What the data for this category looks like
Data sources include the New Energy Project Filing Specifications of the National Energy Administration, provincial development and reform commission grid connection process guidelines, project pre-approval form templates, and local government service center handling rules. Updates are synchronized dynamically with national and local energy policy adjustments and newly added project filing cases. Most documents use a chaptered clause structure, containing fields such as approval node serial numbers, material requirements, handling time limits, and responsible departments. Units include working days, megawatts, ten thousand yuan, and similar units.
What constraints these characteristics impose on the knowledge base retrieval and recall link
The chaptered long clause structure requires avoiding context breaks during retrieval, to ensure complete association of approval nodes. Clear field and unit requirements demand precise keyword matching during retrieval, to avoid result deviations caused by unit confusion. Format differences across multiple source documents require unified preprocessing logic to ensure retrieval consistency. Dynamically updated data requires the knowledge base synchronization frequency to match the pace of policy adjustments, to prevent expired policy content from being recalled.
How to Configure Settings
Reference only; verify against the target FastGPT release and deployment documentation.
| Configuration Item | Recommended Value | Rationale |
|---|---|---|
similarity | 0.75–0.85 | Matches the accuracy requirements for new energy business handling, filters low-relevant government content |
limit | Top 6 entries | Covers the number of core approval nodes for a single new energy project handling, ensures complete contextual support for guidance |
usingReRank | Enabled | Optimizes sorting results after long-text retrieval, improves recall priority for highly relevant approval nodes |
searchMode | Semantic search | Adapts to natural language user query scenarios, matches fuzzy query needs for handling steps |
datasetSearchUsingExtensionQuery | Enabled | Supplements business scenario keywords, improves recall accuracy for new energy-specific documents |
collectionIds | Bind corresponding dataset IDs by project type | Targetedly retrieves new energy development-specific documents, excludes irrelevant government content |
Three Common Misconfigurations
- Phenomenon: After upgrading a private deployment version, clicking the knowledge base file details displays "Invalid dataset file key". Cause: The secret key configuration associated with the dataset was not updated synchronously during the upgrade, causing the original file index to fail to match the new system secret key.
- Phenomenon: Pulling the document library via the API returns a large number of non-new energy development government content. Cause: The
collectionIdsparameter was not specified correctly, causing the retrieval scope to cover all government documents, which interferes with business guidance results. - Phenomenon: When a user asks "grid connection process", the recall results lack core material list items. Cause:
datasetSearchUsingExtensionQuerywas not enabled, and scenario-related keywords were not supplemented, causing some associated documents to not be accurately recalled.
How to Confirm Proper Configuration
- Enter the FastGPT knowledge base management interface, verify whether the bound
collectionIdscorrespond to the new energy development-specific dataset. - Initiate a test query, such as "new energy project filing handling process", check whether the sorting of returned results conforms to business node logic.
- Call the API interface to pull the document library, check whether the returned results only contain business handling content related to new energy development.
- After upgrading the deployment version, click any knowledge base file details, confirm that the "Invalid dataset file key" error prompt does not appear on the interface.
The values provided are common starting points and should be measured against the reader's own samples.
Question material comes from public community discussions. Configuration values are common starting points and should be measured against your own samples. Verified on 2026-09-28.