LLM Providers
LibreFang ships with a comprehensive model catalog covering 3 native LLM drivers, 49 providers, 230+ builtin models, and 23 aliases. Every provider uses one of three battle-tested drivers: the native Anthropic driver, the native Gemini driver, or the universal OpenAI-compatible driver.
This section is split into focused pages so you can jump straight to hosted APIs, local runtimes, platform integrations, developer-tool backends, or provider operations. The model catalog also supports dynamic loading: official provider definitions are pre-installed to ~/.librefang/providers/, and any custom *.toml files you add there are merged into the catalog at boot.
Table of Contents
- Hosted APIs — Anthropic, OpenAI, Gemini, DeepSeek, Groq, OpenRouter, Mistral, Together, Fireworks, Perplexity, Cohere, Cerebras, SambaNova, Hugging Face, xAI, Alibaba Coding Plan, Moonshot, Novita, AWS Bedrock
- Local & Self-Hosted — Ollama, vLLM, LM Studio
- Platforms & Managed Endpoints — Replicate, NVIDIA NIM, DeepInfra, Azure OpenAI, Qwen (DashScope), MiniMax, Qianfan (Baidu), VolcEngine (Doubao), BytePlus ModelArk, Zhipu (GLM), Zhipu Coding, Z.ai, Vertex AI
- Developer Tools — Claude Code, GitHub Copilot, ChatGPT (Session), Kimi Coding, Aider, Claude Code CLI, Codex CLI, Gemini CLI, Qwen Code
- Provider Management — dynamic loading, model catalog, aliases, routing, quotas, fallback, API endpoints, channel commands, env vars, security
Capability Matrix
A quick glance at which providers support which capabilities. ✓ = wired in the driver, ○ = partial / model-dependent, ✗ = not available via this provider in LibreFang. Embeddings shown here are inference-time embeddings, not the local Ollama auto-detect ladder.
| Provider | Vision | Tools | Embeddings | File Upload | Prompt Cache | Image Gen | Audio |
|---|---|---|---|---|---|---|---|
| Anthropic | ✓ | ✓ | ✗ | ✓ | ✓ | ✗ | ✗ |
| OpenAI | ✓ | ✓ | ✓ | ✓ | ✓ | ✓ | ✓ |
| Google Gemini | ✓ | ✓ | ✓ | ○ | ✗ | ✗ | ✗ |
| DeepSeek | ✗ | ✓ | ✗ | ✗ | ✗ | ✗ | ✗ |
| Groq | ✗ | ✓ | ✗ | ✗ | ✗ | ✗ | ✗ |
| OpenRouter | ✓ | ✓ | ✗ | ○ | ○ | ✓ | ○ |
| Mistral AI | ✓ | ✓ | ✓ | ✗ | ✗ | ✗ | ✗ |
| Together AI | ✓ | ✓ | ✓ | ✗ | ✗ | ✗ | ✗ |
| Fireworks AI | ✓ | ✓ | ✓ | ✗ | ✗ | ✗ | ✗ |
| Perplexity | ✓ | ✗ | ✗ | ✗ | ✗ | ✗ | ✗ |
| Cohere | ✗ | ✓ | ✓ | ✗ | ✗ | ✗ | ✗ |
| Cerebras | ✗ | ✓ | ✗ | ✗ | ✗ | ✗ | ✗ |
| SambaNova | ✗ | ✓ | ✗ | ✗ | ✗ | ✗ | ✗ |
| Hugging Face | ○ | ✓ | ✓ | ✗ | ✗ | ✗ | ✗ |
| xAI | ✓ | ✓ | ✗ | ✗ | ✗ | ✗ | ✗ |
| Alibaba Coding | ✓ | ✓ | ✗ | ✗ | ✗ | ✗ | ✗ |
| Moonshot (Kimi) | ✓ | ✓ | ✗ | ✓ | ✗ | ✗ | ✗ |
| Novita AI | ○ | ✓ | ✗ | ✗ | ✗ | ✗ | ✗ |
| AWS Bedrock | ✗ | ✓ | ✓ | ✗ | ○ | ✗ | ○ |
| Replicate | ✓ | ✗ | ✓ | ○ | ✗ | ✓ | ✓ |
| Azure OpenAI | ✓ | ✓ | ✓ | ✓ | ✓ | ✓ | ✓ |
| Qwen (DashScope) | ✓ | ✓ | ✓ | ✗ | ✗ | ✗ | ✗ |
| MiniMax | ✓ | ✓ | ✓ | ✗ | ✗ | ✗ | ✗ |
| Qianfan (Baidu) | ✓ | ✓ | ✓ | ○ | ✗ | ✗ | ○ |
| VolcEngine (Doubao) | ✓ | ✓ | ✗ | ○ | ✗ | ✗ | ✗ |
| BytePlus ModelArk | ✓ | ✓ | ✗ | ○ | ✗ | ✗ | ✗ |
| Zhipu (GLM) | ✓ | ✓ | ✓ | ○ | ✗ | ✗ | ✗ |
| Vertex AI | ✓ | ✓ | ✓ | ✓ | ✗ | ✓ | ○ |
| GitHub Copilot | ✓ | ✓ | ✗ | ✗ | ✗ | ✗ | ✗ |
| ChatGPT (Session) | ✓ | ✓ | ✗ | ✓ | ✗ | ✗ | ○ |
| Kimi Coding | ✓ | ✓ | ✗ | ✓ | ✗ | ✗ | ✗ |
| Ollama (Local) | ○ | ○ | ✓ | ✗ | ✗ | ✗ | ✗ |
For the full provider catalog including every alias and model, run librefang models list against a running daemon — the in-process registry is always the source of truth.
Quick Setup
The fastest path from zero to running:
# Pick ONE provider — set its env var — done.
export GEMINI_API_KEY="your-key" # Free tier available
# OR
export GROQ_API_KEY="your-key" # Free tier available
# OR
export ANTHROPIC_API_KEY="your-key"
# OR
export OPENAI_API_KEY="your-key"
LibreFang auto-detects which providers have API keys configured at boot. Any model whose provider is authenticated becomes immediately available. Local providers (Ollama, vLLM, LM Studio) require no key at all.
For Gemini specifically, either GEMINI_API_KEY or GOOGLE_API_KEY will work.
Picking a Provider Page
- Use Hosted APIs when you want a standard cloud LLM API with an API key.
- Use Local & Self-Hosted when models run on your machine or on your own inference server.
- Use Platforms & Managed Endpoints when the model sits behind a cloud platform, enterprise integration, or regional gateway.
- Use Developer Tools when LibreFang is talking to a local coding assistant runtime or CLI wrapper instead of a normal model endpoint.
- Use Provider Management for catalog behavior, aliases, routing, spend controls, and REST/channel introspection.