Hosted Provider APIs

This page covers internet-hosted model providers that authenticate with an API key and connect through LibreFang's native Anthropic/Gemini drivers or the shared OpenAI-compatible driver.

Included Providers

• Anthropic
• OpenAI
• Google Gemini
• DeepSeek
• Groq
• OpenRouter
• Mistral AI
• Together AI
• Fireworks AI
• Perplexity AI
• Cohere
• Cerebras
• SambaNova
• Hugging Face
• xAI
• Alibaba Coding Plan
• Moonshot (Kimi)
• Novita AI
• AWS Bedrock

Anthropic

Available Models:

• claude-opus-4-20250514 (Frontier)
• claude-sonnet-4-20250514 (Smart)
• claude-haiku-4-5-20251001 (Fast)

Setup:

1. Sign up at console.anthropic.com
2. Create an API key under Settings > API Keys
3. export ANTHROPIC_API_KEY="sk-ant-..."

OpenAI

Available Models:

• gpt-4.1 (Frontier)
• gpt-4o (Smart)
• o3-mini (Smart)
• gpt-4.1-mini (Balanced)
• gpt-4o-mini (Fast)
• gpt-4.1-nano (Fast)

Setup:

1. Sign up at platform.openai.com
2. Create an API key under API Keys
3. export OPENAI_API_KEY="sk-..."

Google Gemini

Available Models:

• gemini-2.5-pro (Frontier)
• gemini-2.5-flash (Smart)
• gemini-2.0-flash (Fast)

Setup:

1. Go to aistudio.google.com
2. Get an API key (free tier included)
3. export GEMINI_API_KEY="AIza..." or export GOOGLE_API_KEY="AIza..."

Notes: The Gemini driver is a fully native implementation. It is not OpenAI-compatible. Model goes in the URL path, system prompt via systemInstruction, tools via functionDeclarations, streaming via streamGenerateContent?alt=sse.

DeepSeek

Available Models:

• deepseek-chat (Smart) -- DeepSeek V3
• deepseek-reasoner (Smart) -- DeepSeek R1, no tool support

Setup:

1. Sign up at platform.deepseek.com
2. Create an API key
3. export DEEPSEEK_API_KEY="sk-..."

Groq

Available Models:

• llama-3.3-70b-versatile (Balanced)
• mixtral-8x7b-32768 (Balanced)
• llama-3.1-8b-instant (Fast)
• gemma2-9b-it (Fast)

Setup:

1. Sign up at console.groq.com
2. Create an API key
3. export GROQ_API_KEY="gsk_..."

Notes: Groq runs open-source models on custom LPU hardware. Extremely fast inference. Free tier has rate limits but is very usable.

OpenRouter

Available Models:

• openrouter/google/gemini-2.5-flash (Smart) -- cheap, fast, 1M context (default)
• openrouter/anthropic/claude-sonnet-4 (Smart) -- strong reasoning + tools
• openrouter/openai/gpt-4o (Smart) -- GPT-4o via OpenRouter
• openrouter/deepseek/deepseek-chat (Smart) -- DeepSeek V3
• openrouter/meta-llama/llama-3.3-70b-instruct (Balanced) -- Llama 3.3 70B
• openrouter/qwen/qwen-2.5-72b-instruct (Balanced) -- Qwen 2.5 72B
• openrouter/google/gemini-2.5-pro (Frontier) -- Gemini 2.5 Pro
• openrouter/mistralai/mistral-large-latest (Smart) -- Mistral Large
• openrouter/google/gemma-2-9b-it (Fast) -- Gemma 2 9B, free
• openrouter/deepseek/deepseek-r1 (Frontier) -- DeepSeek R1 reasoning

Setup:

1. Sign up at openrouter.ai
2. Create an API key under Keys
3. export OPENROUTER_API_KEY="sk-or-..."

Notes: OpenRouter is a unified gateway to 200+ models from many providers. Model IDs use the upstream format (e.g. google/gemini-2.5-flash). You can use any model from OpenRouter's catalog by specifying the full model path with the openrouter/ prefix.

Mistral AI

Available Models:

• mistral-large-latest (Smart)
• codestral-latest (Smart)
• mistral-small-latest (Fast)

Setup:

1. Sign up at console.mistral.ai
2. Create an API key
3. export MISTRAL_API_KEY="..."

Together AI

Available Models:

• meta-llama/Meta-Llama-3.1-405B-Instruct-Turbo (Frontier)
• Qwen/Qwen2.5-72B-Instruct-Turbo (Smart)
• mistralai/Mixtral-8x22B-Instruct-v0.1 (Balanced)

Setup:

1. Sign up at api.together.ai
2. Create an API key
3. export TOGETHER_API_KEY="..."

Fireworks AI

Available Models:

• accounts/fireworks/models/llama-v3p1-405b-instruct (Frontier)
• accounts/fireworks/models/mixtral-8x22b-instruct (Balanced)

Setup:

1. Sign up at fireworks.ai
2. Create an API key
3. export FIREWORKS_API_KEY="..."

Perplexity AI

Available Models:

• sonar-pro (Smart) -- online search-augmented
• sonar (Balanced) -- online search-augmented

Setup:

1. Sign up at perplexity.ai
2. Go to API settings and generate a key
3. export PERPLEXITY_API_KEY="pplx-..."

Notes: Perplexity models have built-in web search. They do not support tool use.

