This page is written based on the configuration logic of FastGPT v4.17.0.
What the data for this category looks like
The data for this category originates from ERP orders and inventory, bills of materials (BOMs) and process routes, equipment capacity, staff shifts, work-in-progress items, and supplier arrivals. Daily documents include sales orders, master production schedules, material requirement plans, weekly and daily scheduling sheets, work orders, material shortage lists, and delivery change notices. Data updates occur daily, weekly, or when triggered by order or exception events. Document structures fall into two categories: tables and text. Fields include order numbers, material codes, capacity man-hours, work-in-progress quantities, delivery dates, and others. Units include pieces, hours, days, and others.
What constraints do these characteristics impose on the source citation and traceability link
Data sources are scattered and update at inconsistent intervals. Match the update time of corresponding data during traceability to avoid citing expired scheduling or material data. Document structures vary. Adapt indexing and retrieval logic for different formats to ensure accurate identification of both tabular scheduling sheets and textual exception reports. Fields and units are diverse. Fully retain original information during citation to avoid impacting subsequent scheduling calculations and exception judgments. Temporarily generated exception reports must be linked to original trigger data to maintain a complete traceability chain.
How to set the configurations
| Configuration Item | Recommended Setting | Rationale |
|---|---|---|
quoteMaxToken | 8000–12000 token | Production planning data includes multiple scheduling sheets and work orders. Sufficient token is required to carry complete cited content and avoid context truncation |
| Single search limit | Top 10–15 results | Data sources for this category are scattered. Too many results will cause context redundancy, while too few will miss key scheduling constraint data |
responseData | Retain original document fields and units | Production planning data includes unit-bearing fields such as material codes and capacity man-hours. Full retention ensures accuracy of subsequent scheduling calculations |
| Citation source display | Enabled | Workshop scheduling and planning personnel need to trace the original order or work order records corresponding to exception reports to ensure the traceability chain is accessible |
detail | Retrieve by document type | Documents in this category have diverse structures. Retrieving by type improves the matching degree between cited content and drafting requirements |
| Response timeout threshold | 300–600 seconds | Retrieval of production planning data requires cross-system access. Sufficient time is reserved to avoid citation failure caused by timeout |
Three common configuration errors
- The symptom is that after setting
quoteMaxTokento 3000, the large language model does not receive context content. The reason is that the data volume of a single scheduling sheet or work order exceeds the configured threshold, causing context truncation and failure to transmit. - The symptom is that configuring HTTP response data as a citation has no effect. The reason is that data is not converted to a structured format that meets FastGPT knowledge base requirements, and document titles and source identifiers are not included.
- The symptom is that knowledge base citations are displayed but the answer does not include professional content. The reason is that the
detailretrieval parameter is configured too low, failing to retrieve sufficient scheduling constraint data, or insufficient context token causes professional content to be compressed.
How to verify correct configuration
- Upload a typical weekly and daily scheduling sheet and material shortage list, test the citation and traceability function, and check whether the returned results include document titles and original field information.
- Adjust the
quoteMaxTokenparameter, upload multiple production planning documents to verify that the context is fully transmitted without truncation prompts. - Call the API to pass the workflow, confirm that the knowledge base ID is correctly associated, and check whether the returned results carry corresponding source identifiers.
- Test the exception report drafting scenario, confirm that the cited exception data can be traced back to the original delivery change notice or material shortage list.
The values provided here are common starting points and should be measured against the reader's own samples.