What the data for this category looks like
Coal mine production document and material data originates from three main scenarios. First, daily team production records. These include shift details, on-site headcount, equipment operating hours, and hidden danger inspection items. Second, monthly production reports and equipment operation and maintenance ledgers. These contain fields such as equipment ID, maintenance cycle, and consumable usage volume. Third, standardized documents including safety production regulations and emergency response plans.
Data update rhythms vary significantly. Team records are updated daily. Operation and maintenance ledgers are updated at each maintenance milestone. Standardized documents are revised once per year.
Individual document lengths range from hundreds of characters in team logs to dozens of pages of regulation files. Fields include timestamps, unique device codes, hidden danger levels, and more. Units used include hours, units, and items.
What constraints these characteristics impose on citation source and traceability
Multi-source, heterogeneous data structures require binding unique identifiers during traceability. This prevents confusion between identically named fields across different documents. Differences in update frequency mean retrieval ranges must separate real-time production data from historical standard documents. This stops incorrect cross-cycle data associations. The mix of short and long documents requires retaining field association information during chunking. This avoids matching irrelevant paragraphs during traceability. Fields contain professional attributes such as equipment numbers and hidden danger levels. This requires enabling precise matching during retrieval. This prevents generic searches from returning irrelevant content.
How to Set Up the Configuration
Reference only; verify against the target FastGPT release and deployment documentation.
| Configuration Item | Recommended Setting | Rationale |
|---|---|---|
| Chunk Length | 800–1200 characters | Adapts to the average paragraph length of coal mine team logs and regulation documents, avoids traceability deviations caused by overly fragmented or overly long chunks |
| Similarity Threshold | 0.75–0.85 | Matches the semantic precision of coal mine professional terminology, prevents generic searches from returning irrelevant operation and maintenance records |
| Single Search Limit | Top 3 entries | Controls the number of citation sources, avoids redundant output content, and meets the rigor requirements of coal mine documents |
quoteMaxToken | 200–300 tokens | Reserves sufficient space to display citation identifiers, prevents truncated garbled text |
responseData | Enabled | Carries unique identifiers and field information of document sources, supports precise traceability |
detail | Enabled | Displays core fields of cited documents, facilitates verification of production data accuracy |
Three Common Misconfigurations
- When querying coal mine scenario questions not present in the knowledge base, irrelevant document citations are still returned. The cause is that the "call large model to generate when no retrieval results" switch is not turned off, forcing the large model to associate existing knowledge base content.
- Unclosed citation marker garbled text appears at the end of output content. The cause is that
quoteMaxTokenis set too small, causing citation identifiers to be truncated without full closure. - After calling the knowledge base tool in a workflow, citation source information cannot be returned. The cause is that the
responseDataparameter is not enabled, so the tool return result does not carry document identification data required for traceability.
How to Confirm Proper Configuration
- Initiate a query containing a known equipment number, verify that the citation identifier at the end of the output matches the unique number of the corresponding document.
- Input a coal mine emergency scenario question not present in the knowledge base, confirm that no citation content is returned.
- Call the knowledge base tool in a workflow, check whether the return result includes detailed information of the document source.
- Initiate a production report query containing long text, confirm that the citation identifier does not appear truncated or garbled.
The values provided here are common starting points, and should be evaluated against the reader's own samples.
Question material comes from public community discussions. Configuration values are common starting points and should be measured against your own samples. Verified on 2026-09-28.