Benchmark and guarantee warm-start: a cold python -c "<fluent chain>" against an already-analyzed project must answer from cache in ≤ ~1s (interpreter start + analysis load + query).
Today, loading inflates the full analysis.json into Pydantic models up front. If cold start is slow, the codeact story loses to tool-protocol round trips — this budget gates Epic C.
Scope
- Benchmark script over the bundled sample apps + one large fixture; publish numbers in-repo.
- Lazy/partial model hydration keyed by what the chain actually touches.
- If the budget cannot be met in-process, evaluate answering cold calls from the Neo4j backend.
Part of #195. Branch: feat/issue-<n>.
Benchmark and guarantee warm-start: a cold
python -c "<fluent chain>"against an already-analyzed project must answer from cache in ≤ ~1s (interpreter start + analysis load + query).Today, loading inflates the full
analysis.jsoninto Pydantic models up front. If cold start is slow, the codeact story loses to tool-protocol round trips — this budget gates Epic C.Scope
Part of #195. Branch:
feat/issue-<n>.