What the data for this category looks like
This category covers basic chemical continuous production entities including chlor-alkali, synthetic ammonia, methanol, and soda ash. Daily documents include operating procedures, process indicator cards, DCS trends, laboratory reports, interlock bypass records, special work permits, and shift logs. Data sources cover DCS/SIS systems, online analyzers, LIMS laboratory management systems, equipment condition monitoring, energy metering devices, and raw material tank monitoring nodes. Data update frequencies include real-time, per-shift, and daily. Different documents correspond to matching update cycles. Fields include process parameter values, interlock logic, work approval status, and more. Some documents include standardized units such as MPa, ℃, t/h.
What constraints do these characteristics impose on the knowledge base retrieval and recall process?
The data characteristics of this category create multiple constraints for the knowledge base retrieval and recall process. Real-time updated DCS trends and online analyzer data require retrieval logic to support sorting and recalling by update time, prioritizing matching the latest working condition data. Documents related to safety and compliance such as interlock bypass records and special work permits require strict matching of keywords and fields to avoid recalling irrelevant content. Multi-source, multi-update frequency data requires setting filtering rules by data type and update cycle to ensure the timeliness and matching accuracy of recalled documents. The large volume of documents and high number of standardized process parameters require limiting the number of single-recall results to avoid exceeding the context window.
How to configure the system
This document is based on FastGPT v4.17.0 configuration guidelines. Specific configuration items and recommendations are listed below:
| Configuration Item | Recommended Setting | Rationale |
|---|---|---|
similarity | 0.75–0.85 | Matches highly compliant documents such as interlock bypass records and process indicator cards, preventing low-relevance results from being included |
limit | Top 8 entries | Controls the number of single-recall results, adapting to the document volume and context window limits of basic chemical continuous production |
searchMode | fullTextRecall | The query examples emphasize identifiers or specialist terms; test this mode with representative data, not as a product default. |
usingReRank | Enabled | Re-sorts recalled results, prioritizing displaying process parameters and compliance records related to the bidding quotation scenario |
datasetSearchUsingExtensionQuery | Enabled | Automatically expands retrieval keywords to cover synonyms and unit variants of process parameters, supporting standardized field retrieval |
collectionIds | Collection IDs corresponding to documents of this category | Divides knowledge base collections by document type such as operating procedures and interlock records, accurately limiting the retrieval scope |
Three common mistakes
- Symptom: An "access denied" error appears during knowledge base search debugging, or the retrieval result is empty. Cause: The access permission of the corresponding document collection has not been granted to the current retrieval account, or the
collectionIdsparameter is configured incorrectly and does not include the target document collection. - Symptom: The number of results returned by a single retrieval exceeds the expected upper limit. Cause: The
limitparameter is not set correctly, or the configuredlimitvalue does not match the context window limit. - Symptom: Retrieved documents do not include the latest process indicator card data. Cause: Recall results are not filtered by update time, or the retrieval logic for sorting by update time is not enabled.
How to confirm correct configuration
- Perform a retrieval containing core process keywords, and check whether the similarity of the recalled results falls within the preset 0.75–0.85 range.
- View the document source and update time of the retrieval results, verify that filtering and sorting follow the configured rules, and prioritize displaying recently updated documents.
- Trigger an extended query retrieval, check whether relevant documents containing synonyms and unit variants are recalled, and confirm that the
datasetSearchUsingExtensionQueryconfiguration is active. - Check the permissions of the current retrieval account, verify that the configured knowledge base collection can be accessed normally, and confirm that the
collectionIdsparameter is configured correctly.
The values given are common starting points and should be measured against the reader's own samples.