Message Transforms

BluesMinds can transform message formats on the fly, allowing you to use a single OpenAI-compatible format and have it routed to providers that use different native protocols (Anthropic Messages API, Google Gemini API, etc.).

OpenAI → Anthropic Transform

When you route a request to a Claude model via the standard /v1/chat/completions endpoint, BluesMinds automatically converts the OpenAI message format to Anthropic's Messages API format.

You write:

{
  "model": "claude-sonnet-4-5",
  "messages": [
    {"role": "system", "content": "You are a concise assistant."},
    {"role": "user", "content": "Summarize AI in one sentence."}
  ]
}

BluesMinds internally sends:

{
  "model": "claude-sonnet-20240229",
  "system": "You are a concise assistant.",
  "messages": [
    {"role": "user", "content": "Summarize AI in one sentence."}
  ],
  "max_tokens": 1024
}

Your code never needs to know about Anthropic's format.

Native Anthropic Messages Endpoint

If you prefer to use Anthropic's native message format directly, BluesMinds also exposes it:

POST https://api.bluesminds.com/v1/messages
Authorization: Bearer sk-...
Content-Type: application/json

{
  "model": "claude-sonnet-4-5",
  "messages": [{"role": "user", "content": "Hello Claude!"}],
  "max_tokens": 256
}

This is useful for applications already built against the Anthropic SDK.

OpenAI → Gemini Transform

Similarly, requests to Gemini models via /v1/chat/completions are automatically transformed into Google's generateContent format internally. The response is converted back to the OpenAI format before being returned to you.

System Messages

All providers handle system messages differently. BluesMinds normalizes them:

Provider	Handling
OpenAI	Native {"role": "system", ...}
Anthropic	Extracted to top-level "system" field
Google Gemini	Mapped to system_instruction

Just use standard OpenAI {"role": "system"} in your messages array — BluesMinds handles the rest.

Streaming Normalization

All providers use different streaming formats. BluesMinds normalizes all of them to OpenAI's SSE format:

data: {"id":"chatcmpl-...","choices":[{"delta":{"content":"Hello"}}]}
data: {"choices":[{"delta":{"content":" world"}}]}
data: [DONE]

You never need to handle Anthropic event streams or Google SSE formats.

Limitations

• Embeddings are not currently available
• Audio STT (speech-to-text) transforms are provider-specific; check model availability
• Some provider-specific parameters (e.g., Anthropic's top_k) are not forwarded through the OpenAI compat layer — use the native endpoint if you need them