77 lines (57 loc) · 1.95 KB

Contributing

Thanks for contributing to Pi-Memory.

Philosophy

keep the code small
preserve the C + SQLite core simplicity
prefer reliability over cleverness
keep cross-platform behavior explicit and testable

Local development

npm run setup
npm run doctor
npm run test:smoke
npm pack --dry-run

Notes:

there are no runtime npm dependencies for the core memory engine
npm run setup copies a matching prebuilt when available, otherwise compiles from source
bun run setup is also supported

Repo structure

native/ — C source + makefiles + SQLite amalgamation
extensions/ — Pi extension(s)
skills/ — Pi skill(s)
prompts/ — prompt templates
docs/ — user/developer docs
prebuilds/ — platform-specific prebuilt binaries when committed

Pull request process

Open an issue first for non-trivial changes.
Keep the PR scoped to one concern.
Include before/after behavior.
Update docs for user-visible behavior.
Run the validation steps and include output.
Avoid bundling research artifacts or unrelated local tooling into the PR.

PR checklist

No unrelated files changed
Native build still succeeds cleanly
npm run doctor passes
npm run test:smoke passes
Docs updated where needed
Backward compatibility considered
Cross-platform impact considered (macOS / Linux / Windows)

Coding guidelines

C

prefer straightforward control flow over abstraction layers
keep allocations bounded and obvious
use prepared statements for SQL
do not modify the bundled SQLite amalgamation unless intentionally updating upstream
preserve compatibility with existing memory.db files

Extension / JS / TS

keep lifecycle hooks best-effort and non-fatal
use timeouts on external calls
avoid hardcoded provider assumptions where generic behavior is possible
keep install/runtime helpers consistent with scripts/setup.mjs