Metadata-Version: 2.5
Name: pytracer-replay
Version: 0.1.0
Summary: pyTracer replay & test engine - deterministic replay, assertions, diffing, and cross-branch classification.
Author: pyTracer
License-Expression: Apache-2.0
Requires-Python: >=3.10
Requires-Dist: pydantic>=2.6
Requires-Dist: pytracer-sdk==0.1.0
Provides-Extra: dev
Requires-Dist: mypy>=1.10; extra == 'dev'
Requires-Dist: pytest-asyncio>=0.23; extra == 'dev'
Requires-Dist: pytest>=8.0; extra == 'dev'
Requires-Dist: ruff>=0.5; extra == 'dev'
Description-Content-Type: text/markdown

# pytracer-replay

Layer 4 of pyTracer: reconstructs a past agent run deterministically, injects
recorded tool/model responses instead of live calls, runs assertions, diffs two
runs, and classifies a cross-branch reproduction as fixed / still failing /
diverged.

```python
from pytracer_replay import MockSet, ReplayEngine, ReplayMode, classify
import pytracer_sdk as pytracer

# 1. Build a mock set from a captured trace (its recorded tool/model outputs).
mocks = MockSet.from_trace(recorded_spans)

# 2. Replay the agent deterministically — tools and model are mocked.
engine = ReplayEngine(mode=ReplayMode.FULL)
result = engine.replay(lambda: agent_entrypoint(task), mocks)

# 3. Assert, diff, fingerprint, or classify against another branch's run.
```

Modes: `FULL` (recorded model + tools), `PARTIAL` (recorded tools, live model),
`LIVE` (both live, for drift detection).
