Langfuse observability — every LLM call traced and scored

Langfuse v2.95.11 running on :3001 (Docker + Postgres).
Login: j@lakehouse.local / lakehouse2026

tracing.ts: startTrace → logGeneration/logRetrieval/logSpan → scoreTrace → flush.
Every hybrid search, SQL generation, RAG pipeline, and co-pilot
briefing gets a full trace: model, prompt, output, latency, tokens.

The observer can now score traces based on verification results —
Langfuse aggregates accuracy over time so we can see which models
and approaches actually work in production, not just in tests.

Services: lakehouse(:3100) + sidecar(:3200) + agent(:3700) +
observer + langfuse(:3001) + minio(:9000) + mariadb(:3306)

Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>

mcp-server/tracing.ts

@@ -0,0 +1,52 @@
+/**
+ * Langfuse tracing for the Lakehouse agent gateway.
+ *
+ * Every LLM call gets traced: model, prompt, output, latency, tokens.
+ * The observer uses this to build a picture of what's working.
+ *
+ * Langfuse UI: http://localhost:3001
+ * Login: j@lakehouse.local / lakehouse2026
+ */
+
+import { Langfuse } from "langfuse";
+
+const langfuse = new Langfuse({
+  publicKey: process.env.LANGFUSE_PUBLIC_KEY || "pk-lf-staffing",
+  secretKey: process.env.LANGFUSE_SECRET_KEY || "sk-lf-staffing-secret",
+  baseUrl: process.env.LANGFUSE_URL || "http://localhost:3001",
+  enabled: true,
+});
+
+export type TraceContext = ReturnType<typeof langfuse.trace>;
+
+export function startTrace(name: string, input?: any, metadata?: any) {
+  return langfuse.trace({ name, input, metadata });
+}
+
+export function logGeneration(
+  trace: TraceContext, name: string,
+  opts: { model: string; prompt: string; completion: string;
+          duration_ms: number; tokens_in?: number; tokens_out?: number; },
+) {
+  trace.generation({
+    name, model: opts.model, input: opts.prompt, output: opts.completion,
+    usage: { promptTokens: opts.tokens_in, completionTokens: opts.tokens_out },
+    metadata: { duration_ms: opts.duration_ms },
+  });
+}
+
+export function logSpan(trace: TraceContext, name: string, input: any, output: any, duration_ms: number) {
+  trace.span({ name, input, output, metadata: { duration_ms } });
+}
+
+export function logRetrieval(trace: TraceContext, name: string, query: string, results: any[], duration_ms: number) {
+  trace.span({ name, input: { query }, output: { results_count: results.length }, metadata: { duration_ms, type: "retrieval" } });
+}
+
+export function scoreTrace(trace: TraceContext, name: string, value: number, comment?: string) {
+  trace.score({ name, value, comment });
+}
+
+export async function flush() { await langfuse.flushAsync(); }
+export async function shutdown() { await langfuse.shutdownAsync(); }
+export { langfuse };