Observability
Langfuse tracing, trace viewer, chat forking, prompt lab, and split-screen comparison.
README
Langfuse Observability
Agent Zero plugin for LLM observability via Langfuse.
Features
- Tracing — Automatic tracing of LLM calls, tool executions, and agent interactions
- Trace Viewer — Browse and inspect traces from the Agent Zero sidebar
- Chat Forking — Fork any conversation at any point to explore alternatives
- Prompt Lab — Test and refine prompts with judge-based evaluation
- Split View — Side-by-side comparison of chat responses
Installation
Clone or download this repo into your Agent Zero plugins directory:
git clone <repo-url> /path/to/agent-zero/usr/plugins/langfuse-observability
Configuration
Configure via the Agent Zero settings UI (Agent tab > Observability section), or set environment variables:
LANGFUSE_PUBLIC_KEY— Your Langfuse project public keyLANGFUSE_SECRET_KEY— Your Langfuse project secret keyLANGFUSE_HOST— Langfuse instance URL (default:https://cloud.langfuse.com)
Requirements
The langfuse Python package is auto-installed on first use.