qianfan-knowledgebase-search

What This Skill Does

The qianfan-knowledgebase-search skill is a specialized agent tool designed to interface with the Baidu Qianfan Knowledgebase platform. It enables AI agents to perform sophisticated, context-aware information retrieval from private, user-defined knowledge repositories. By utilizing this skill, agents can transition from relying solely on pre-trained internal data to accessing live, domain-specific information, effectively creating a Retrieval-Augmented Generation (RAG) pipeline. The skill supports advanced search methodologies including semantic search (vector-based), fulltext search, and hybrid combinations, alongside reranking capabilities that ensure the most relevant information is prioritized for the agent's reasoning process.

Installation

To integrate this capability into your OpenClaw environment, execute the following command in your terminal: clawhub install openclaw/skills/skills/hannatao/qianfan-knowledgebase-search

Ensure you have your BAIDU_API_KEY set in your environment variables. Optionally, you may define QIANFAN_KNOWLEDGEBASE_IDS if you prefer to have a default knowledge base assigned to your agent's workflow.

Use Cases

This skill is indispensable for enterprise scenarios requiring private data access. Common use cases include:

• Customer Support Automation: Searching through product manuals, service contracts, and internal FAQ databases to resolve support tickets.
• Corporate Knowledge Management: Querying internal policy documents, HR handbooks, or technical specifications.
• Research Assistance: Analyzing large volumes of industry-specific documents uploaded to the Qianfan platform to generate summarized reports.
• Technical Documentation Navigation: Quickly surfacing API implementation details or troubleshooting steps from proprietary documentation.

Example Prompts

1. "Search the product documentation to find the troubleshooting steps for error code 503 and explain how to reset the connection."
2. "Based on our internal policy document, what is the standard procedure for requesting remote work equipment?"
3. "Retrieve the latest service level agreement details from the knowledge base and summarize the primary response time requirements."

Tips & Limitations

• Performance Optimization: Always tune the top_k and score_threshold parameters to balance between precision and comprehensiveness. If you receive too much noise, increase the threshold; if you lack context, increase the top_k.
• Hybrid Recall: Use recall_type="hybrid" for the most accurate results, as this combines the semantic understanding of vector search with the keyword precision of fulltext search.
• Data Privacy: Since this skill communicates with external APIs, ensure that sensitive documents are managed within the secure parameters of your Baidu Qianfan account and that your environment variables are handled securely.