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.
41 lines
1.3 KiB
Go
41 lines
1.3 KiB
Go
package langfuse
|
|
|
|
import (
|
|
"context"
|
|
|
|
"github.com/Tencent/WeKnora/internal/types"
|
|
)
|
|
|
|
// traceCtxKey is the exported context key defined in types/const.go. It lives
|
|
// there (not inside this package) so that logger.CloneContext — which rebuilds
|
|
// a stripped-down context on every request — can preserve the Langfuse trace
|
|
// without importing this package. If we kept the key private here, every
|
|
// CloneContext call would drop the trace and downstream LLM wrappers would
|
|
// each auto-create their own shallow trace, fragmenting a single HTTP request
|
|
// into many unrelated traces in the Langfuse UI.
|
|
var traceCtxKey = types.LangfuseTraceContextKey
|
|
|
|
// withTrace stores a *Trace on the context so downstream LLM wrappers can
|
|
// attach their generations to it.
|
|
func withTrace(ctx context.Context, t *Trace) context.Context {
|
|
if t == nil || ctx == nil {
|
|
return ctx
|
|
}
|
|
return context.WithValue(ctx, traceCtxKey, t)
|
|
}
|
|
|
|
// traceFromCtx retrieves the active trace, if any.
|
|
func traceFromCtx(ctx context.Context) (*Trace, bool) {
|
|
if ctx == nil {
|
|
return nil, false
|
|
}
|
|
t, ok := ctx.Value(traceCtxKey).(*Trace)
|
|
return t, ok && t != nil
|
|
}
|
|
|
|
// TraceFromContext is the public accessor used by HTTP middlewares and
|
|
// handlers that want to set the trace input/output on the active trace.
|
|
func TraceFromContext(ctx context.Context) (*Trace, bool) {
|
|
return traceFromCtx(ctx)
|
|
}
|