What the data for this category looks like
Typical participants in this category are project procurement staff, integrator business staff, and warehouse administrators. Core daily documents include purchase orders, arrival acceptance forms, serial number lists, and invoice materials. Data mainly comes from ERP systems, supplier shipping documents, logistics information platforms, and equipment manufacturer backends. Data update frequency is per order or weekly. Each document contains fields such as equipment model, serial number, arrival date, and supplier name. The most time-consuming business links are model configuration verification, arrival acceptance, and serial number registration. These links usually differ from general enterprise IT procurement data.
What constraints these characteristics impose on knowledge base retrieval and recall
Data sources for this category are scattered and have inconsistent formats. Documents exported from different systems may have inconsistent field names. This requires the retrieval system to support multi-source format adaptation and field mapping. Data updates are performed per order or weekly. The knowledge base must match this incremental sync frequency to avoid data lag affecting work order responses. Each document links to specific project equipment information. Retrieval must be limited to documents within the corresponding project to prevent invalid cross-project results from interfering with work. Business links have high requirements for equipment matching accuracy. Retrieval and recall must prioritize core fields such as model and serial number. Otherwise, issues such as equipment mismatches and project delays may occur.
How to set configurations
This page's configurations are based on FastGPT v4.17.0.
| Configuration Item | Recommended Value | Rationale |
|---|---|---|
similarity | 0.75–0.85 | Matches precise business fields such as equipment model and serial number. A threshold that is too low will introduce irrelevant documents. A threshold that is too high will miss valid matching results |
limit | Top 8 entries | Covers the number of conventional devices linked to a single purchase order. Prevents excessive recall from causing context overflow |
usingReRank | Enabled | Corrects ranking bias from initial vector retrieval. Prioritizes business key fields such as arrival date and serial number |
datasetSearchUsingExtensionQuery | Enabled | Supplements synonyms and abbreviations for equipment models. Adapts to different naming formats in supplier shipping documents |
collectionIds | Knowledge base collection IDs bound per project | Limits retrieval scope. Prevents invalid cross-project documents from interfering with business work order processing |
Three common errors
- Phenomenon: Retrieval result ranking does not align with business expectations. Core equipment serial number matching results are ranked low. Cause: Re-ranking function is not enabled. Initial vector retrieval only uses global semantic similarity. It does not adjust weights for business key fields.
- Phenomenon:
getaddrinfo EAI_AGAIN mongoerror occurs during knowledge base synchronization. Restarting the service does not resolve the issue. Cause: MongoDB connection timeout settings are not adjusted for concurrent requests from multi-source system imports. This leads to domain name resolution timeouts. - Phenomenon: Serial number text in images cannot be extracted from imported PDF equipment documents. Cause: Document OCR recognition configuration is not enabled. Only plain PDF text is extracted. This does not cover serial number lists in image format.
How to confirm configurations are correct
- Import a single purchase order document. Run retrieval for the corresponding work order question. Check if the recalled results include the core fields of that order.
- Adjust the
similaritythreshold. Verify if the number and relevance of recalled results meet business expectations. - After enabling the re-ranking function, compare retrieval ranking results before and after the change. Confirm if the ranking of key business documents has improved.
- Import a PDF document containing images. Verify that serial number information in images is correctly extracted.
The values listed here are common starting points. They should be validated against the reader's own datasets and use cases.
Question material comes from public community discussions. Configuration values are common starting points and should be measured against your own samples. Verified on 2026-10-19.