v0.7.3

Highlights

• harden bounded Google/Vertex review runs with preflight sizing, chunked --files execution, and per-phase prompt telemetry
• fix Anthropic structured output strictness handling and improve provider error classification
• correct the README so the project is clearly multi-agent and not framed as a Claude-only framework

Validation

• ./.venv/bin/python -m pytest -q → 282 passed, 2 warnings
• ./.venv/bin/python -m compileall src tests → passed
• live local vertex-ai bounded review run on an oversized 3-file context completed successfully on the shipping candidate state with chunked_context preflight and a valid JSON result