Skip to main content

🔄 Transparent Provider Switching

Your application talks to one LLM API (e.g., OpenAI), but you want to route traffic to a different provider (e.g., Anthropic) without modifying any application code.

📋 Scenario​

Your app sends OpenAI-format requests to http://127.0.0.1:11510/v1/chat/completions. You want these requests forwarded to Anthropic with format conversion.

⚙️ Setup​

1️⃣ Configure Downstreams​

downstreams:
  - id: openai
    name: OpenAI
    api_formats: [openai]
    base_url: https://api.openai.com/v1
    api_key: sk-proj-...
    output_model_ids:
      - gpt-4o

  - id: anthropic
    name: Anthropic
    api_formats: [anthropic]
    base_url: https://api.anthropic.com
    api_key: sk-ant-...
    output_model_ids:
      - claude-sonnet-4-20250514

2️⃣ Create a Rule with Format Conversion (Optional)​

rules:
  - id: chat-anthropic
    name: Chat Completions -> Anthropic
    pattern_path: /v1/chat/completions
    match_format: [openai]
    match_downstream_format: [anthropic]
    match_downstreams: [anthropic]
    pipeline_config:
      - plugin_id: openai2anthropic
    is_enabled: true

The openai2anthropic plugin:

  • Converts OpenAI Chat Completion request format to Anthropic Messages format
  • Maps model names (e.g., gpt-4o → claude-sonnet-4)
  • Extracts system prompts into Anthropic's system field
  • Handles streaming responses (SSE event conversion)

Tip: Since the Anthropic downstream declares api_formats: [anthropic], auto-translation kicks in automatically — the explicit openai2anthropic plugin in the rule is redundant but harmless (deduplication prevents double-conversion). The format filters on this rule ensure it only matches when the request is OpenAI format and the resolved downstream is Anthropic.

3️⃣ Point Your App​

Configure your application to use Tresor as the endpoint:

# Python example
client = OpenAI(
    base_url="http://127.0.0.1:11510/v1",
    api_key="sk-proxy-123"
)

response = client.chat.completions.create(
    model="gpt-4o",
    messages=[{"role": "user", "content": "Hello!"}]
)

The app thinks it's talking to OpenAI. Tresor converts the request, forwards it to Anthropic, converts the response back, and returns it — all transparently.