What this category of data looks like
Data for this category comes from government portals, document management systems, government information disclosure directories, policy document repositories, and press conference materials. The update frequency is daily. Daily documents include policy documents, policy interpretation drafts, public disclosure directories, information release approval forms, press releases, and website column update records. Most document structures include title, official document number, issuing authority, drafting date, main body content, and public attribute label. All fields follow relevant government disclosure standards, with no custom business fields.
What constraints these characteristics impose on the source citation and traceability step
Data sources are scattered and must align with government disclosure standards. A unique source identifier must be bound to each document to ensure accurate traceability. Daily incremental updates must synchronize with the index to avoid mixing old and new data. Documents include multiple types such as policy files and approval forms. Some documents have complex structural hierarchies. Original title hierarchies must be preserved to accurately locate cited passages. Public disclosure directories must be linked to policy release paths. Directory entries must be bound to corresponding policy documents to ensure citations can be traced back to official release channels.
How to configure
Reference only; verify against the target FastGPT release and deployment documentation.
| Configuration Item | Recommended Value | Rationale |
|---|---|---|
| Chunk Length | 800–1200 characters | Policy document main bodies are mostly long paragraphs; this range preserves complete policy clauses and title hierarchies |
| Overlap Length | 100–200 characters | Prevents policy clauses from being truncated by chunking, ensures cited passages can be fully linked to the original text |
| Single Search Limit | Top 10 entries | Matches the release volume of government disclosure data, covers the scope of commonly used policy documents |
quoteMaxToken | 800–1200 token | Ensures cited content can fully display fragments of original policy text, meets traceability requirements |
responseData | Enabled | Forces return of original cited data from knowledge base searches for traceability display |
detail | Enabled | Displays complete metadata of cited documents, including key government disclosure information such as issuing authority and drafting date |
Three common mistakes
- After configuring dynamic knowledge base variable calls, returned results have no cited documents. The cause is that the
responseDataparameter is not enabled, so search citation data is not forced to be returned. - The citation module is not displayed when the terminal is exposed externally. The cause is that the
detailparameter is not enabled, or the interface switch for citation display is not configured. - When long documents use title hierarchies as separators, citation matching accuracy is low. The cause is that reasonable chunk overlap length is not set, resulting in policy clauses being truncated by chunking, making accurate matching of original text passages impossible.
How to confirm the configuration is correct
- Launch a policy Q&A test, check if the returned results include a citation module, confirm that the
responseDataparameter is active. - View the metadata in the citation module, confirm that it includes fields related to government disclosure such as issuing authority and drafting date, confirm that the
detailparameter is enabled. - Adjust the chunk length and overlap length, launch multiple tests, verify that cited passages can fully match original policy clauses.
- Call the corresponding API interface, check if the returned fields include complete citation-related data, confirm that the interface configuration is correct.
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.