docs: add Rewind to ecosystem integrations

[risjai]

Adds Rewind to the ecosystem integrations list in the tracing documentation.

What is Rewind?

Rewind is a time-travel debugger for AI agents. It records every LLM call, tool execution, and handoff — then lets you:

• Fork timelines at any step
• Replay from failure — cached steps served instantly (0 tokens, 0ms), only the fixed step re-runs live
• Diff original vs replayed timelines
• Regression test with rewind assert in CI

Integration

Zero config — auto-registers a TracingProcessor via add_trace_processor():

import rewind_agent
from agents import Agent, Runner

rewind_agent.init()  # auto-detects Agents SDK, registers tracing

agent = Agent(name="researcher", instructions="...", tools=[web_search])
result = await Runner.run(agent, "What is the population of Tokyo?")
# rewind show latest → full trace with agent names, tool calls

Install: pip install rewind-agent[agents]

Links

• Integration docs: https://github.com/agentoptics/rewind/blob/master/docs/openai-agents-sdk.md
• GitHub: https://github.com/agentoptics/rewind
• PyPI: https://pypi.org/project/rewind-agent/

[seratch]

Thanks for sharing this. However, the page lists only the integrations that support tracing interface and your one does not have the functionality. If you build something with tracing, please feel free to create a new PR.

[risjai]

Thanks for the feedback! We've updated our integration to properly implement the tracing interface:

• RewindTracingProcessor now formally subclasses agents.tracing.TracingProcessor
• Verified: issubclass(RewindTracingProcessor, TracingProcessor) == True
• Registers via add_trace_processor() — either auto-detected through rewind_agent.init() or manually:

from agents.tracing import add_trace_processor
from rewind_agent.openai_agents import RewindTracingProcessor
from rewind_agent.store import Store

store = Store()
session_id, timeline_id = store.create_session("my-agent")
processor = RewindTracingProcessor(store, session_id, timeline_id)
add_trace_processor(processor)

Maps GenerationSpanData → llm_call steps (with model, tokens from usage), FunctionSpanData → tool_call steps, HandoffSpanData → handoff steps.

Updated integration docs: https://github.com/agentoptics/rewind/blob/master/docs/openai-agents-sdk.md

Published on PyPI: pip install rewind-agent[agents]