Cohere

Available Models:

• command-r-plus (Smart)
• command-r (Balanced)

Setup:

1. Sign up at dashboard.cohere.com
2. Create an API key
3. export COHERE_API_KEY="..."

Cerebras

Available Models:

• cerebras/llama3.3-70b (Balanced)
• cerebras/llama3.1-8b (Fast)

Setup:

1. Sign up at cloud.cerebras.ai
2. Create an API key
3. export CEREBRAS_API_KEY="..."

Notes: Cerebras runs inference on wafer-scale chips. Ultra-fast and ultra-cheap ($0.06/M tokens for both input and output on the 70B model).

SambaNova

Available Models:

• sambanova/llama-3.3-70b (Balanced)

Setup:

1. Sign up at cloud.sambanova.ai
2. Create an API key
3. export SAMBANOVA_API_KEY="..."

Hugging Face

Available Models:

• hf/meta-llama/Llama-3.3-70B-Instruct (Balanced)

Setup:

1. Sign up at huggingface.co
2. Create a token under Settings > Access Tokens
3. export HF_API_KEY="hf_..."

xAI

Available Models:

• grok-2 (Smart) -- supports vision
• grok-2-mini (Fast)

Setup:

1. Sign up at console.x.ai
2. Create an API key
3. export XAI_API_KEY="xai-..."

Alibaba Coding Plan

Available Models:

• alibaba-coding-plan/qwen3.6-plus (Smart) — vision support, 1M context
• alibaba-coding-plan/qwen3.5-plus (Smart) — vision support, 1M context
• alibaba-coding-plan/qwen3-coder-plus (Smart) — 1M context
• alibaba-coding-plan/qwen3-coder-next (Frontier) — 262K context
• alibaba-coding-plan/qwen3-max-2026-01-23 (Frontier) — 262K context
• alibaba-coding-plan/glm-5 (Frontier) — 202K context
• alibaba-coding-plan/glm-4.7 (Smart) — 202K context
• alibaba-coding-plan/kimi-k2.5 (Smart) — vision support, 262K context
• alibaba-coding-plan/MiniMax-M2.5 (Balanced) — 196K context

Setup:

1. Subscribe at Coding Plan page
2. Get plan-specific API key (format: sk-sp-xxxxx)
3. export ALIBABA_CODING_PLAN_API_KEY="sk-sp-..."

Quota Limits (subscription-based, not token-based):

• 90,000 requests/month (resets on subscription anniversary date at 00:00 UTC+8)
• 45,000 requests/week (resets every Monday at 00:00 UTC+8)
• 6,000 requests per 5 hours (sliding window — each request resets exactly 5 hours after use)

Notes:

• Uses OpenAI-compatible API format
• Plan-specific API key (sk-sp-xxxxx) differs from pay-as-you-go DashScope key
• Metering shows $0 cost (subscription-based), but token usage still tracked
• Monitor request quotas via Alibaba Cloud console
• Not for automated scripts or batch API calls — coding tools only
• For more info: Official Documentation

Moonshot (Kimi)

Setup:

1. Sign up at Moonshot Platform (mainland) or Moonshot AI (international)
2. Create an API key from the console
3. export MOONSHOT_API_KEY="sk-..."

Minimal config.toml:

[default_model]
provider = "moonshot"
model = "moonshot-v1-128k"

Capabilities: Chat completions, tool use, vision (on kimi-latest / vision-capable models), and large-context windows up to 128K. File uploads supported via the /files endpoint for multi-document RAG.

Notes: Kimi models are operated by Moonshot AI. Use the kimi alias for shorter config.

Novita AI

Setup:

1. Sign up at novita.ai
2. Generate an API key from the console
3. export NOVITA_API_KEY="..."

Minimal config.toml:

[default_model]
provider = "novita"
model = "<model-id-from-novita-catalog>"

Capabilities: Tools and streaming are supported via the shared OpenAI-compatible driver. Vision support depends on the upstream model — check the Novita model card before enabling it.

Notes: Novita exposes its catalog over the OpenAI Chat Completions API format. Auto-detection picks Novita up automatically when NOVITA_API_KEY is set; no explicit provider line is required if you only want a fallback.

AWS Bedrock

Setup:

1. In the AWS console, create a Bedrock API Key (long-lived bearer token). SigV4 is not used by this driver.

2. Decide on a region with the model you want enabled in the Bedrock model catalog.

3. Export the credentials:

   export AWS_BEARER_TOKEN_BEDROCK="..."
   export AWS_REGION="us-east-1"
   

   CopyCopied!

Minimal config.toml:

[default_model]
provider = "bedrock"
model = "anthropic.claude-sonnet-4-20250514-v1:0"

The model field is passed verbatim into the endpoint path, so use the full Bedrock model identifier (including region prefixes like eu. or inference-profile IDs when required by the region).

Capabilities:

• Tool use (function calling) via the Converse toolConfig shape, with full message-shape repair for toolResult / toolUse pairing.
• Streaming and non-streaming completions.
• Vision is not wired today: image content blocks are dropped before the request because Bedrock Converse rejects the LibreFang Image / ImageFile shapes.
• Prompt-cache token counters are not surfaced — Bedrock Converse does not expose cache_creation_input_tokens / cache_read_input_tokens separately, so they report as zero in metering.

Notes: Region resolution order is the explicit driver argument, then AWS_REGION, then AWS_DEFAULT_REGION, then us-east-1. If you want a regional inference profile (e.g. eu.anthropic.…), set AWS_REGION to a matching region (eu-west-1, etc.) and use the prefixed model ID.