What the Data for This Category Looks Like
Data sources for this category include enterprise procurement platforms, public resource trading platforms, supplier databases, contract management systems, and audit issue databases. The update frequency is daily. Daily documents cover procurement plans, requirement specifications, bidding documents, bid evaluation records, contracts, acceptance documents, and supplier complaint materials. Data fields include procurement identifiers, supplier information, budget amounts, procurement methods, bid evaluation results, performance and acceptance status, and more. Some documents are structured tables, some are unstructured text, and the size of individual data entries varies.
Constraints Imposed on Knowledge Base Retrieval and Recall
Decentralized data sources create differences between structured and unstructured document formats. This creates requirements for retrieval and recall adaptation. Daily updated data volume must match retrieval response speeds to avoid exceeding business processing time limits. Documents covering the entire procurement process must accurately match compliance review-related keywords. This ensures recall results cover core compliance points such as procurement methods and budget basis. Associated data from supplier databases and audit issue databases must be recalled synchronously. This ensures compliance reviews cover past risks. Field differences across different business departments must be unified during retrieval configuration. This avoids missing key information in recall results.
Configuration Settings
Reference only; verify against the target FastGPT release and deployment documentation.
| Configuration Item | Recommended Setting | Rationale |
|---|---|---|
similarity | 0.75–0.85 | Matches the compliance keyword matching accuracy of tender and procurement documents, balances recall coverage and irrelevant result filtering |
limit | Top 8–12 entries | Covers core compliance review requirements for documents across the entire procurement process, avoids excessive redundant data occupying context space |
searchMode | Hybrid retrieval | Adapts to mixed structured and unstructured document types, matches both keywords and semantic associations |
usingReRank | Enabled | Improves ranking accuracy for long texts such as bid evaluation records and contracts, ensures results with high compliance relevance are recalled first |
datasetSearchUsingExtensionQuery | Enabled | Generates extended query terms based on procurement plans and bidding documents, covers potential compliance risk points |
collectionIds | Bind procurement-related dataset IDs | Precise retrieval scope limitation, excludes interference from unrelated business data |
Three Common Misconfigurations
- Phenomenon: The number of retrieval results is far lower than the configured
limitvalue. Cause:collectionIdsare not correctly bound, leading the retrieval scope to cover non-procurement-related datasets and dilute valid data. - Phenomenon: The semantic matching accuracy of retrieval results is insufficient, and long-text documents rank lower. Cause:
usingReRankis not enabled, and no secondary re-ranking is performed on recall results, making it impossible to accurately match compliance keywords in long texts such as bid evaluation records and contracts. - Phenomenon: An "insufficient reference quota" error is returned when calling the knowledge base, or response time is excessively long. Cause: The
limitparameter is not set reasonably, and too many documents are recalled in a single retrieval, exceeding the platform context limit or processing upper limit, leading to reference failure or slower search speeds.
How to Verify Correct Configuration
- Enter the FastGPT knowledge base management page, check whether the bound datasets only include procurement-related data sources.
- Initiate a simulated retrieval, input "procurement method compliance review", and verify whether the number of recall results matches the
limitconfiguration. - Check the ranking of retrieval results, confirm that documents with high compliance relevance such as bid evaluation records and contracts are displayed first.
- Check whether retrieval results include associated data from supplier databases and audit issue databases, ensuring coverage of full-process compliance review requirements.
The values provided here are common starting points, and should be evaluated 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.