OpenRouter Provider on Bifrost

1) Chat Completions

Primary path at /v1/chat/completions. Bifrost delegates to the OpenAI implementation with special handling for reasoning models. OpenRouter supports all standard OpenAI chat parameters — see OpenAI Chat Completions for full parameter reference. See Chat Completions in Bifrost docs.

Reasoning parameter handling

Extended thinking on compatible models. Bifrost converts the internal reasoning object to OpenRouter's reasoning_effort. Reasoning models include gpt-oss-120b and other models with special reasoning content handling.

// Bifrost request
{
  "reasoning": {
    "effort": "high",
    "max_tokens": 10000
  }
}

// OpenRouter conversion
{
  "reasoning_effort": "high"
}

Cache control stripping

Anthropic-specific cache_control directives on message content blocks are removed before the request is sent upstream.

// cache_control stripped from content blocks
{
  "messages": [{
    "role": "user",
    "content": [{
      "type": "text",
      "text": "...",
      "cache_control": {"type": "ephemeral"}
    }]
  }]
}

Filtered parameters

Removed for OpenRouter compatibility: prompt_cache_key, verbosity (Anthropic-specific), store, and service_tier (OpenAI-specific).

• Messages, tools, responses, and streaming follow standard OpenAI formats
• Tool calling: full function definitions and execution on compatible models
• Streaming: SSE with usage tracking

curl -X POST http://localhost:8080/v1/chat/completions \
  -H "Content-Type: application/json" \
  -d '{
    "model": "openrouter/openai/gpt-4o",
    "messages": [{"role": "user", "content": "Hello"}]
  }'

2) Responses API

For compatible models on OpenRouter. Maps to upstream /v1/responses endpoint in beta.

curl -X POST http://localhost:8080/v1/responses \
  -H "Content-Type: application/json" \
  -d '{
    "model": "openrouter/anthropic/claude-3.5-sonnet",
    "messages": [{"role": "user", "content": "Hello"}]
  }'

3) Text Completions

Legacy completions endpoint. Maps to upstream /v1/completions. Supports streaming.

curl -X POST http://localhost:8080/v1/completions \
  -H "Content-Type: application/json" \
  -d '{
    "model": "openrouter/meta-llama/llama-3.1-70b-instruct",
    "prompt": "Hello, my name is",
    "max_tokens": 50
  }'

4) Embeddings

Vector embeddings via OpenRouter. Maps to upstream /v1/embeddings. Does not support streaming.

curl -X POST http://localhost:8080/v1/embeddings \
  -H "Content-Type: application/json" \
  -d '{
    "model": "openrouter/openai/text-embedding-3-small",
    "input": "Hello world"
  }'

5) List Models

Enumerate available models on OpenRouter. Maps to upstream /v1/models endpoint.