-
released this
2026-02-10 01:52:20 +00:00 | 8 commits to main since this releaseAgentLens v0.1.0
Agent observability that traces decisions, not just API calls. See why your AI agents chose what they chose.
Install
pip install vectry-agentlensQuick Start
import agentlens agentlens.init(api_key="your-key", endpoint="https://agentlens.vectry.tech") with agentlens.trace("my-agent-task"): agentlens.log_decision( type="TOOL_SELECTION", chosen={"name": "search_web", "confidence": 0.92}, alternatives=[{"name": "search_docs", "reason_rejected": "query too broad"}], reasoning="User query requires real-time data" )What's Included
Python SDK
- Decision tracing with 7 decision types (TOOL_SELECTION, ROUTING, PLANNING, RETRY, ESCALATION, MEMORY_RETRIEVAL, CUSTOM)
- OpenAI auto-instrumentation:
wrap_openai(client) - LangChain callback handler
- Nested trace contexts for multi-agent workflows
- Async batch transport with automatic flushing
Dashboard
- Real-time trace streaming via SSE
- Decision tree visualization (React Flow + Dagre)
- Analytics: token usage, cost tracking, duration timelines
- Advanced filtering: status, tags, date range, full-text search
- Trace detail with tabs: Tree, Analytics, Decisions, Spans, Events
Infrastructure
- Docker Compose one-command deployment
- PostgreSQL 16 + Redis 7
- Health checks on all services
- Memory limits and log rotation
Links
- Live demo: https://agentlens.vectry.tech
- PyPI: https://pypi.org/project/vectry-agentlens/
- Documentation: See README.md and packages/sdk-python/README.md
- Examples: See examples/ directory
Downloads