About
Deep Tutor is not a chat app or code generator. It is a context layer that shapes how you learn in the terminal: prompt packs, orchestrator logic, and lightweight memory integrated with tools you already use — primarily OpenCode and local models via Ollama (Qwen / DeepSeek-class weights).
Framing: A system designed to improve how developers think.
Two environments
- Deep Tutor (this repo) — Prompt packs, orchestrator,
memory/, domain configs, scripts, and pedagogy experiments - Leveling environment — Sibling workspace (e.g.
leveling-arc/) where skill progression and reflection live
Intended flow: Leveling environment → Deep Tutor context → OpenCode CLI → local LLM
Key workflows
Route a concrete error or question into the right domain context:
./scripts/route -m "IndexError on line 5 when I access nums[i]"
./scripts/preamble -m "What is a deadlock?" --cwd ../leveling-arc/domains/backend
./scripts/opencode-install --target ../leveling-arcA DSA pilot binds sibling leveling-arc/domains/dsa/ via CONTEXT.md; ./scripts/pilot-dsa verifies preamble assembly.
Culture
Engineering narrative lives in devlogs/; structured tries in experiments/. The project practices the same honest, iterative posture it teaches.