v0.2.0 — Server-Sent Events (SSE) Streaming Token Replayer

🚀 What's New in v0.2.0

🌊 Server-Sent Events (SSE) Streaming Token Replayer (#13)

Deterministic recording and offline replay for streaming LLM responses:

• Automatic Stream Detection: Detects content-type: text/event-stream responses automatically.
• Chunk-by-Chunk Replay: Records sequential streaming chunks into cassettes and replays them through custom synchronous and asynchronous byte streams without repeat network calls.
• Full Async & Sync Support: Works seamlessly in both CassetteTransport and AsyncCassetteTransport.
• OpenAI Python SDK Integration: Offline support for client.chat.completions.create(..., stream=True) with 0 outbound calls and $0 token cost during replay.
• Configurable Replay Latency: Optional chunk_delay_ms parameter on transports to simulate realistic token generation latency in UI and integration tests.

👥 Contributors

Special thanks to @abhandary14 for contributing PR #14! 👏

📦 Installation

pip install --upgrade rerunner

v0.1.2 — Google GenAI Support & OpenTelemetry Trace Exporter

🚀 What's New in v0.1.2

🌐 Google GenAI & Vertex AI 1-Line Adapter

• Seamless integration with Google Cloud Gemini via wrap_google_genai(), wrap_vertexai(), and wrap_gemini().
• Drop-in deterministic replay support for multi-step Gemini agents in CI.

🔭 OpenTelemetry (OTel) Distributed Trace Exporter

• Export recorded agent flight cassettes to standard OpenTelemetry OTLP trace JSON:

  rerunner export-otel session.jsonl --output session.otel.json --service-name "agent-service"

v0.1.1: Compressed Cassettes & Enterprise CI Caching

Release v0.1.1: Compressed Cassettes & Enterprise CI Caching

Rerunner v0.1.1 brings enterprise-grade storage efficiency and continuous integration recipes to make AI agent regression testing faster, deterministic, and storage-friendly.

✨ What's New

1. Transparent Gzip-Compressed Cassettes (.jsonl.gz) (#6)

• Automatic Compression & Decompression: Recording to any path ending in .jsonl.gz automatically compresses the cassette stream, reducing disk footprint by 70–85% for multi-turn agent runs and large RAG context payloads.
• Cryptographic Chain Integrity Preserved: Maintains SHA-256 block hash verification and tamper detection across compressed archives.
• Atomic File Writing: Ensures safe, corruption-resistant disk updates via atomic temporary-file replacement.
• CLI & Backward Compatibility: rerunner inspect, play, and stats seamlessly support both .jsonl and .jsonl.gz.
  Contributed by @abhandary14 in #6.

2. GitHub Actions CI Matrix & Cassette Caching Recipe (#5)

• Zero-Secret CI Replay: Detailed production recipe for running agent test suites in CI without exposing model API credentials or making paid API calls.
• Enterprise Security Hardening: Guidance on caching scoped .jsonl fixtures with actions/cache@v4 while preventing pull_request_target security exposures.
  Contributed by @KeuqS in #5.

📦 Installation

`�ash
pip install --upgrade rerunner
``n