# Chatty

CLI chat client for OpenAI models with conversation persistence.

## Quick reference

```bash
# One-shot message (non-interactive, best for scripting / agent use)
python main.py "Your message here"

# One-shot with a named conversation (creates or resumes)
python main.py -n my-thread "Follow-up question"

# One-shot with a specific model
python main.py -m gpt-4o "Hello"

# Interactive mode
python main.py
python main.py -n my-thread

# Resume by conversation ID
python main.py --resume 3

# List saved conversations
python main.py --list

# Custom system prompt
python main.py -s "You are a pirate." "Ahoy?"
```

## Architecture

Single-file app: `main.py`. SQLite database `conversations.db` stores all conversations and messages.

Key functions:
- `send_message()` - one-shot: send a message, get a response, no UI
- `chat_loop()` - interactive REPL with streaming output
- `get_or_create_conversation()` - upserts by name
- `load_messages()` / `save_message()` - DB read/write

## Environment

Requires `OPENAI_API_KEY` in environment or `.env` file. Dependencies: `openai`, `python-dotenv`.

## Conventions

- Default model: `gpt-4.1` (set via `DEFAULT_MODEL`)
- Messages are saved to DB only after a successful API response
- System prompt is stored as the first message in a conversation
- Named conversations are resumed automatically; unnamed ones always create new entries