What the data for this category looks like
The data sources for this category are tender announcements, tender documents, enterprise qualification databases, and supplier quotations. Data is updated once per project, and is entered synchronously when the tender task starts. Daily processed documents include tender documents, qualification certificates, business deviation tables, and technical deviation tables. Core fields cover tenderer qualification requirements, response clauses, quotation details, format specifications, etc. There are no unified fixed units; values adjust dynamically based on tender requirements.
What constraints do these characteristics impose on the knowledge base retrieval and recall link
The data characteristics of this category impose multiple constraints on the retrieval process. First, data is updated dynamically per project, requiring the retrieval process to support associating the latest tender documents, enterprise qualifications, and supplier quotations on demand to avoid calling expired content. Second, core documents include structured response files such as business deviation tables and technical deviation tables, which require precise matching of the tenderer’s clause requirements and format specifications, placing higher demands on the relevance and format adaptability of recalled content. In addition, data sources are scattered, requiring support for joint retrieval across multiple data sources, while also limiting the scope of recalled results to ensure only content that meets tender qualification and response requirements is returned.
Configuration Settings
The following configurations are based on FastGPT v4.17.0.
| Configuration Item | Recommended Setting | Rationale |
|---|---|---|
similarity | 0.75–0.85 | Matches the precision requirement for tender clauses, avoiding recall of irrelevant non-response content |
limit | Top 10 entries | Controls the number of results recalled per single retrieval, adapting to the page response length requirements of tender documents |
searchMode | Full-text retrieval | Covers matching needs for multiple types of text such as deviation tables and qualification items |
usingReRank | Enabled | Improves the sorting priority of highly relevant content, ensuring core response clauses are returned first |
datasetSearchUsingExtensionQuery | Enabled | Expands the retrieval scope to cover implicit format and clause requirements in tender announcements |
collectionIds | Bind corresponding datasets per project | Ensures only exclusive data for the current tender task is called, avoiding interference from cross-project content |
Three Common Misconfigurations
- Issue: The total character count of knowledge base retrieval results in workflow scenarios is less than 3000, or the number of recalled entries is far lower than expected. Cause: Confused the configuration upper limit parameters between independent knowledge base search and in-workflow knowledge base search, and failed to adjust corresponding settings for the current tender task.
- Issue: Retrieval results include historical qualification or quotation data unrelated to the current tender project. Cause: Failed to bind the exclusive dataset ID for the current tender task, and called the full public dataset, resulting in cross-project content being mixed in.
- Issue: Retrieved content does not match the tenderer’s clause response requirements, and contains a large number of non-targeted general document fragments. Cause: The re-ranking function is not enabled, or the similarity threshold is set too low, causing low-relevance content to be recalled first.
How to Verify Proper Configuration
- Navigate to the corresponding workflow or knowledge base configuration page, check the
limitparameter setting value, and confirm it matches the response length requirement of the single-project tender task. - Enter keywords from a tender clause, perform a retrieval, and verify that the number of returned results and similarity meet the preset configuration.
- Check the bound
collectionIds, confirm they only include exclusive datasets such as the current tender task’s tender documents and qualification database. - After enabling the
usingReRankfunction, perform another retrieval, and confirm that highly relevant response clause content appears at the top of the results.
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 2026-10-19.