What This Category of Data Looks Like
This category of data comes from the national unified construction project environmental impact assessment (EIA) declaration and approval system, ecological environment zoning control results, pollution discharge permit platform, and monitoring data. Update frequencies are irregular. Daily documents include environmental impact assessment reports or report forms, approval applications, public participation descriptions, technical evaluation opinions, and EIA approval documents. Fields include project category, EIA level, pollution prevention measures, public participation status, approval document number, and more. Most document structures use a hybrid format combining structured forms and long text.
Constraints Imposed by These Characteristics on Knowledge Base Retrieval and Recall
Scattered data sources are updated irregularly, so retrieval must cover homologous data across multiple systems to avoid inconsistent information between sources. Documents mix structured content and long text, so retrieval logic must adapt to different formats to ensure accurate matching of field-level and full-text content. EIA approval processes include compliance reviews, so retrieval results must link to specific approvals and technical opinions, which demands high relevance for recalled content. A single retrieval must cover multiple document types to avoid missing key compliance basis, and retrieval results must maintain timeliness to match irregularly updated data.
Configuration Settings
Reference only; verify against the target FastGPT release and deployment documentation.
| Configuration Item | Recommended Value | Rationale |
|---|---|---|
similarity | 0.75–0.85 | EIA approval requires highly relevant results to avoid irrelevant documents with low matching scores interfering with compliance review |
limit | Top 8 entries | Cover multiple types of EIA-related documents while controlling context length to avoid exceeding processing limits |
searchMode | Full-text retrieval | Adapt to mixed structured and long-text EIA documents, balancing recall of field and full-text content |
usingReRank | Enabled | Improve the ranking priority of highly relevant results to ensure key documents such as technical evaluation opinions and approval documents are returned first |
datasetSearchUsingExtensionQuery | Enabled | Supplement extended queries for fields such as project category and EIA level to improve recall accuracy of structured information |
Three Common Misconfigurations
- When retrieving from a knowledge base split from Word documents, image links can be identified, but answers cannot be generated based on image content. This occurs because Word splitting only extracts text content and does not bind image links to corresponding text blocks, so retrieval cannot recall context information associated with images.
- Retrieval results from the knowledge base automatically add citations at the end of answers, and "no permission to operate this conversation record" is displayed in some scenarios. This occurs because citation hiding settings are not configured, and conversation permissions are not adapted to the internal collaboration scenario of EIA approval.
- Single tabulation summary retrieval takes too long, causing delays in report generation. This occurs because a reasonable
limitparameter is not set orusingReRankis not enabled, and recalling too many low-relevance documents increases processing overhead.
How to Verify Proper Configuration
- Upload a typical environmental impact assessment report and EIA approval document, perform a retrieval test, and check that the returned results include both types of documents and the ranking meets expectations.
- Adjust the
similarityparameter to 0.8, and verify whether the matching degree of retrieval results meets the accuracy requirements of compliance review. - After enabling
datasetSearchUsingExtensionQuery, test retrieval for the project category field to confirm that corresponding documents can be recalled accurately. - Test the image link scenario in Word documents, and confirm that retrieval results can be linked to corresponding text blocks to support question answering based on image content.
Note: The configuration values provided here are common starting points. Adjust them based on specific local samples and use cases.
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.