Cross-Project Recall
When you frame a new problem in project B, quint-code automatically searches for related decisions from all your other projects. If you solved a similar problem six months ago in project A, that decision surfaces — with a congruence level tag indicating how transferable the context is.
How it works
Every time you record a decision (/q-decide), a summary is written to a shared
index at ~/.quint-code/index.db. This index contains: project name, decision title,
selected variant, rationale, weakest link, and primary language.
When you frame a new problem (/q-frame), quint-code searches this index using FTS5
(full-text search). Matching decisions from other projects appear in a
"Cross-Project History" section:
## Cross-Project History
- [dec-20260319-001] Tokio-based parallel executor with semaphore-bounded concurrency
— Tokio is already the standard async runtime in the Rust ecosystem...
(CL1 (different context), from test-rust) Congruence levels
Not all cross-project knowledge transfers equally. A caching decision from another Go project is more relevant than one from a Python project with completely different constraints.
| CL | When | R_eff penalty | Meaning |
|---|---|---|---|
| CL2 | Same primary language | 0.1 | Similar context — likely applicable with minor adjustments |
| CL1 | Different primary language | 0.4 | Different context — the rationale may apply but implementation differs |
Primary language is auto-detected from your codebase modules. A Go project recalling a decision from another Go project gets CL2. Recalling from a Rust project gets CL1.
What gets shared
The index stores summaries only — not the full decision body:
- Decision title and selected variant name
- Why it was selected (rationale)
- Weakest link
- Primary language of the source project
- Project name (for display)
The index is a projection — a read-optimized copy. If it's corrupted or deleted, it can be rebuilt from per-project databases.
Privacy
Everything is local. The index lives at ~/.quint-code/index.db on your machine.
No data leaves your computer. No cloud. No analytics. No telemetry.
In future team mode, decisions will flow through a central server that you host yourself. See the roadmap for what's planned.