What Data for This Category Looks Like
Core data for this category comes from daily documents of manufacturing enterprise PMC, planning, shop scheduling, and material control departments. These documents include sales orders, master production schedules, material requirement plans, weekly and daily production schedules, work orders, material shortage lists, and delivery change notices. Data sources cover ERP orders and inventory, BOMs and process routes, equipment capacity, staff shifts, work in progress, and supplier deliveries. Data update frequencies include daily, weekly, and real-time updates when orders or exceptions occur. Document structures include plain text and tables. Fields include order numbers, capacity values, delivery times, material codes, and more, adjusted dynamically based on business scenarios.
Constraints Imposed on Vector Models and Indexes
Data sources for this category are scattered, update frequencies vary, and the dataset includes a large volume of structured tables and unstructured text. These factors create multiple constraints for the vector model and index workflow. First, field differences across multi-source data require the vectorization process to cover core business fields such as orders, capacity, and delivery times, to avoid missing key constraint conditions. Second, real-time updated exception data such as material shortage lists and delivery change notices requires indexes to support incremental updates, to avoid performance losses caused by full index rebuilding. Third, the mixed document structure requires chunking rules adapted to different lengths of text and table content, to ensure the completeness and accuracy of vector embeddings. Fourth, cross-process resource constraint data requires retrieval results to match both order and equipment information, to avoid retrieval bias from a single dimension.
Configuration Settings
The following configurations are formulated based on FastGPT v4.17.0.
| Configuration Item | Recommended Values | Rationale |
|---|---|---|
| Chunk Length | 800–1200 characters | Adapts to the typical length of single-paragraph text and table row groups from work orders and production schedules |
| Overlap Length | 50–100 characters | Prevents business logic across chunks from being truncated, and ensures vector integrity for associated fields such as capacity and delivery times |
maxToken | 8000 | Accommodates contextual information from long texts such as master production schedules and complete work orders, to meet the needs of complex queries |
| Similarity Threshold | 0.75–0.85 | Filters low-match irrelevant data, and ensures relevance of retrieval results to equipment inspection and scheduling constraints |
| Single Search Limit | Top 10 results | Balances query response speed and retrieval coverage, to adapt to real-time decision-making needs for shop scheduling |
defaultToken | 100 | Controls the base length of single vector embeddings, to adapt to short-text documents such as material shortage lists and delivery change notices |
Three Common Misconfigurations
- Symptom: In the FastGPT knowledge base settings, enabling the Doubao-embedding-large index model, filling in a custom request address and apikey, then clicking test results in an immediate error. Cause: The official request path format for the model was not matched correctly, or the apikey field contains extra spaces and line breaks.
- Symptom: Table data in the knowledge base does not generate corresponding vectors, and search results cannot extract field information such as capacity and material quantity from tables. Cause: Multi-vector support for table data was not enabled, or chunking rules were not adapted to the row-column structure of tables, leading to overall truncation of table content.
- Symptom: The number of search results does not match the configured single search limit, with either too many or too few results. Cause: The retrieval quantity threshold was not set correctly in the vector index configuration, or the latest scheduling and work order data was not synchronized during index updates.
How to Verify Successful Configuration
- Upload a test work order and weekly/daily production schedule, and check if the vector generation task status in the FastGPT v4.17.0 backend shows as successful.
- Enter a test query such as "What are the equipment capacity constraints for the current work order?", and check if the returned results include equipment capacity field information from the ERP system.
- Modify a test delivery change notice to trigger incremental index updates, and confirm that search results match the latest delivery data.
- Navigate to the knowledge base settings page, and check if all configuration item parameters match the preset values, with no missing or altered values.
The values provided are common starting points and should be evaluated against the user's own samples.