Skip to content

developing.md

Latest commit

 

History

History
38 lines (29 loc) · 1.2 KB

developing.md

File metadata and controls

38 lines (29 loc) · 1.2 KB

Developing On Codex Blackbox

Codex Blackbox has fake, preflight, dogfood, and live evidence categories. Keep those separate in code, docs, tests, and release notes.

Useful local commands:

cargo fmt --check
cargo test --workspace
cargo clippy --workspace -- -D warnings
./test/validate-openai-config.sh
./test/e2e-openai-responses-full.sh

Start the local stack:

codex-blackbox up
docker compose logs -f codex-blackbox-core

Run a real dogfood check only when explicitly intended:

codex-blackbox run -- codex exec --sandbox read-only "Read README.md and summarize Codex Blackbox in 3 bullets. Do not edit files."
codex-blackbox postmortem last

Release-facing claims require a real Codex model turn observed by codex-blackbox-core with provider="codex_responses". Fake fixture tests validate local parser, persistence, API, watch, and dashboard contracts only. Keep the fake regression in CI, but do not describe it as live support proof.

Before changing behavior, read the repository AGENTS.md and the module map in that file. The highest-risk paths are request parsing, Responses SSE accumulation, token accounting, persistence, watch replay, metrics labels, and the CLI wrapper.