Prompt Fixer

Research-backed Prompt Fixer web app that improves prompts, runs a multi-agent hallucination reduction workflow, tests outputs against selected models, and stores user ratings for future learning.

Features

• Prompt input interface.
• Multi-agent prompt improvement workflow (writer, editor, proof-reader, fact-checker).
• Improved prompt preview + consensus status.
• Model test execution endpoint.
• Instruction-following validation step.
• Feedback storage (prompt/output/rating/rank) in SQLite.
• Automated tests for core workflows.

Run locally

python -m venv .venv
source .venv/bin/activate
pip install -e .[dev]
uvicorn app.main:app --reload

Open http://127.0.0.1:8000.

Optional: set OPENAI_API_KEY for live model calls. Otherwise, the app runs in demo mode.

Run tests

pytest -q

Files

• RESEARCH_REPORT.md: planning and research summary with sources.
• app/: FastAPI application and multi-agent services.
• tests/: automated tests.