Description

LangfuseProvider multi-framework support + Langfuse v4 compatibility

• Updated LangfuseProvider to support newer version of langfuse SDK since the callback feature is added after v4
• Makes LangfuseProvider work with traces from any framework (LangChain, Strands, LlamaIndex, CrewAI, etc.) — not just Strands — and adds support for the Langfuse v4 SDK.

Previously, LangfuseProvider only handled Strands-specific observation names ("execute_tool", "invoke_agent"). Traces from LangChain or other frameworks were silently skipped. This PR makes the provider framework-agnostic by routing on obs.type (universal across all frameworks in Langfuse) with a Strands-specific fallback for backward compatibility.

Changes

• langfuse_provider.py — Multi-framework observation routing + Langfuse v4 support:
  • _convert_observation(): Route by universal obs.type (GENERATION → InferenceSpan, TOOL → ToolExecutionSpan, CHAIN/AGENT root → AgentInvocationSpan) with Strands SPAN name-based fallback preserved
  • _convert_tool_execution(): Handle both Strands format (obs.input.name) and LangChain/universal format (obs.name + obs.input as arguments)
  • _parse_tool_result(): Support LangChain ToolMessage format ({"content": "...", "type": "tool"})
  • _convert_message(): Parse LangChain tool_calls field on assistant messages
  • _extract_user_prompt(): Handle LangChain CHAIN formats ({"input": "..."}, {"messages": [...]}) and Strands message-list format via OTEL ([{"role": "user", "content": '[{"text": "..."}]'}])
  • _extract_agent_response(): Handle LangChain CHAIN output formats ({"output": "..."}, {"messages": [...]}) and Strands message-list output format
  • _extract_text_from_content(): New helper to unwrap Strands content (plain string, JSON-encoded [{"text": "..."}], or parsed list)
  • _fetch_observations(): Use legacy.observations_v1.get_many() on Langfuse v4+ (v4's new endpoint returns lightweight objects missing input/output/name)
  • _fetch_all_pages(): Auto-detect page-based (v2/v3: meta.total_pages) vs cursor-based (v4+: meta.cursor) pagination
• pyproject.toml: Bump langfuse dependency from <3 to <5 to support v4 SDK

Backward compatibility

• Strands traces: the SPAN + obs.name fallback is preserved — existing Strands→Langfuse flows work unchanged
• Langfuse v2/v3 SDK: page-based pagination and observations.get_many() paths are preserved
• All 27 pre-existing tests continue to pass

Related Issues

#91
#159

Documentation PR

N/A

Type of Change

New feature

Testing

How have you tested the change? Verify that the changes do not break functionality or introduce warnings in consuming repositories: agents-docs, agents-tools, agents-cli

• I ran hatch run prepare

Checklist

• I have read the CONTRIBUTING document
• I have added any necessary tests that prove my fix is effective or my feature works
• I have updated the documentation accordingly
• I have added an appropriate example to the documentation to outline the feature, or no new docs are needed
• My changes generate no new warnings
• Any dependent changes have been merged and published

By submitting this pull request, I confirm that you can use, modify, copy, and redistribute this contribution, under the terms of your choice.