Apppublic

yashodipmore/queryforge-v1

sourceHugging Faceupdated 6mo agoView on Hugging Face

0likes

App README

QueryForge-v1

A real-world SQL query optimization and debugging environment for AI agents.

QueryForge trains and evaluates agents on tasks that database engineers do every day: fixing broken queries, optimizing slow ones, and redesigning inefficient schemas.

Submission Quick Links

—GitHub Repository: https://github.com/yashodipmore/queryforge-v1
—Hugging Face Space: https://huggingface.co/spaces/yashodipmore/queryforge-v1
—Team Name: Sarthak
—Team Members: Yashodip More, Komal Kumavat, Jaykumar Girase

Environment Overview

Property	Value
Backend	SQLite in-memory
Tasks	3 (Easy to Hard)
Max steps	8 / 10 / 12 per task
Reward range	0.0 to 1.0 (partial, per step)
Action space	4 action types
Observation space	Schema + Query + Execution Result + Metrics

Tasks

Task 1: Fix Broken SQL Query (Easy)

Agent receives a query with syntax or logic errors. Goal: fix the query so it executes and returns the correct rows.

Task 2: Optimize Slow Query (Medium)

Agent receives a syntactically correct but slow query. Goal: add indexes and/or rewrite the query to eliminate expensive operations.

Task 3: Redesign Inefficient Schema (Hard)

Agent receives a denormalized table. Goal: propose a normalized schema, write migration queries, and verify data integrity.

Action Space

Action	Description	Required Fields
rewrite_query	Replace current query with new SQL	query
add_index	Add an index to the schema	index_definition
analyze_table	Inspect a table columns and statistics	table_name
submit	Submit current query as final answer	none

Reward Function

reward = syntax_score (0.3)
       + correctness_score (0.4)
       + performance_score (0.2)
       + efficiency_bonus (0.1)
       - penalties

Reward is computed at every step to provide continuous signal.

Setup

Local Development

bash

pip install -r requirements.txt
uvicorn app:app --host 0.0.0.0 --port 7860 --reload

Docker

bash

docker build -t queryforge-v1 .
docker run -p 7860:7860 queryforge-v1

API Usage

bash

curl -X POST http://localhost:7860/reset \
  -H "Content-Type: application/json" \
  -d '{"task_id": "fix_broken_query"}'

curl -X POST http://localhost:7860/step \
  -H "Content-Type: application/json" \
  -d '{"action_type": "rewrite_query", "query": "SELECT customer_id, SUM(amount) FROM orders WHERE status = '\''paid'\'' GROUP BY customer_id"}'

curl http://localhost:7860/state

Run Baseline Inference

bash

# HF_TOKEN is required
export HF_TOKEN=your_token_here
export API_BASE_URL=https://router.huggingface.co/v1
export MODEL_NAME=Qwen/Qwen2.5-72B-Instruct
export QUERYFORGE_URL=http://localhost:7860

python inference.py

OpenEnv Validation

bash

pip install openenv-core

# If the CLI is available in your environment
openenv validate

Pre-Submission Checklist

—openenv.yaml present and task IDs match implementation
—inference.py in repository root and START/STEP/END log format preserved
—/health, /reset, /step, /state endpoints return valid responses
—Rewards stay in [0.0, 1.0] and graders remain deterministic
—docker build succeeds and container serves on port 7860
—GitHub repository and Hugging Face Space links are public and accessible

License

MIT