou and keeps the per-feed eou_events count.
- AudioTranscriptionStream + streamJSON merge into one driver-based function;
  streamSegmenter is generalized to the unified event with a text-only fallback
  that preserves the legacy (no-words) library's per-utterance segmentation.

Verified: build/vet/gofumpt clean, golangci-lint 0 issues, all coordinator and
parakeet packages under -race, the fail-closed conformance gate green, and
make test-realtime (12 e2e WS+WebRTC).

Assisted-by: Claude:claude-opus-4-8 [Claude Code]
Signed-off-by: Richard Palethorpe <io@richiejp.com>

---------

Signed-off-by: Richard Palethorpe <io@richiejp.com>
2026-06-30 09:01:22 +02:00

247 lines
7.9 KiB
Go

package trace
import (
"encoding/json"
"fmt"
"slices"
"sync"
"time"
"github.com/emirpasic/gods/v2/queues/circularbuffer"
"github.com/mudler/LocalAI/core/schema"
"github.com/mudler/xlog"
)
type BackendTraceType string
const (
BackendTraceLLM BackendTraceType = "llm"
BackendTraceEmbedding BackendTraceType = "embedding"
BackendTraceTranscription BackendTraceType = "transcription"
BackendTraceImageGeneration BackendTraceType = "image_generation"
BackendTraceVideoGeneration BackendTraceType = "video_generation"
BackendTraceTTS BackendTraceType = "tts"
BackendTraceSoundGeneration BackendTraceType = "sound_generation"
BackendTraceRerank BackendTraceType = "rerank"
BackendTraceTokenize BackendTraceType = "tokenize"
BackendTraceDetection BackendTraceType = "detection"
BackendTraceDepth BackendTraceType = "depth"
BackendTraceFaceVerify BackendTraceType = "face_verify"
BackendTraceFaceAnalyze BackendTraceType = "face_analyze"
BackendTraceVoiceVerify BackendTraceType = "voice_verify"
BackendTraceVoiceAnalyze BackendTraceType = "voice_analyze"
BackendTraceVoiceEmbed BackendTraceType = "voice_embed"
BackendTraceAudioTransform BackendTraceType = "audio_transform"
BackendTraceModelLoad BackendTraceType = "model_load"
BackendTraceScore BackendTraceType = "score"
BackendTraceTokenClassify BackendTraceType = "token_classify"
BackendTracePatternPII BackendTraceType = "pattern_pii"
BackendTraceVectorStore BackendTraceType = "vector_store"
)
type BackendTrace struct {
Timestamp time.Time `json:"timestamp"`
Duration time.Duration `json:"duration"`
Type BackendTraceType `json:"type"`
ModelName string `json:"model_name"`
Backend string `json:"backend"`
Summary string `json:"summary"`
// Body is the full request payload sent to the backend, when one
// applies (currently: cloud-proxy passthrough forwards). Summary
// is a short preview for the trace list; Body is the full
// payload shown when the row is expanded. Capped by the recorder
// to keep the in-memory ring buffer bounded.
Body string `json:"body,omitempty"`
Error string `json:"error,omitempty"`
Data map[string]any `json:"data"`
}
// MaxTraceBodyBytes caps the per-trace stored request body. Roomy
// enough to keep typical chat histories intact while preventing a
// runaway buffer when a caller streams MB-scale payloads.
const MaxTraceBodyBytes = 1 << 20
var (
backendTraceBuffer *circularbuffer.Queue[*BackendTrace]
backendMu sync.Mutex
backendLogChan = make(chan *BackendTrace, 100)
backendInitOnce sync.Once
)
// backendMaxBodyBytes caps each captured string value in a BackendTrace.Data
// field to keep the /api/backend-traces JSON small enough for the admin UI to
// load on every 5s auto-refresh. Mirrors the API-trace body cap added in
// commit 61bf34ea: without it a chatty LLM workload (full message history per
// trace) or any TTS run (~1.3 MiB of audio_wav_base64 per trace) blows the
// payload past tens of MiB and locks the Traces page in a loading state.
//
// 0 disables the cap. Guarded by backendMu; refreshed on EVERY
// InitBackendTracingIfEnabled call — see below.
var backendMaxBodyBytes int
func InitBackendTracingIfEnabled(maxItems, maxBodyBytes int) {
backendInitOnce.Do(func() {
if maxItems <= 0 {
maxItems = 100
}
backendMu.Lock()
backendTraceBuffer = circularbuffer.New[*BackendTrace](maxItems)
backendMu.Unlock()
go func() {
for t := range backendLogChan {
backendMu.Lock()
if backendTraceBuffer != nil {
backendTraceBuffer.Enqueue(t)
}
backendMu.Unlock()
}
}()
})
