Langfuse LLM observability for DeepSeek Harness
TtTRz/dsh-langfuse · productivity
Langfuse LLM observability for DeepSeek Harness — one OpenTelemetry trace tree per agent session, with feedback scores and subagent lineage.
★ 2GitHub stars
0Forks
2026-08-15Last updated
TypeScriptLanguage
MITLicense
Key features
- Full session tracing — every turn, generation, and tool call lands in Langfuse with model, provider, usage (incl.
- The harness logs everything model-visible to the canonical session log
- cache-read & reasoning tokens), and TTFT.
- Content-Length transport — spans are delivered as a single write with an explicit Content-Length header, never chunked.
- Some Langfuse deployments sit behind gateways that mangle chunked POST bodies (400 Failed to parse OTel JSON Trace)
Requirements
- One backend per context (the seam throws on duplicates — the bundled patch disables the official OTLP-logs row).
- Subagent lineage is same-process best-effort (a resumed parent's trace id is unknown across restarts).
Install command
dsh plugin --profile web add github:TtTRz/dsh-langfuse