What This Category of Data Looks Like
Ledger data for refinery and chemical production comes from distributed control systems (DCS), tank farm level monitoring systems, laboratory quality inspection reports, production scheduling logs, and other sources. There are three update frequencies: real-time process parameters updated every second, daily shift ledgers updated daily, and monthly energy consumption reports updated monthly. Documents are structured around unit number, operation time, and material type, and include process parameters such as flow rate, temperature, and pressure, as well as fields like quality inspection pass rate and energy consumption indicators. Units follow refinery and chemical industry standards: flow rate is measured in cubic meters per hour, temperature in degrees Celsius, and pressure in megapascals.
Constraints Imposed on Knowledge Base Retrieval and Recall
The characteristics of this data impose the following constraints on the knowledge base retrieval and recall link:
- Second-level updates of real-time process parameters require the knowledge base to support incremental synchronization to avoid retrieving expired data.
- Integration of multi-source data requires limiting the retrieval scope to the dedicated refinery and chemical ledger knowledge base to prevent mixing in other business data.
- Complex professional fields and units require associating professional terms with fixed fields during retrieval matching, rather than relying solely on semantic matching.
- Batch records of long documents require retaining contextual information such as unit and batch during chunking, to avoid losing key associations after splitting.
How to Configure the Settings
Reference only; verify against the target FastGPT release and deployment documentation.
| Configuration Item | Recommended Value Range | Rationale |
|---|---|---|
similarity | 0.75–0.85 | Refinery and chemical ledgers contain a large number of professional terms and fixed formats. A threshold that is too low will introduce irrelevant unit data, while a threshold that is too high will miss associated ledgers of the same batch |
limit | Top 10–15 entries | A single refinery and chemical ledger contains multiple sets of process parameters. Retrieving too many entries will exceed the context window, while retrieving too few will fail to cover the complete process chain |
searchMode | vector + fulltext | Balances semantic matching and keyword matching, adapting to the structure of ledgers that include both process descriptions and fixed fields |
usingReRank | Enabled | Refinery and chemical data has strong professional relevance. Reranking can filter out results that are semantically similar but do not match the process scenario |
datasetSearchUsingExtensionQuery | Enabled | Supplements retrieval terms for professional fields such as unit number and batch number, improving accurate matching rate |
collectionIds | Fill in according to the bound refinery and chemical production ledger knowledge base ID | Limits the retrieval scope to avoid mixing in knowledge base data from other categories |
Three Common Misconfigurations
- Phenomenon: Non-refinery and chemical production ledger data appears in retrieval results. Cause: The
collectionIdsparameter was not filled correctly, and the retrieval scope was not limited to the target refinery and chemical knowledge base. - Phenomenon: Local knowledge base retrieval requests trigger external network access restrictions, and results cannot be returned normally. Cause: External network access permissions were not configured in the FastGPT backend, and external calls to local knowledge bases are restricted by default.
- Phenomenon: Retrieval results do not include process parameters for the specified batch. Cause: The
similaritythreshold was set too high, resulting in failure to match ledger data that is semantically similar but not exactly identical in fields.
How to Confirm the Configuration Is Successfully Applied
- View the knowledge base list in the FastGPT backend, confirm that the knowledge base ID corresponding to the
collectionIdsparameter matches the target refinery and chemical ledger knowledge base. - Initiate a retrieval for a known unit number, check whether the returned results include the ledger data for that unit.
- Adjust the
similaritythreshold, observe changes in the number of retrieval results, and confirm that the threshold meets business requirements. - After enabling
usingReRank, compare the result ranking before and after reranking, confirm that the reranking function is working.
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.