Multi-Model Manager

You have GPT-4 for reasoning, Claude for coding, DeepSeek for cost efficiency, and a local llama.cpp for privacy-sensitive data.

But each one needs a different API client, different auth, different message format. Switching between them is a pain.

Tian AI Agent 14.0 solves this with a unified model manager:

# Add any model backend
POST /api/config {"action":"add", "name":"gpt4", "endpoint":"https://api.openai.com/v1", "api_key":"sk-..."}
POST /api/config {"action":"add", "name":"claude", "endpoint":"https://api.anthropic.com/v1", "api_key":"sk-ant-..."}
POST /api/config {"action":"add", "name":"local", "endpoint":"http://localhost:8080"}

# Switch between them instantly
POST /api/config {"action":"switch", "name":"local"}

What's Supported

Provider	Protocol	Capabilities
OpenAI	OpenAI-compat	chat, image, audio, embedding, video
Anthropic Claude	Anthropic native	chat
DeepSeek	OpenAI-compat	chat
Google Gemini	Gemini native	chat, image
xAI Grok	OpenAI-compat	chat
Mistral	OpenAI-compat	chat
Groq	OpenAI-compat	chat
OpenRouter	OpenAI-compat	chat, image
Stability AI	OpenAI-compat	image
Runway	OpenAI-compat	video
ElevenLabs	OpenAI-compat	audio
llama.cpp	Completion API	chat
Ollama	Ollama native	chat, embedding
Any OpenAI-compatible	Auto-detected	chat

Auto Protocol Detection

Just paste the endpoint URL — the tool figures out the format:

from model_manager import ModelConnector
mc = ModelConnector(endpoint="https://api.anthropic.com/v1", api_key="sk-ant-...")
mc.chat([{"role": "user", "content": "Hello"}])
# → Automatically uses Anthropic's /v1/messages format

It works by:

Checking known domains (openai.com → OpenAI format, anthropic.com → Anthropic format, etc.)
Probing the endpoint for common API paths (/v1/chat/completions, /api/chat, /completion)
Falling back to OpenAI-compatible format for unknown endpoints

Multi-Model Routing

The ModelManager keeps all your models in one place. When you send a request, it routes to the right model based on capability:

mm = ModelManager()
mm.add("gpt4", endpoint="https://api.openai.com/v1", api_key="sk-...", 
       capabilities=["chat", "image", "embedding"])
mm.add("sdxl", endpoint="https://api.stability.ai", api_key="sk-...",
       capabilities=["image"])

# Chat goes to GPT-4
mm.chat("Hello")

# Image generation auto-routes to Stability AI
mm.generate_image("A cat in a spacesuit")