SafeAdapt

Research prototype for studying alignment drift in continually interacting LLM agents.

Status

Phase 10 (LLM) — OpenAI-compatible (DeepSeek), Ollama, and independent LLM judge.

Phases 1–10 complete for the research prototype pipeline (mock + real LLM).

Setup

python -m venv .venv
source .venv/bin/activate  # Windows: .venv\Scripts\activate
pip install -e ".[dev]"
cp .env.example .env       # set DEEPSEEK_API_KEY / OPENAI_API_KEY

Usage

# Mock (no API)
safeadapt run configs/experiments/filesystem_drift.yaml
safeadapt matrix configs/experiments/matrix_dev.yaml

# DeepSeek (OpenAI-compatible)
safeadapt run configs/experiments/filesystem_deepseek.yaml

# Ollama (local)
safeadapt run configs/experiments/filesystem_ollama.yaml

Web showcase

Public UI (Next.js on Vercel) + API (FastAPI on Render):

# API
pip install -e ".[dev]" && pip install -r api/requirements.txt
uvicorn api.main:app --host 0.0.0.0 --port 8000

# Frontend
cd web && cp .env.example .env.local
# set NEXT_PUBLIC_API_URL=http://127.0.0.1:8000
npm install && npm run dev

Regenerate committed showcase plots: PYTHONPATH=src python scripts/export_showcase.py

Tests

pytest -v

See plan.md for the full specification.