Reference only; verify against the target FastGPT release and deployment documentation.
What the data for this category looks like
Power generation group ledger data mainly comes from production operation SCADA systems, ERP fuel management systems, equipment maintenance management platforms, and environmental monitoring systems. Core operating parameters are synced once every 8 hours per shift. A monthly summary ledger draft is generated daily. Most documents are structured Excel/CSV files with fixed fields including equipment ID, operating hours, power generation (unit: megawatt-hours), coal consumption (grams of standard coal per kilowatt-hour), maintenance date, responsible team, and others. They also include unstructured maintenance report PDFs and monthly analysis Word documents.
What constraints do these characteristics impose on the knowledge base retrieval and recall link?
There are many structured fields with professional units in power generation ledgers. Single vector retrieval cannot accurately match queries with precise conditions such as "May 2024 #3 Unit coal consumption". Ledger data from multiple sources is scattered across different business systems. Knowledge base collections must be divided by business type to avoid interference from cross-category data. Frequently updated ledgers require the retrieval scope to cover newly uploaded files in real time. Duplicate uploaded ledgers for the same equipment and period must also be handled. Both unstructured maintenance reports and structured ledgers must be retrieved to generate complete analysis reports.
How to configure the settings
| Config Item | Recommended Approach | Rationale |
|---|---|---|
collectionIds | List of collection IDs grouped by business, such as power generation equipment ledgers, fuel procurement ledgers | Precisely limit the retrieval scope to avoid irrelevant business data mixing into summary results |
similarity | 0.75–0.85 | Adapt to the matching accuracy of power generation professional terminology, filter low-relevance non-professional expressions |
limit | Top 10–20 entries | Balance the information completeness of summary reports and reading load |
searchMode | vector+keyword | Balance semantic retrieval and exact matching of structured fields, suitable for multi-type data in power generation ledgers |
usingReRank | Enabled | Perform secondary sorting on initial retrieval results by relevance to correct local deviations from single vector retrieval |
datasetSearchUsingExtensionQuery | Enabled | Supplement synonymous expansion of power generation professional terminology, such as expanding "coal consumption" to "standard coal consumption rate" |
Three common misconfigurations
- Phenomenon: Empty results or irrelevant data are returned when retrieving power generation ledger data from a specified collection. Cause: The
collectionIdsparameter is not configured correctly, or the collection ID does not match the actually uploaded knowledge base collection, resulting in an unexpected retrieval scope. - Phenomenon: Entries in Excel files that meet the filter criteria are not fully retrieved. Cause: The keyword mode of
searchModeis not enabled. Relying only on vector retrieval cannot match the exact values of structured fields, resulting in some eligible entries being filtered out. - Phenomenon: Duplicate ledger entries for the same period and equipment appear in the knowledge base. Cause: Deduplication is not performed on uploaded Excel files by equipment ID + timestamp, or old version ledger files for the same period are not overwritten during upload.
How to confirm the configuration is properly set
- Manually specify
collectionIdsas the collection ID of a single power generation equipment ledger, retrieve a known existing equipment ID, and verify that the returned results only include ledger data from that collection. - Adjust the
similarityparameter to 0.8, retrieve the query term "#3 Unit coal consumption", and check whether the relevance of the returned results meets the expectations of professional scenarios. - Upload a test power generation ledger Excel file, retrieve specific field values from it, and confirm that the number of retrieved entries matches the source file.
- After enabling
usingReRank, compare the result ranking before and after enabling to confirm that the reranking logic takes effect.
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.