Personalization
Codistry can adapt to how you work: the conventions you follow, the corrections you tend to make, and what it has learned about your codebase. You control all of it from Settings → Personalization, and none of it leaves your machine.
Custom rules, learned preferences, and project memory are stored locally. They're never sent to Codistry servers.
Custom system-prompt rules
Add your own instructions that Codistry appends to every request: coding standards, framework conventions, testing expectations, or any preference it should always honor. Keep it tight (roughly 100–300 words) so it stays effective.
Good rules are specific
For example: “Always use TypeScript strict mode. Prefer functional patterns. Include error handling in every code suggestion.” Stored in contextRules.
Adaptive learning & your worldview
With Adaptive Learning on, Codistry continuously picks up your preferences and working style, across all your projects, and applies them automatically. What it has learned is kept as plain, editable text called your Worldview Content.
Learns as you go
Corrections and feedback become durable preferences. Toggle it with enableAdaptiveLearning.
Yours to edit
Read or curate the learned text directly in adaptiveLearningContent. Add, refine, or remove anything.
Worldview history
Codistry keeps previous versions of your worldview, so you can review how it has changed over time and roll back if needed.
Project memory
Project Memory lets Codistry build and maintain a private knowledge graph about your codebase: what different parts do, how they relate, and useful observations about structure. This persists across sessions, so the agent stays oriented without re-analyzing everything each time.
Project memory is the long-lived, learned layer; the Context Engine is the structural index of your code. Together they give the agent durable, context-aware understanding. Toggle memory with projectMemory.
Alongside the knowledge graph, the agent also keeps repo-specific notes as files you can read and edit: .codistry/context.md (a concise project map) and Dynamic Skills — step-by-step playbooks it captures for recurring workflows. Both live in your repo, so they travel with your code and your team.
- Settings Reference: Context & Memory
- Dynamic Skills, repo-specific playbooks the agent maintains
- Context Engine, the structural side of context
- Expertise, the mined who-knows-what profile (distinct from what you teach it here)