What this category’s data looks like
This category primarily covers rubber product enterprises including tire, seal, and hose manufacturers. Daily documents include formula sheets, weighing records, mixing curves, rubber material quick test reports, vulcanization process cards, mold records, and finished product performance reports. Data sources include formula systems, raw material batch ledgers, internal mixers, Mooney viscometers, rheometers, vulcanizers, and laboratory equipment. Updates occur synchronously for each raw material truck, each batch of rubber compound, and each mold run. Document fields include rubber type ratio, additive dosage, temperature segment parameters, batch numbers, equipment models, timestamps, and more. Some documents are structured tables, while others are long-text curve data.
Constraints for Knowledge Base Retrieval and Recall
High-frequency updates require the knowledge base to support incremental synchronization, to avoid performance loss caused by full reindexing. Mixed document types and non-uniform formats require retrieval configurations to support matching both structured tables and long-text curve data. Auxiliary fields such as batch numbers and equipment models require retrieval to balance semantic matching and precise field matching, preventing recall of unrelated batch process data. The update cadence of each raw material truck, rubber batch, and mold run requires retrieval response speed to match the real-time needs of bid preparation. Errors in this category’s data can lead to product scrapping or failure, so recall results must have relatively high precision to avoid incorrect formulas or processes being used for quotation and proposal preparation.
How to Configure Settings
These configurations are based on FastGPT v4.17.0.
| Configuration Item | Recommended Setting | Rationale |
|---|---|---|
similarity | 0.75–0.85 | Match professional terminology and parameters related to rubber compounding and vulcanization, filter irrelevant documents with low similarity |
limit | Top 6 entries | Cover multi-batch process and formula data to support multi-dimensional reference for bid proposals and quotations |
searchMode | embedding | Prioritize matching document semantic content, adapt to text retrieval needs for professional formulas and processes |
usingReRank | Enabled | For long-text content such as mixing curves and vulcanization process cards, reordering can improve precise matching accuracy |
datasetSearchUsingExtensionQuery | Enabled | Supplement semantic expansion of auxiliary fields such as batch numbers and equipment models, expand matching coverage |
collectionIds | Specify collection IDs for the mixing process library and vulcanization parameter library | Isolate category-specific documents to avoid interference from other business data on retrieval results |
Three Common Mistakes
- Phenomenon: Calling the
apiCollectioninterface to upload rubber compounding documents returnsInvalid URL, code: 500. Cause: The API access domain name of the knowledge base collection is not configured correctly, or the incomingcollectionIdsparameter format does not meet requirements. - Phenomenon: Retrieval results only match the main content of chunks, and do not associate with auxiliary fields such as batch numbers and equipment models. Cause: The
datasetSearchUsingExtensionQueryconfiguration is not enabled, or auxiliary field content is not fully retained during chunking. - Phenomenon: The generated bid document or quotation reply includes reference links to knowledge base content. Cause: The source display configuration is not disabled in the conversation interface, or reference link output is not blocked in the retrieval configuration.
How to Confirm Configurations Are Correctly Set
- Upload 3 to 5 typical documents of this category (formula sheets, vulcanization process cards) and run a retrieval test, verify the relevance and field completeness of recall results.
- View retrieval logs to confirm the similarity of matching results falls within the preset range, adjust the
similarityvalue as needed. - Call the knowledge base retrieval interface to verify that returned results include matching information for auxiliary fields such as batch numbers and equipment models.
- Test the complete conversation generation flow to confirm the reply does not include reference links to knowledge base sources.
The values provided here are common starting points and should be measured against the reader's own samples.