// The body cap tracks the LATEST call, not the first: tracing_max_body_bytes
// is runtime-mutable via the settings API (ApplyRuntimeSettings), and every
// recording path calls this right before RecordBackendTrace with the current
// appConfig value. Freezing the cap on first init meant a raised setting let
// producers (e.g. trace.AudioSnippet, which reads the live value) embed
// payloads that this recorder then stomped with the "<truncated: N bytes>"
// marker — corrupting audio_wav_base64 into an unplayable string. maxItems
// keeps first-call semantics: resizing the ring buffer would drop entries.
backendMu.Lock()
backendMaxBodyBytes = maxBodyBytes
backendMu.Unlock()
}
func RecordBackendTrace(t BackendTrace) {
backendMu.Lock()
maxBody := backendMaxBodyBytes
backendMu.Unlock()
if t.Data != nil && maxBody > 0 {
t.Data = capDataStrings(t.Data, maxBody)
}
select {
case backendLogChan <- &t:
default:
xlog.Warn("Backend trace channel full, dropping trace")
}
}
// capDataStrings walks a trace Data map and replaces any string value (at any
// depth) that exceeds maxBytes with a fixed-size marker that names the
// original byte count. The replacement is intentionally short and not valid
// base64/JSON: the goal is to flag "this was dropped" cheaply, not to keep a
// partial value that the UI might try to render. Non-string scalars and
// non-map containers pass through untouched so structural fields like
// total_deltas or audio_sample_rate remain useful.
func capDataStrings(data map[string]any, maxBytes int) map[string]any {
out := make(map[string]any, len(data))
for k, v := range data {
out[k] = capValue(v, maxBytes)
}
return out
}
func capValue(v any, maxBytes int) any {
switch val := v.(type) {
case string:
if len(val) > maxBytes {
return fmt.Sprintf("<truncated: %d bytes>", len(val))
}
return val
case map[string]any:
return capDataStrings(val, maxBytes)
default:
return v
}
}
func GetBackendTraces() []BackendTrace {
backendMu.Lock()
if backendTraceBuffer == nil {
backendMu.Unlock()
return []BackendTrace{}
}
ptrs := backendTraceBuffer.Values()
backendMu.Unlock()
traces := make([]BackendTrace, len(ptrs))
for i, p := range ptrs {
traces[i] = *p
}
slices.SortFunc(traces, func(a, b BackendTrace) int {
return b.Timestamp.Compare(a.Timestamp)
})
return traces
}
func ClearBackendTraces() {
backendMu.Lock()
if backendTraceBuffer != nil {
backendTraceBuffer.Clear()
}
backendMu.Unlock()
}
func GenerateLLMSummary(messages schema.Messages, prompt string) string {
if len(messages) > 0 {
last := messages[len(messages)-1]
text := ""
switch content := last.Content.(type) {
case string:
text = content
default:
b, err := json.Marshal(content)
if err == nil {
text = string(b)
}
}
if text != "" {
return TruncateString(text, 200)
}
}
if prompt != "" {
return TruncateString(prompt, 200)
}
return ""
}
func TruncateString(s string, maxLen int) string {
if len(s) <= maxLen {
return s
}
return s[:maxLen] + "..."
}
// TruncateToBytes caps a string at exactly maxBytes, preserving the leading
// content and appending a marker so the UI knows the value was clipped.
// Unlike TruncateString it guarantees output <= maxBytes, which matters for
// fields that feed back into the trace pipeline: capDataStrings in
// RecordBackendTrace re-checks size and would otherwise replace a producer's
// head-preserving truncation with the bare marker, losing the prefix.
//
// maxBytes <= 0 disables the cap, matching backendMaxBodyBytes semantics.
func TruncateToBytes(s string, maxBytes int) string {
if maxBytes <= 0 || len(s) <= maxBytes {
return s
}
suffix := fmt.Sprintf("...[truncated, %d bytes]", len(s))
if len(suffix) >= maxBytes {
// Pathologically small caps can't fit the marker; fall back to a
// hard cut so the contract (output <= maxBytes) still holds.
return s[:maxBytes]
}
return s[:maxBytes-len(suffix)] + suffix
}
// TruncateBytes is the []byte counterpart of TruncateString — it copies
// at most maxLen bytes, avoiding a full string([]byte) allocation when
// the input is a large request body.
func TruncateBytes(b []byte, maxLen int) string {
if len(b) <= maxLen {
return string(b)
}
return string(b[:maxLen]) + "..."
}