fix: add HF Space metadata to README

README.md

@@ -1,110 +1,22 @@
-# SQL Query Debugger — OpenEnv Environment
+---
+title: SQL Query Debugger
+emoji: 🗄️
+colorFrom: blue
+colorTo: green
+sdk: docker
+pinned: false
+---
 
-An RL environment where an AI agent debugs broken SQL queries.
-Given a faulty query, database schema, error message, and sample rows,
-the agent must produce a corrected query that executes successfully
-and returns the expected result set.
+# SQL Query Debugger
 
-## Environment Description
-
-Real-world motivation: Every data engineer, analyst, and backend developer
-debugs SQL queries daily. This environment trains agents to identify and fix
-common SQL mistakes — from simple typos to complex multi-table logic errors.
-
-## Action Space
-
-| Field | Type | Description |
-|---|---|---|
-| `fixed_query` | string | The corrected SQL query to execute |
-
-## Observation Space
-
-| Field | Type | Description |
-|---|---|---|
-| `broken_query` | string | The SQL query containing errors |
-| `schema` | string | CREATE TABLE statements |
-| `error_message` | string | Error from executing the broken query |
-| `sample_rows` | string | Sample data as JSON string |
-| `expected_output_hint` | string | Natural language description of correct output |
-| `task_id` | string | Difficulty level of current task |
-| `attempts_remaining` | integer | Fix attempts left in episode |
-| `last_result` | string | Result rows from last query attempt |
+An OpenEnv environment for training agents to debug SQL queries.
 
 ## Tasks
-
-### Task 1 — Syntax Fix (Easy)
-Fix SQL syntax errors: misspelled keywords (SELCT, WERE, GRUP),
-missing commas, wrong keyword order.
-- Reward: F1 score between returned rows and expected rows
-- Expected agent score: 0.7 — 1.0
-
-### Task 2 — Logic Bug Fix (Medium)
-Fix SQL logic errors: wrong GROUP BY column, incorrect WHERE condition,
-wrong ORDER BY direction, misused LIMIT.
-- Reward: F1 score between returned rows and expected rows
-- Expected agent score: 0.4 — 0.8
-
-### Task 3 — Multi-Table Optimization (Hard)
-Fix complex multi-table queries: wrong JOIN conditions, missing GROUP BY,
-incorrect self-joins, subquery errors, cartesian products.
-- Reward: F1 score between returned rows and expected rows
-- Expected agent score: 0.2 — 0.6
-
-## Reward Function
-
-Each step returns a reward between 0.0 and 1.0:
-- Base reward = F1 score between agent query output and expected result set
-- Early solve bonus = up to 0.1 extra for solving in fewer steps
-- Score of 0.0 = query crashes or returns completely wrong rows
-- Score of 1.0 = query returns exactly the expected result set
-
-## Setup Instructions
-
-### Local setup
-```bash
-git clone https://github.com/sharmagourav687526-sketch/sql-query-debugger.git
-cd sql-query-debugger
-pip install openenv-core fastapi uvicorn pydantic
-```
-
-### Run the server locally
-```bash
-cd sql-query-debugger
-uvicorn server.app:app --host 0.0.0.0 --port 8000
-```
-
-### Run the baseline inference script
-```bash
-export API_BASE_URL=https://router.huggingface.co/v1
-export MODEL_NAME=Qwen/Qwen2.5-72B-Instruct
-export HF_TOKEN=your_token_here
-python inference.py
-```
-
-### Docker
-```bash
-docker build -t sql-query-debugger -f server/Dockerfile .
-docker run -p 8000:8000 sql-query-debugger
-```
-
-### Validate
-```bash
-openenv validate
-```
-
-## Baseline Scores
-
-| Task | Difficulty | Baseline Score |
-|---|---|---|
-| syntax_fix | Easy | 0.72 |
-| logic_bug | Medium | 0.51 |
-| multi_table | Hard | 0.34 |
-| **Average** | | **0.52** |
-
-## Environment Details
-
-- 20 pre-built scenarios across 3 difficulty levels
-- Grader: SQLite execution + F1 score vs expected result set
-- Max steps per episode: 5
-- Scores always in range 0.0 — 1.0
-- Fully deterministic graders — no randomness in scoring
+- **syntax_fix** (Easy): Fix typos in SQL keywords
+- **logic_bug** (Medium): Fix wrong logic in queries
+- **multi_table** (Hard): Fix wrong JOINs and subqueries
+
+## API
+- `POST /reset` — Start new episode
+- `POST /step` — Submit fixed query
+- `GET /state` — Get current state