[ajianaz]

What

Third step of #1515 (option B, staged). uteke_core::EngineSettings holds [tier], [recall], [dream], [lifecycle] and [embed_fallback] (FallbackSettings is deserializable like the other four). The CLI config and the server config embed it with #[serde(flatten)], so their five mirrored sections and the CLI's fallback struct are gone, and Uteke::apply_settings replaces the sequence of setters (Jaccard weight, dream thresholds, lifecycle, fallback) that both binaries repeated. Hand-written code removed: ~200 lines.

Why

The sections were written twice and applied twice. MCP still applies none of them; that is part 4 (a loader for MCP, which changes MCP recall results and gets its own contract note). Refs #1515, #1355.

Testing

• clippy -D warnings, fmt, cargo test --workspace pass; cargo run -p docgen changes nothing
• 4 new core tests (the five sections parse next to other and unknown sections through flatten, an empty file is all defaults, apply_settings sets what it should, an incomplete fallback is not applied)
• side by side with the real ONNX model: the CLI and the server built from develop and from this branch, same config with all five sections, a partial file and an invalid one (negative tier days, bad strategy), give identical recall results, lifecycle status, aging status; the only difference is that the server logs the dream and lifecycle settings on every start
• reviewed locally with the Cora MCP agent reviewer before pushing: no blocking finding; two notes of mine (the settings now carry the fallback key but have no Debug; two log lines always appear) are recorded in the code and the changelog

[github-actions]

🔍 Cora AI Code Review

✅ No issues found. Code looks good!

---

Review powered by cora-code · BYOK · MIT