Anthropic-Compatible RAG

Endpoint: POST https://api.yourgpt.ai/chatbot/v1/anthropic/v1/messages

Set base_url to https://api.yourgpt.ai/chatbot/v1/anthropic in the Anthropic SDK — it appends /v1/messages automatically. Add a yourgpt:knowledgebase server tool and the endpoint handles retrieval automatically.

import anthropic

client = anthropic.Anthropic(
    api_key="apk-your-api-key",
    base_url="https://api.yourgpt.ai/chatbot/v1/anthropic",
)

response = client.messages.create(
    model="claude-sonnet-4-6",
    max_tokens=1024,
    system="You are a helpful support agent.",
    messages=[{"role": "user", "content": "What is your refund policy?"}],
    tools=[
        {"type": "yourgpt:knowledgebase", "parameters": {"limit": 5}}
    ],
)

print(response.content[0].text)

Authentication

Pass your apk-... project API key as the Anthropic api_key. The SDK sends it in x-api-key, which YourGPT accepts automatically.

Request limits

model, max_tokens, and messages are all required. At least one user message must contain text.

Server tools

Add these to the tools array alongside any of your own Anthropic tool definitions. Your own tools are passed through to the model unchanged — see User-defined tools.

yourgpt:knowledgebase

Injects a search_knowledgebase tool into Claude. Claude calls it when it decides retrieval is needed.

{
  "type": "yourgpt:knowledgebase",
  "parameters": {
    "limit": 5,
    "mode": "tool_only"
  }
}

tool_only — best when the conversation mixes KB and non-KB questions. Claude only searches when it decides to.

hybrid — best for always-on KB answers (support bots, FAQ). Context is retrieved up front so Claude can answer immediately. Costs slightly more per request.

yourgpt:web_search

Injects a web_search tool. Claude calls it for live information outside your knowledgebase. Each result includes title, URL, and up to 3,000 chars of text.

{
  "type": "yourgpt:web_search",
  "parameters": {
    "max_results": 5
  }
}

User-defined tools

Any tool in tools[] without a yourgpt:* type is your own tool. YourGPT forwards it to Claude unchanged and, when Claude calls it, hands the call back to you exactly as the Anthropic API would — stop_reason: "tool_use" with a tool_use content block. Execute it and POST back a tool_result block to continue the conversation.

Knowledgebase and web search run server-side and never appear as tool_use blocks in your response.

Response

Standard Anthropic response shape plus a sources array.

{
  "content": [{ "type": "text", "text": "..." }],
  "stop_reason": "end_turn",
  "usage": { "input_tokens": 312, "output_tokens": 48 },
  "sources": [
    {
      "type": "knowledgebase",
      "content": "You can reset your password from...",
      "score": 0.921,
      "doc_id": "doc_abc123"
    },
    {
      "type": "web",
      "title": "EU Consumer Rights",
      "url": "https://example.com/eu-consumer-rights",
      "text": "Consumers in the EU have..."
    }
  ]
}

Every entry is tagged with type — knowledgebase chunks and web results share one array.

sources is empty when nothing was retrieved, and is de-duplicated — the same document or URL appears at most once.

Streaming

Set "stream": true — response is standard Anthropic SSE format. sources is attached to the message_delta event.

Examples

Rate limits & errors

Errors returned by Anthropic itself are forwarded with their original status and message.