Reference only; verify against the target FastGPT release and deployment documentation.
Data Profile
Coal mine production declaration materials and file review data originate from multiple sources. These include daily shift reports from coal mine production scheduling systems, ledger archives from safety supervision agencies, gas and ventilation monitoring reports, tunneling and mining volume records, and internal enterprise license approval documents.
Update frequencies vary across data types: real-time shift reports and monitoring data, weekly aggregated safety supervision ledgers, and license documents updated quarterly or annually.
Documents contain both structured fill-in fields and unstructured supporting attachments. Fields have clear physical units such as meters, percent, and cubic meters per minute. Attachments are mostly scanned approval forms or monitoring screenshots.
Impact on Citation Sourcing and Traceability
The mixed structure of coal mine production data requires the traceability process to match both structured fields and unstructured attachment content. Avoid only searching text while overlooking quantitative indicators with units.
Data with different update frequencies requires differentiated synchronization trigger rules. Real-time data uses incremental synchronization. Periodic data uses scheduled full updates.
Fields with units require retaining unit matching logic during retrieval. This prevents confusion between tunneling footage and gas concentration.
Unstructured attachments must first undergo OCR recognition and be associated with the original file path. This ensures direct access to the original scanned documents during traceability.
Configuration Settings
| Configuration Item | Recommended Value | Rationale |
|---|---|---|
quoteMaxToken | 800–1200 token | The longest single coal mine declaration material is approximately 1000 characters, so this aligns with the token-to-character ratio |
| Single recall limit | Top 6 entries | Supporting materials for coal mine file review usually do not exceed 5 copies, so this retains a reasonable redundancy margin |
responseData | Return original document ID and storage path | Traceability requires linking to specific shift reports or safety supervision ledger entries, which aligns with public service sector audit requirements |
detail | Enable structured field extraction | Coal mine data includes unit-bearing fields such as gas concentration and footage, so field matching accuracy must be preserved |
| Traceability log storage period | 90 days | Complies with record retention specifications for state-owned enterprises and public service sectors |
Three Common Configuration Errors
- Symptom: The reference variable dropdown menu is empty when selecting a variable application for a knowledge base search node. Cause: The dedicated coal mine production ledger and report knowledge base was not configured as a bindable variable group in advance, or the variable group was not correctly associated with the corresponding data source.
- Symptom: The large language model outputs forged reference IDs that do not match the actually retrieved documents. Cause: The original document ID verification logic for
responseDatawas not enabled, or the knowledge base index was not correctly associated with the unique number field of the coal mine ledger. - Symptom: The text "Reference Marker: [1]" appears in the output content. Cause: FastGPT's native reference marker prefix output was not disabled, or custom reference display rules were not configured.
How to Verify Successful Configuration
- Submit an audit query that includes tunneling footage parameters, and check if the returned results are associated with the document ID of the corresponding shift report.
- Enter the FastGPT knowledge base management interface, and confirm that the dynamically bound variable group includes two dedicated data sources: coal mine production ledgers and safety supervision reports.
- Trigger a complete file review process, and check if the system log records the retrieved document path and reference ID.
- Check the large language model output content, and confirm that reference markers are only displayed as inline text in the body, without the "Reference Marker:" prefix.
The values provided are common starting points and should be measured 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.