WEAREAI mod notes

• Client-side Fabric mod that hijacks chat when a message starts with the trigger word (default !ai), sends the prompt to your AI endpoint, and posts the reply back to chat.
• Configurable via Cloth Config (keybind \ opens the screen) with options for trigger word, max character length (applied to prompts and replies), model name, base URL, optional API key + header name, and whether to send the AI reply to the server chat or keep it client-side.
• Sends a single JSON POST to the configured base URL shaped like an OpenAI/Ollama-style chat request: {"model": "<model>", "messages": [{"role": "user", "content": "<prompt>"}]} with Content-Type: application/json and optional API-key header.
• Response text is pulled from common fields (choices[0].message.content, choices[0].text, message.content, or response) and trimmed to the configured length before showing in chat; failures are reported locally instead of blocking chat.
• The mod runs entirely on the client; no server install is required. Keep API keys optional and set only if your endpoint needs them.

Known limitations / ideas:

• Assumes your endpoint accepts an OpenAI/Ollama-like JSON body; Gemini/Claude endpoints may need a compatibility proxy.
• No streaming support; replies appear once the HTTP call finishes.
• Single-turn prompts only; conversation memory is not persisted between messages.