chatline 0.2.10

A pretty command line interface for LLM chat.

Chatline

A lightweight CLI library for building terminal-based LLM chat interfaces with minimal effort. Provides rich text styling, animations, and conversation state management.

• Terminal UI: Rich text formatting with styled quotes, brackets, emphasis, and more
• Response Streaming: Real-time streamed responses with loading animations
• State Management: Conversation history with edit and retry functionality
• Multiple Providers: Run with AWS Bedrock, OpenRouter, or connect to a custom backend
• Keyboard Shortcuts: Ctrl+E to edit previous message, Ctrl+R to retry

Installation

pip install chatline

With Poetry:

poetry add chatline

Usage

There are two modes: Embedded (with built-in providers) and Remote (requires response generation endpoint).

Embedded Mode with AWS Bedrock (Default)

The easiest way to get started is to use the embedded generator with AWS Bedrock (the default provider):

from chatline import Interface

chat = Interface()

chat.start()

For more customization:

from chatline import Interface

# Initialize with AWS Bedrock (default provider)
chat = Interface(
    provider="bedrock",  # Optional: this is the default
    provider_config={
        "region": "us-west-2",  
        "model_id": "anthropic.claude-3-5-haiku-20241022-v1:0", 
        "profile_name": "development", 
        "timeout": 120  
    }
)

# Add optional welcome message
chat.preface(
    "Welcome", 
    title="My App", 
    border_color="green")

# Start the conversation with custom system and user messages
chat.start([
    {"role": "system", "content": "You are a friendly AI assistant that specializes in code generation."},
    {"role": "user", "content": "Can you help me with a Python project?"}
])

Embedded Mode with OpenRouter

You can also use OpenRouter as your provider: (Just make sure to set your OPENROUTER_API_KEY environment variable first)

from chatline import Interface

# Initialize with OpenRouter provider
chat = Interface(
    provider="openrouter",
    provider_config={
        "model": "deepseek/deepseek-chat-v3-0324", 
        "temperature": 0.7, 
        "top_p": 0.9, 
        "frequency_penalty": 0.5, 
        "presence_penalty": 0.5,
        "timeout": 60 
    }
)

chat.start()

Remote Mode (Custom Backend)

You can also connect to a custom backend by providing the endpoint URL. Passing an empty array allows for the initial messages to be instantiated on the backend:

from chatline import Interface

# Initialize with remote mode
chat = Interface(endpoint="http://localhost:8000/chat")

# Start the conversation with custom system and user messages
chat.start([])

You can use generate_stream function (or build your own) in your backend. Here's an example in a FastAPI server:

import json
import uvicorn
from fastapi import FastAPI, Request
from fastapi.responses import StreamingResponse
from chatline import generate_stream

app = FastAPI()

CONVERSATION_STARTER = [
    {"role": "system", "content": "The Assistant is an Alien!!!"},
    {"role": "user", "content": "Introduce yourself to me!"},
]

@app.post("/chat")
async def stream_chat(request: Request):
    # Parse the request body
    body = await request.json()

    # Get conversation state
    state = body.get("conversation_state", {}) or {}

    # Get messages directly from the request body
    messages = body.get("messages", [])

    # Filter out any messages with empty content
    messages = [msg for msg in messages if msg.get("content", "").strip()]

    if not messages:
        messages = CONVERSATION_STARTER.copy()
        state["messages"] = messages

    # Return streaming response with state
    headers = {
        "Content-Type": "text/event-stream",
        "X-Conversation-State": json.dumps(state),
    }

    return StreamingResponse(
        generate_stream(messages, provider="bedrock"),
        headers=headers,
        media_type="text/event-stream",
    )

if __name__ == "__main__":
    uvicorn.run("server:app", host="127.0.0.1", port=8000)

Acknowledgements

Chatline was built with plenty of LLM assistance, particularly from Anthropic, Mistral and Continue.dev.