Actions13
Overview
This node, named "FastGPT," provides an AI-powered knowledge base search functionality. Specifically, the "知识库检索" (Knowledge Base Search) operation allows users to query a selected knowledge dataset using various search modes such as semantic vector search, full-text search, or a hybrid approach combining both. This is useful for retrieving relevant information from large collections of documents or datasets based on user input text.
Common scenarios include:
- Searching internal company knowledge bases to quickly find relevant documents or answers.
- Implementing intelligent FAQ systems that return precise results based on natural language queries.
- Enhancing customer support by integrating semantic search over product manuals or help articles.
Practical example:
A user inputs a question or keyword phrase, selects a specific knowledge dataset, and chooses a search mode (e.g., semantic embedding). The node returns the most relevant documents or passages ranked by similarity or relevance scores.
Properties
| Name | Meaning |
|---|---|
| 选择知识库 | Select the knowledge dataset to search within. Options are dynamically loaded from available datasets filtered by type "dataset". |
| 检索文本 | The text string to search for in the selected knowledge base. |
| 最大Tokens数量 | Maximum number of tokens to consider in the search result. |
| 最低相关度 | Minimum similarity threshold (0 to 1) for returned results; filters out less relevant matches. |
| 类型 | Search mode to use: - 语义检索 (embedding): Semantic vector-based search. - 全文检索 (fullTextRecall): Traditional full-text keyword search. - 混合检索 (mixedRecall): Combination of semantic and full-text search with ranking fusion. |
| 使用ReRank | Boolean flag to enable re-ranking of search results for improved ordering. |
Output
The node outputs JSON data containing the search results from the knowledge base. The structure typically includes matched documents or entries along with metadata such as similarity scores or relevance rankings depending on the search mode used.
If binary data were involved (not indicated here), it would represent attachments or files retrieved from the knowledge base, but this node focuses on textual search results.
Dependencies
- Requires an API key credential for the FastGPT platform.
- Needs access to the FastGPT API endpoint specified in the credentials (
baseUrl). - The node dynamically loads knowledge datasets via an API call to
/core/dataset/list. - Proper configuration of the API authentication token is necessary for successful requests.
Troubleshooting
- Empty or no results: Check if the selected knowledge dataset contains indexed data. Also verify the search text is not empty and the similarity threshold is not set too high.
- API errors or timeouts: Ensure the API key credential is valid and the base URL is correctly configured. Network connectivity issues can also cause failures.
- Invalid property values: For example, setting
最大Tokens数量to an excessively high number might cause performance issues or API rejections. - Search mode mismatch: Using full-text search on datasets not optimized for keyword indexing may yield poor results; try semantic or mixed modes instead.
Links and References
- FastGPT Official Documentation (replace with actual URL)
- Concepts of semantic search and embeddings: https://en.wikipedia.org/wiki/Semantic_search
- Full-text search basics: https://www.elastic.co/what-is/full-text-search