[ajianaz]

What

First part of the single write path (#1514, epic #1523). remember_precomputed_detailed already held the shared tail; it is now split into building the Memory and commit_memory (row, source hash, Created event, edge wiring, cache invalidation, vector, cosine links), which later parts reuse for import and consolidation. The embedding step is one function, embed_for_write (JSON flattened, retried, FTS5-only with a reason on failure), and type validation is one function. remember_with_contradiction uses both.

Why

The contradiction path had drifted from plain remember: it stored an unknown memory type, embedded JSON as raw text, tried the embedding once, and failed the whole write when the embedding failed. It now behaves like remember. On a persistent embedding failure, or with no embedder, the memory is stored FTS5-only and the contradiction check is skipped, with a warning that says so when the caller asked for it. Import and consolidation (the CORE-visible parts) wait for uteke-cloud#242. Refs #1514.

Testing

• clippy -D warnings, fmt, cargo test --workspace pass; cargo run -p docgen changes nothing
• 6 new tests in write_path_tests.rs; run against an unmodified develop worktree, 4 of the first 5 fail there (type check, JSON text, retry, FTS5-only degradation) and the side-effect test passes, as intended
• reviewed locally with the Cora MCP agent reviewer before pushing: no blocking finding; one minor finding of mine (a skipped contradiction check was silent) was fixed and tested before this push

[github-actions]

🔍 Cora AI Code Review

✅ No issues found. Code looks good!

---

Review powered by cora-code · BYOK · MIT