vintage-ai

Telnet to AI bridge

Running Locally

Ensure that any of these AI providers are installed:

• google-genai for Google Gemini models
• openai for OpenAI models
• anthropic for Anthropic Claude models

Set the AI_PLATFORM environment variable to either gemini, openai, or anthropic. It will default to gemini if not set.

Set the AI_MODEL environment variable to be the model name you want to use. e.g o4-mini or gemini-2.0-flash.

If this is not set, Gemini will default to gemini-2.0-flash, OpenAI will default to gpt-4o-mini, and Anthropic will default to claude-3-5-sonnet-latest

Gemini

• First set the GEMINI_API_KEY environment variable to be your Google AI Studio API key
• Install the Gemini library pip install google-genai

OpenAI Compatible API's (OpenAI, Ollama, vLLM etc)

• First set the OPENAI_API_KEY environment variable to be your OpenAI/Ollama/whatever API key
• IF you want to use Ollama/vLLM/any other OpenAI compatible API, set the OPENAI_BASE_URL environment variable to the relevant base URL for your provider. You WILL need to set the AI_MODEL environment variable if you do this, else it can't default to an OpenAI model.
• Install the OpenAI library pip install openai

Anthropic Claude

• First set the ANTHROPIC_API_KEY environment variable to be your Anthropic API key
• Install the Anthropic library pip install anthropic

Run the server with python3 ./start_server.py then connect to it from whatever telnet client you want to use (it runs on port 2323).

Performance & Memory Management

The server has been optimized for efficiency and includes:

• Memory management: Chat history is automatically limited to 50 messages per connection to prevent memory leaks
• Improved threading: Proper cleanup of spinner animation threads with timeout handling
• Optimized text processing: Pre-compiled regex patterns for better performance
• Unified streaming: Single codebase handles all AI platforms efficiently
• Clean shutdown: Daemon threads allow graceful server termination

Docker

This can be run via docker, if you so wanted. Edit the docker-compose.yml as per the instructions within, setting the relevant API keys and AI platform to use etc. Then run docker compose up -d to start the container. Connect to it on port 2323 from your telnet/bbs client.

Commands

The server now supports runtime commands:

• /model <name> - Change AI model during your session (e.g. /model gpt-4o, /model claude-3-5-sonnet-20241022)
• /status - Show current platform, model, and chat history count
• /help - Display available commands

Other Things to Know

Double enter sends what you've typed. This allows you to send multi-line prompts.

See here for a video of it working: https://bsky.app/profile/did:plc:xt3aino56gwrypmn2ovfmg3a/post/3lott7mkcwc2f