Skip to content

Solution for long term memory for agent coding CLIs and to facilitate handoff between different agent vendors

Rust

docs(contributing): align workflow guidance with canonical rules

zydoa2d9542d88d6
diff --git a/CONTRIBUTING.md b/CONTRIBUTING.md
index 73ca1292..462eb475 100644
--- a/CONTRIBUTING.md
+++ b/CONTRIBUTING.md
@@ -131,14 +131,12 @@ The full rules are in [`AGENTS.md`](AGENTS.md), the single
canonical agent/contributor rules file (`CLAUDE.md` is just a pointer to
it). Short version:
-1. Work milestone by milestone. Do not start M(n+1) until every "Done when"
- bullet in M(n) passes (see `docs/design-decisions.md`).
-2. No dead code, no half-built features. Stubs are documented with
- `// M<n> TODO` in the module doc-comment.
+1. Keep changes small and scoped to the task; avoid adjacent feature work.
+2. No dead code or half-built features. Document future work in `docs/`
+ design notes rather than shipping unreachable stubs.
3. Write tests before claiming done. Parsers, ID derivation, and
retention/decay math especially.
-4. Do not refactor outside the milestone. Only touch what the current
- milestone requires.
+4. Avoid opportunistic refactors or speculative abstractions.
5. Comments explain *why*, never *what*. No comments that restate the line
above them.
6. The core stays Rust-only. The server, its built-in web UI and everything
@@ -151,15 +149,18 @@ it). Short version:
## Cross-cutting invariants
Never violate any of the invariants in [`AGENTS.md`](AGENTS.md) (see the
-"Rust Engineering Rules" and "Project Maintenance Rules" sections).
+"Cross-cutting invariants" and "Project maintenance rules" sections).
Highlights for contributors:
- All SQLite writes go through the single writer actor (`WriterHandle`).
-- Config is read once at startup; never call `std::env::var` outside `Config::load`.
+- ai-memory settings are read once by `Config::load()` and passed down.
+ Process and third-party harness environment reads use injectable lookups.
- Atomic file writes only: tmp + rename + fsync; never write in-place.
-- Every wiki page is namespaced by `(workspace_id, project_id)`.
-- The CLI is always a thin HTTP client to the running server. It never
- opens the SQLite file or the wiki directory directly.
+- Every wiki page is identified by `(workspace_id, project_id, path)`.
+- Keep server-facing CLI commands thin HTTP clients. Offline lifecycle
+ operations (`reset`, `restore`, `reindex`, and `uninstall --purge-data`)
+ access disk directly and must retain their live-process guards. `backup`
+ uses the running server's online SQLite backup API.
## Versioning and deprecation policy