wizardchen
492e92580b
Closes #620 #497. Add opt-in Langfuse observability covering all five model types (chat, embedding, rerank, VLM, ASR) with HTTP-request-scoped traces and Docker Compose support (both cloud and self-hosted). Core package internal/tracing/langfuse: - HTTP client with batched async ingestion (non-blocking in request path) - Sampling, environment / release tagging, and graceful fallback when LANGFUSE_* env vars are absent (wrappers become no-ops) - Gin middleware opens one trace per traced request and finishes it after the handler chain returns, attaching method / path / user / session - Trace context is stored under a typed key exported from internal/types so logger.CloneContext can preserve it across handler / goroutine boundaries (otherwise each LLM call auto-created an orphan trace, fragmenting one request into many) Per-model generation wrappers (opt-in via NewChat/NewEmbedder/...): - chat: captures prompt, streaming output, token usage + TTFT - embedding: approximates tokens when the provider omits usage - rerank: previews query/docs, summarizes results to keep payload small - vlm: records image count and total bytes, never uploads raw pixels - asr: records file size and audio duration, never uploads audio bytes Async title generation (GenerateTitleAsync) now forwards the trace key into the goroutine so title calls appear under the parent chat trace. Docker Compose: - LANGFUSE_* env passthrough on the `app` service for cloud deployments - Optional `langfuse` profile spins up a self-hosted Langfuse stack that reuses WeKnora's existing PostgreSQL (separate database via an idempotent init container that fixes ICU collation drift) and Redis (separate DB number), adding only ClickHouse, MinIO, web and worker containers - web/worker entrypoints URL-encode DB_PASSWORD / REDIS_PASSWORD at start to avoid Prisma P1013 when passwords contain @ / # / etc. Docs: docs/Langfuse集成.md covers cloud vs self-hosted, per-model usage strategy, code map, and resource footprint.
73 lines
2.0 KiB
Go
73 lines
2.0 KiB
Go
package vlm
|
|
|
|
import (
|
|
"context"
|
|
|
|
"github.com/Tencent/WeKnora/internal/tracing/langfuse"
|
|
)
|
|
|
|
// langfuseVLM wraps a VLM and reports each Predict call as a Langfuse
|
|
// generation. The raw image bytes are NOT uploaded — Langfuse traces are
|
|
// designed for text. We include image count and total byte size in the
|
|
// metadata, plus the text prompt, which matches how Langfuse's own VLM
|
|
// integrations report multimodal calls.
|
|
type langfuseVLM struct {
|
|
inner VLM
|
|
}
|
|
|
|
func (l *langfuseVLM) GetModelName() string { return l.inner.GetModelName() }
|
|
func (l *langfuseVLM) GetModelID() string { return l.inner.GetModelID() }
|
|
|
|
func (l *langfuseVLM) Predict(ctx context.Context, imgBytes [][]byte, prompt string) (string, error) {
|
|
mgr := langfuse.GetManager()
|
|
if !mgr.Enabled() {
|
|
return l.inner.Predict(ctx, imgBytes, prompt)
|
|
}
|
|
|
|
totalImgSize := 0
|
|
for _, b := range imgBytes {
|
|
totalImgSize += len(b)
|
|
}
|
|
|
|
genCtx, gen := mgr.StartGeneration(ctx, langfuse.GenerationOptions{
|
|
Name: "vlm.predict",
|
|
Model: l.inner.GetModelName(),
|
|
Input: map[string]interface{}{
|
|
"prompt": prompt,
|
|
"image_count": len(imgBytes),
|
|
},
|
|
Metadata: map[string]interface{}{
|
|
"model_id": l.inner.GetModelID(),
|
|
"image_count": len(imgBytes),
|
|
"image_bytes_total": totalImgSize,
|
|
},
|
|
})
|
|
|
|
result, err := l.inner.Predict(genCtx, imgBytes, prompt)
|
|
|
|
// VLMs don't return token usage; approximate prompt tokens for cost
|
|
// tracking in Langfuse (users can configure per-model pricing in the UI).
|
|
promptTokens := len([]rune(prompt))/4 + 1
|
|
outputTokens := len([]rune(result)) / 4
|
|
usage := &langfuse.TokenUsage{
|
|
Input: promptTokens,
|
|
Output: outputTokens,
|
|
Total: promptTokens + outputTokens,
|
|
Unit: "TOKENS",
|
|
}
|
|
|
|
gen.Finish(result, usage, err)
|
|
return result, err
|
|
}
|
|
|
|
// wrapVLMLangfuse applies the Langfuse decorator when the manager is enabled.
|
|
func wrapVLMLangfuse(v VLM, err error) (VLM, error) {
|
|
if err != nil || v == nil {
|
|
return v, err
|
|
}
|
|
if !langfuse.GetManager().Enabled() {
|
|
return v, nil
|
|
}
|
|
return &langfuseVLM{inner: v}, nil
|
|
}
|