Data profile for this category
The core entities in this category are manufacturers of power battery packs, energy storage cabinets, and BMS systems. Daily generated documents include assembly rules, BOM lists, welding and torque records, BMS flashing records, insulation and voltage withstand reports, EOL test reports, and system configuration lists. Data sources include cell capacity grading data, MES systems, welding equipment, torque tools, BMS software version repositories, and EOL test benches. Documents are updated either per produced pack or cabinet, or synchronously with BMS software version changes. Documents contain professional fields such as cell gear, voltage tolerance, torque value, and communication protocol version. Some log documents include timestamps and device identification numbers.
Constraints for document parsing and chunking
Documents in this category are mostly structured or semi-structured files exported by equipment or systems. Each single document maps to one independent production unit. Retain the complete semantics of a single record when chunking, and avoid splitting core parameters. Keep units intact for professional fields such as torque (N·m) and voltage (V) to prevent field breaks during chunking. Avoid mixing test steps and results across chunks for long-sequence EOL test reports. Preserve parent node associations for BOM hierarchical structures to enable correct component information association during retrieval. Bind fault codes and timestamps in exception reports, and do not split them across different chunks.
Configuration Settings
These recommendations are based on FastGPT v4.17.0.
| Configuration Item | Recommended Setting | Rationale |
|---|---|---|
chunkSize | 800–1200 characters | Adapts to the complete semantics of a single process step or test record, avoids splitting core assembly rule parameters |
chunkSplitter | Split by paragraph + fixed characters | Adapts to the structured line break format of welding records and torque logs, preserves complete fields for single records |
customPdfParse | Enable text layer-first parsing | Works with formatted PDF reports exported from MES, avoids field misalignment from OCR recognition |
PARSE_FILE_TIMEOUT_SECONDS | 600 seconds | Supports long text parsing for single EOL test reports, prevents timeout interruptions |
UPLOAD_FILE_MAX_SIZE | 1000 MB | Meets upload requirements for large single EOL test reports, prevents system file rejection |
chunkSettingMode | Manual configuration | Adjust chunk boundaries manually for BOM hierarchical structures, preserves parent node associations |
Three Common Configuration Errors
- Tool HTTP request parameters are parsed normally, but no execution occurs Phenomenon: Tool call logs show successful parameter parsing, but no corresponding execution result is generated. Cause: Tool call permission is not enabled in the knowledge base configuration, or the parameter format does not meet the preset trigger rules.
- Imported Feishu spreadsheet documents cannot be parsed for indexing Phenomenon: Uploaded Excel or multi-dimensional documents show successful parsing, but no matching content appears during retrieval. Cause: The non-PDF format adaptation switch for
customPdfParseis not enabled, or unstandardized merged cells in the document cause field recognition failures. - Imported PDF knowledge base files cannot be opened in search results Phenomenon: Clicking the PDF link in search results produces no response or a "file does not exist" prompt. Cause: The file hosting path is not configured, or the
UPLOAD_FILE_MAX_SIZEsetting is smaller than the uploaded file size, resulting in file truncation.
How to Verify Correct Configuration
- Upload a standard BOM Excel file, confirm parsed chunks retain hierarchical associations and have no field splitting misalignment.
- Submit a single EOL test report, check the parsing log for no timeout errors, and confirm the number of chunks matches the number of document paragraphs.
- Enable tool call testing, pass preset process parameters, confirm the tool triggers normally and returns matching results.
- Check the file upload limit, upload a file near the
UPLOAD_FILE_MAX_SIZEthreshold, confirm parsing is not interrupted.
The values provided are common starting points and should be measured against the reader's own samples.