What the Data for This Category Looks Like
Typical entities in this category are manufacturers of power battery packs, energy storage cabinets, and BMS systems. Daily generated documents include grouping rules, BOM lists, welding and torque records, BMS flashing records, insulation and voltage withstand reports, EOL test data, and system configuration lists. Data sources cover cell capacity grading data, MES systems, welding equipment, torque tools, BMS software version files, and data collected from EOL test benches. Updates are synchronized each time a power battery pack or energy storage cabinet is produced, or each time a BMS software version is changed.
Constraints on Knowledge Base Retrieval and Recall
Data sources are scattered and linked to specific production batches. Retrieval processes use collectionIds to limit the corresponding knowledge base collection, preventing interference from cross-batch documents. Update frequency is high. Configure a near-real-time index update mechanism to ensure retrieved documents are the latest versions. Documents include structured parameter records and unstructured test reports. A single retrieval mode cannot cover all scenarios. Combine multi-mode retrieval to adapt to content in different formats. Errors directly affect system safety. Retrieved results must have high confidence. Use a re-ranking mechanism to filter low-relevance content.
Configuration Settings
These configurations are based on the knowledge base retrieval function of FastGPT v4.17.0, adapted to the document characteristics of this category:
| Configuration Item | Recommended Setting | Rationale |
|---|---|---|
similarity | 0.75–0.85 | Meets the high-precision parameter matching requirements of battery PACK documents, and avoids recalling low-relevance grouping rules or test records |
limit | Top 8 entries | Covers the number of core documents for a single batch of PACKs, and avoids excessive context length that affects inference efficiency |
searchMode | mixedRecall | Supports both keyword matching and semantic retrieval, adapting to different retrieval needs for structured parameters and unstructured reports |
usingReRank | Enabled | Performs secondary sorting on retrieved results, improves the display priority of high-confidence documents, and reduces the probability of missed detection of documents related to insulation and thermal management |
datasetSearchUsingExtensionQuery | Enabled | Expands relevant parameter queries based on the original question, supplementing hidden associated content such as cell capacity grading data and torque values |
collectionIds | Collection IDs bound to production batches | Limits the retrieval scope to only documents from the current batch, avoiding interference from cross-batch data |
Three Common Misconfigurations
- Phenomenon: Retrieval results only return documents from some knowledge bases, or cross-batch irrelevant documents are recalled. Cause: The
collectionIdsparameter is not configured correctly, and the retrieval scope of the knowledge base collection is not limited. - Phenomenon: The context reference area does not correctly render the original document format, or only displays meaningless segmented fragments. Cause: The retrieval and analysis configuration adapted to the document structure is not enabled, or the segmentation logic does not match the structured fields of the document.
- Phenomenon: Key parameters such as insulation withstand voltage values and BMS flashing versions are not recalled. Cause:
datasetSearchUsingExtensionQueryis not enabled, and associated parameter queries are not expanded, resulting in semantic retrieval failing to cover hidden associated content.
How to Verify Proper Configuration
- Retrieve the PACK grouping rule document for a specified production batch, check whether the returned results only include documents from the target batch, to verify that the
collectionIdsconfiguration is effective. - Enter a query containing insulation withstand voltage parameters, check whether the corresponding insulation withstand voltage report is included in the retrieved results, to verify the effectiveness of the
similarityandusingReRankconfigurations. - View the context reference area, confirm whether the format of unstructured documents matches the original document structure, to verify the correctness of the retrieval and analysis configuration.
- Compare the result differences between keyword retrieval and semantic retrieval, confirm whether the
searchModeconfiguration is adapted to the current document type.
The values given are common starting points and should be measured against the reader's own samples.