DraftForge Repair Agent

DraftForge Repair Agent is an AI-powered code debugging and repair assistant built for the AkashML Hackathon (winning project).

It accepts broken code + error context, identifies probable root causes, and generates patch-ready fixes with explanations.

Core Capabilities

• Bug triage with severity and root-cause hypotheses
• Automated patch generation (unified diff style)
• Optional OpenAI-compatible AkashML model integration
• Fallback deterministic repair mode for offline demos
• Lightweight web studio for fast live demos
• Batch repair runner for processing multiple failures in one pass

Stack

• FastAPI backend
• Static web studio (HTML/CSS/JS)
• httpx model client (OpenAI-compatible endpoint)

Quick Start

python3 -m venv .venv
source .venv/bin/activate
pip install -r requirements.txt
cp .env.example .env
uvicorn apps.api.main:app --reload --port 8010

Open apps/studio/index.html and point API URL to http://localhost:8010.

Endpoints

• GET /health
• POST /api/v1/analyze
• POST /api/v1/repair

Batch Mode

Run repair jobs from JSONL input:

python3 scripts/batch_repair.py \
  --api-url http://localhost:8010 \
  --input ./jobs.jsonl \
  --output ./results.jsonl

Sample input row (jobs.jsonl):

{"id":"job-1","language":"python","strategy":"minimal_patch","error_log":"TypeError: ...","code":"def foo(): ..."}

AkashML Configuration

Set these in .env to run with AkashML-hosted models:

• AKASH_BASE_URL
• AKASH_API_KEY
• AKASH_MODEL

Beyond Hackathon

See docs/ROADMAP.md for productionization milestones.