Spaces:
Running
Running
File size: 9,652 Bytes
fcc5471 a39d8ef deedaad a39d8ef deedaad a39d8ef | 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 | ---
title: NL2SQL Bench
emoji: π
colorFrom: blue
colorTo: indigo
sdk: docker
pinned: false
---
# NL2SQL-Bench
**Natural Language to SQL Analytics Environment for RL Training**
[](https://github.com/meta-pytorch/OpenEnv)
[](https://www.python.org)
[](LICENSE)
---
## What is NL2SQL-Bench?
NL2SQL-Bench is an OpenEnv-compliant RL training environment where an AI agent must iteratively write and refine **SQLite queries** to answer natural-language business questions against a synthetic e-commerce database.
This fills a genuine gap in the OpenEnv ecosystem β no SQL query environment currently exists. Every data-driven company employs analysts who translate business questions into SQL. Training agents to do this well (and to recover from errors) is immediately valuable.
**Why it's a great RL domain:**
- Rewards are **100% deterministic** β no LLM-as-judge, no subjectivity
- Multi-turn episodes create **dense reward signal** across the trajectory
- The error β fix β retry loop is a novel mechanic not present in existing environments
- Three clearly graduated difficulty levels challenge models across the full skill range
---
## Environment Description
The agent interacts with a **synthetic e-commerce SQLite database** containing ~150 customers, 64 products across 8 categories, ~600 orders, ~1000 order items, and ~400 reviews. The database is seeded deterministically (seed=42) so results are reproducible across any machine.
The agent receives a natural-language question and iteratively submits SQL queries. Each query is executed, graded against the ground truth, and the reward + error/result is fed back as the next observation.
---
## Database Schema
```
categories(id, name)
products(id, name, category_id, price, stock_quantity)
customers(id, name, email, country, tierβ{bronze|silver|gold}, created_at)
orders(id, customer_id, statusβ{pending|processing|shipped|delivered|cancelled},
created_at, total_amount)
order_items(id, order_id, product_id, quantity, unit_price)
reviews(id, product_id, customer_id, ratingβ1-5, created_at)
```
All dates are ISO-8601 strings sortable by text comparison. SQLite window functions and CTEs are fully supported.
---
## Action & Observation Space
### Action
```python
@dataclass
class NL2SQLAction(Action):
query: str # A SQLite SELECT query string
```
### Observation
```python
@dataclass
class NL2SQLObservation(Observation):
question: str # The NL question to answer
schema_context: str # Compact schema description
task_name: str # Active task identifier
last_query: str # SQL submitted on previous step
last_result: List[Dict] # Up to 10 result rows
last_error: Optional[str] # SQLite error string or None
result_columns: List[str] # Column names of last_result
step: int # Current step (0 after reset)
max_steps: int # Maximum steps per episode
done: bool # Episode ended?
reward: Optional[float] # Step reward [0.0, 1.0]
score: float # Cumulative normalised score
```
---
## Tasks & Expected Difficulty
### Task 1 β `simple-filter` (easy)
Single-table SELECT queries with WHERE, ORDER BY, LIMIT. Tests basic SQL fluency. Example questions:
- "List all gold-tier customers ordered by name alphabetically."
- "Return the top 5 most expensive products."
**Expected solve rate (frontier model, 5 steps):** ~80%
### Task 2 β `join-aggregation` (medium)
Multi-table JOINs with GROUP BY, HAVING, and aggregation functions. Example questions:
- "How many orders has each customer placed? Include customers with zero orders."
- "Which customers have spent more than $500 total on delivered orders?"
**Expected solve rate (frontier model, 5 steps):** ~55%
### Task 3 β `analytics-window` (hard)
CTEs, window functions (DENSE_RANK, ROW_NUMBER, running SUM), and nested subqueries. Example questions:
- "Rank customers by total spending using DENSE_RANK."
- "Show monthly revenue and running total for delivered orders in 2024."
**Expected solve rate (frontier model, 5 steps):** ~30%
---
## Reward Function
Rewards are computed by deterministic comparison of the agent's result set against the ground truth:
| Component | Score | Description |
|---|---|---|
| `syntax_ok` | +0.10 | Query runs without SQLite error |
| `columns_match` | +0.20 | Returned column names match ground truth |
| `row_count_match` | +0.20 | Number of rows matches |
| `exact_match` | +0.50 | Full result set equals ground truth |
| `step_penalty` | β0.05/step | Deducted per step beyond the first |
Final reward is clamped to `[0.0, 1.0]`. Order sensitivity matches the ground-truth query: ORDER BY queries require correct row ordering; others are order-agnostic.
