usage_example.md

# Quick Start Example

```python
from transformers import AutoModelForCausalLM, AutoTokenizer
import torch

# Load model
model = AutoModelForCausalLM.from_pretrained(
    "pawlaszc/DigitalForensicsText2SQLite",
    torch_dtype=torch.float16,
    device_map="auto"
)
tokenizer = AutoTokenizer.from_pretrained("pawlaszc/DigitalForensicsText2SQLite")

# Example schema
schema = """
CREATE TABLE messages (
    _id INTEGER PRIMARY KEY,
    address TEXT,
    body TEXT,
    date INTEGER,
    read INTEGER
);
"""

# Example request
request = "Find all unread messages from yesterday"

# Generate SQL
prompt = f"""Generate a valid SQLite query for this forensic database request.

Database Schema:
{schema}

Request: {request}

SQLite Query:
"""

inputs = tokenizer(prompt, return_tensors="pt", truncation=True, max_length=2048)
inputs = {k: v.to(model.device) for k, v in inputs.items()}

with torch.no_grad():
    outputs = model.generate(**inputs, max_new_tokens=200, do_sample=False)

# Extract generated SQL
input_length = inputs['input_ids'].shape[1]
generated_tokens = outputs[0][input_length:]
sql = tokenizer.decode(generated_tokens, skip_special_tokens=True)

print(sql.strip())
```

## GGUF Usage (llama.cpp)

```bash
# Download GGUF file (Q4_K_M recommended)
wget https://huggingface.co/pawlaszc/DigitalForensicsText2SQLite/resolve/main/forensic-sql-q4_k_m.gguf

# Run with llama.cpp
./llama-cli -m forensic-sql-q4_k_m.gguf -p "Your prompt here"
```

## Available Files

- **Full model (FP16):** ~6 GB - Best quality
- **Q4_K_M.gguf:** ~2.3 GB - Recommended (95% quality, 2.5× faster)
- **Q5_K_M.gguf:** ~2.8 GB - Higher quality (97% quality)
- **Q8_0.gguf:** ~3.8 GB - Highest quality (99% quality)

## Performance

- Overall: 79% accuracy
- Easy queries: 94.3%
- Medium queries: 80.6%
- Hard queries: 61.8%