Skip to main content
WeKnora is an open-source document understanding and semantic retrieval framework based on Large Language Models (LLMs), developed by Tencent. It is specifically designed for document scenarios with complex structures and heterogeneous content.
The framework adopts a modular architecture, integrating multimodal preprocessing, semantic vector indexing, intelligent recall, and large model generation inference to build an efficient and controllable document Q&A workflow. The core retrieval process is based on the RAG (Retrieval-Augmented Generation) mechanism, which combines contextually relevant passages with language models to achieve high-quality answer generation. LangBot currently supports two application types for WeKnora: Knowledge Base Q&A (chat) and Agent Intelligent Conversation (agent).

​
Obtaining API Key from WeKnora

Please deploy the WeKnora service and complete knowledge base creation and document upload according to the WeKnora Official Documentation. weknora After deployment, access the WeKnora WebUI (default http://localhost:80), navigate to Settings -> API Keys page, click Generate API Key and copy it for safekeeping.
  • API Key: Used for authentication. Please keep it secure and do not expose it.
  • Knowledge Base ID: You can view the ID (formatted as kb-00000001) for each knowledge base on the knowledge base management page, which can be specified in LangBot to invoke the corresponding knowledge base.
After clicking, you can obtain the basic URL and key information. weknora
  • If LangBot and WeKnora are deployed on the same host and both use Docker deployment, you can refer to the article: Network Configuration Details. Please add the WeKnora container to langbot-network when starting, and set the API Base URL in LangBot configuration to the container interconnection address, for example http://weknora-app:8080/api/v1.
  • For other scenarios, please consult your company’s operations team.

​
Obtaining WeKnora Agent Information

weknora On the ‘Agents’ page, click on the Agent you want to select to enter its details page. In Agent Configuration, you can view the Agent ID (formatted as builtin-smart-reasoning), and in Tool List, you can view the tool names (such as web-search, file-search, etc.) that the Agent can call. These information will be needed when configuring LangBot. weknora At this point, you have prepared the necessary information to connect to the WeKnora API. Next, please follow the steps below to complete the configuration in LangBot.

​
Configuring LangBot

Open the LangBot WebUI page, add a new pipeline or switch to the AI Capabilities configuration page in an existing pipeline. weknora
  1. Select WeKnora API in Runner.
  2. In the runner configuration items, fill in:
    • API Base URL: The API address of the WeKnora service, default http://localhost:8080/api/v1.
    • API Key: Paste the API Key generated on WeKnora.
    • Application Type: Select agent (Agent Intelligent Conversation, supports tool invocation, web search, chain of thought) or chat (Knowledge Base RAG Q&A).
    • Agent ID: Built-in Agent identifier, common values:
      • builtin-smart-reasoning: Intelligent Reasoning Agent (recommended for agent mode)
      • builtin-quick-answer: Quick Answer Agent (recommended for chat mode)
    • Knowledge Base ID List: Fill in the knowledge base IDs to retrieve from (formatted as kb-00000001), multiple can be filled.
    • Enable Web Search: Only effective in agent mode, allows the Agent to call Web search tools when enabled.
    • Timeout (seconds): Request timeout, default 120.
    • Default Prompt: The default question used when the user does not send text content.
  3. Click Save to complete the configuration.
After completion, you can invoke WeKnora in the pipeline to complete knowledge base-based intelligent conversation.

​
Output Processing

  • When using the agent application, if track-function-calls is enabled in LangBot pipeline Output Processing, a message Calling function xxx will be output to the user each time WeKnora executes a tool call.
  • The thinking (chain of thought) content returned by WeKnora will be sent wrapped in <think> tags, which can be collapsed and displayed on the frontend.
  • The references (knowledge base citation sources) returned by WeKnora will be attached after the answer as reference source messages.

​
Frequently Asked Questions

  • Save failed or invocation error: Please verify that the API Key and API Base URL are correct. The URL must include /api/v1 at the end.
  • No knowledge base content returned: Please verify that the Knowledge Base ID is correct, and that documents have been successfully uploaded and parsed in that knowledge base.
  • Cannot select WeKnora API: Please verify that your current LangBot version supports this runner; if not supported, please upgrade to a version that includes the WeKnora runner.
For more pipeline general configuration instructions, please refer to Modifying Conversation Pipeline Configuration.