---
## Baseline Scores
Run by the `inference.py` script using `Qwen/Qwen2.5-72B-Instruct` via HuggingFace router:
| Task | Expected Score |
|---|---|
| `simple-filter` | ~0.70 |
| `join-aggregation` | ~0.45 |
| `analytics-window` | ~0.25 |
---
## Setup & Usage
### Prerequisites
- Python 3.10+
- Docker (for containerised deployment)
- A HuggingFace account + token
### Local Development (no Docker)
```bash
# Clone the repository
git clone https://huggingface.co/spaces/your-username/nl2sql-bench
cd nl2sql-bench
# Quick start
chmod +x scripts/run_local.sh
./scripts/run_local.sh
# Or manually:
python3 -m venv .venv && source .venv/bin/activate
pip install openenv-core fastapi "uvicorn[standard]" openai pydantic
export PYTHONPATH=".:server"
cd server && uvicorn app:app --reload --port 8000
```
### Test the Running Server
```bash
# Run smoke tests
chmod +x scripts/smoke_test.sh
./scripts/smoke_test.sh http://localhost:8000
# Run full test suite
pip install pytest pytest-asyncio
PYTHONPATH=".:server" pytest tests/ -v
```
### Docker
```bash
# Build
docker build -t nl2sql-bench:latest .
# Run
docker run -p 7860:7860 nl2sql-bench:latest
# Test
./scripts/smoke_test.sh http://localhost:7860
```
### Pre-submission Validation
```bash
# Run the official validator (replace with your HF Space URL)
chmod +x pre_validation_script.sh
./pre_validation_script.sh https://your-username-nl2sql-bench.hf.space .
```
### Running the Baseline Inference
```bash
# Set mandatory variables
export API_BASE_URL="https://router.huggingface.co/v1"
export MODEL_NAME="Qwen/Qwen2.5-72B-Instruct"
export HF_TOKEN="hf_your_token_here"
export SPACE_URL="https://your-username-nl2sql-bench.hf.space"
python inference.py
```
### Using the Client Programmatically
```python
import asyncio
from client import NL2SQLEnv
from models import NL2SQLAction
async def main():
async with NL2SQLEnv(base_url="http://localhost:8000") as env:
result = await env.reset()
print(result.observation.question)
result = await env.step(NL2SQLAction(
query="SELECT id, name FROM customers WHERE tier='gold' ORDER BY name"
))
print(f"Reward: {result.reward:.2f}")
print(f"Done: {result.done}")
print(f"Error: {result.observation.last_error}")
asyncio.run(main())
```
---
## Project Structure
```
nl2sql-bench/
βββ models.py # NL2SQLAction, NL2SQLObservation, NL2SQLState
βββ client.py # NL2SQLEnv(HTTPEnvClient)
βββ inference.py # Baseline inference script (mandatory name)
βββ openenv.yaml # OpenEnv manifest
βββ pyproject.toml
βββ Dockerfile # HF Spaces compatible (port 7860)
βββ .env.example
βββ server/
β βββ app.py # FastAPI entry point
β βββ environment.py # Core RL environment logic
β βββ grader.py # Deterministic reward computation
β βββ requirements.txt
β βββ db/
β β βββ schema.sql # 6-table e-commerce schema
β β βββ seed.py # Deterministic data generator (seed=42)
β βββ tasks/
β βββ base.py # BaseTask + registry
β βββ easy.py # simple-filter (5 examples)
β βββ medium.py # join-aggregation (5 examples)
β βββ hard.py # analytics-window (5 examples)
βββ tests/
β βββ conftest.py
β βββ test_all.py # 30+ pytest tests
βββ scripts/
βββ run_local.sh # Local dev server
βββ smoke_test.sh # Endpoint smoke tests
```
---
## Design Decisions
**Why SQLite in-memory?** Zero runtime dependency, deterministic, and it runs comfortably within the 2 vCPU / 8 GB constraint. The database loads in ~50ms.
**Why multi-turn (up to 5 steps)?** A single-shot SQL environment gives binary rewards. Multi-turn with error feedback gives the agent β and the GRPO trainer β a rich signal: the model learns not just to write SQL, but to debug and refine its queries.
**Why step penalty?** Without it, an agent that accidentally gets the right answer on step 5 scores the same as one that gets it on step 1. The penalty creates pressure to solve efficiently, which is realistic.
**Why order-sensitive comparison for ORDER BY queries?** Business questions that say "rank by spending" expect a ranked output. Order-agnostic comparison would give spurious credit.
---
## License
MIT β see [LICENSE](LICENSE)
|