Quest

Socratic AI tutor — no explanations, only the next better question. Separate evaluator, concept DAGs, mastery tracking, resumable sessions.

Requires Python 3.11+ · Author: yuvrxj

pip install -U quest-ai
mkdir -p ~/.quest
echo 'ANTHROPIC_API_KEY=sk-ant-...' >> ~/.quest/.env
quest

Quick start

At the catalog prompt: 3, rag_pipeline, rag_pipeline --fresh, new react hooks, help, quit.

Installed data lives in ~/.quest/ (quest.db, checkpoints, custom topics). In a git checkout, data stays beside cli.py.

How sessions work

1. You pick a topic (bundled catalog or quest topic new).
2. Quest selects a concept from a prerequisite graph (mastery + spaced repetition).
3. Tutor asks one Socratic question → you answer → evaluator scores that answer (1–5).
4. Repeat on the same concept until mastery (≥4 on three turns), then next concept.
5. /quit pauses; run quest TOPIC again to resume.

Memory: the tutor sees your full Q↔A on the current concept. The evaluator scores only the latest question and answer (/last in-session for detail).

vs “just use ChatGPT”

Quest is for learning: the tutor does not teach; a separate model scores you; topics are structured graphs; progress persists across sessions.

Links (GitHub)

Use these absolute URLs — PyPI cannot serve files under docs/ on the package page.

• User guide
• Changelog
• Repository

Development

git clone https://github.com/preoracle/quest.git
cd quest
python -m venv .venv && source .venv/bin/activate
pip install -e ".[dev]"
cp .env.example .env
pytest
python cli.py

Optional: QUEST_WIZARD=1 quest for arrow-key menus instead of the Rich catalog.

Publish to PyPI (token in .env as PYPI_TOKEN or exported in shell):

chmod +x scripts/publish_pypi.sh
./scripts/publish_pypi.sh          # build + upload
./scripts/publish_pypi.sh --check  # build + validate only

Internal design notes: BRIEF · PHASES