Reference only; verify against the target FastGPT release and deployment documentation.
What this category of data looks like
Data for this category comes from online monitoring platforms, automatic pollution source monitoring data, ecological environment law enforcement systems, and third-party test reports, and is updated daily. Daily documents include on-site inspection records, sampling records, monitoring reports, order for correction decisions, and administrative penalty decisions. Each document’s fields include law enforcement officer information, sampling points, pollutant detection values, law enforcement time, rectification requirements, and more. The length of individual documents varies widely, and the field structures of different documents differ significantly.
What constraints do these characteristics impose on the citation sources and traceability link
The daily update feature of the data requires that the traceability link must associate the latest online monitoring and law enforcement system data to avoid referencing outdated law enforcement or monitoring records. Diverse document types and obvious field differences require precise matching of the citation logic for corresponding documents to ensure that annotated source information corresponds to document content. The wide range of individual document lengths requires that citations of long documents avoid truncating key law enforcement evidence to prevent missing traceability information. Multi-source data must clearly mark the original system or report subject to ensure the authority of traceability and adapt to compliance requirements in party building learning scenarios.
How to set the configuration
| Configuration Item | Recommended Value | Rationale for This Setting |
|---|---|---|
quoteMaxToken | 8000–12000 token | Matches the length range of individual law enforcement or monitoring documents to avoid truncating key evidence content |
| Single recall result limit | Top 6 entries | Covers multi-source law enforcement and monitoring data while controlling the total context length |
detail | Full source fields | Ensures that key information such as document type, generation time, and source system is included in the output |
| Context recall range | Data from the past 7 days | Matches the daily update frequency of the data to ensure references to the latest law enforcement and monitoring records |
responseData | Standard citation markup | Avoids garbled output and ensures compliance with citation format requirements |
Three common configuration errors
- Symptom: Garbled citation marks appear at the end of each paragraph in dialogue output, and then automatically revert to quotation marks afterward. Cause: The standard citation format corresponding to
responseDatais not configured correctly, and the system’s default markup fails to render properly. - Symptom: Knowledge base query takes too long, and a single retrieval consumes a large number of tokens. Cause: The value of
quoteMaxTokenis not restricted, and the full knowledge base data is called by default, exceeding the actually required length. - Symptom: The knowledge base module in the workflow cannot return citation content. Cause: The
detailconfiguration item is not enabled, and the return of complete source information is not specified, causing the module to only return summary content.
How to confirm the configuration is correct
- Initiate a law enforcement or monitoring question related to party building learning, and check whether the output includes source information such as document type and generation time.
- View the conversation log to confirm that the number of tokens consumed by a single retrieval does not exceed the preset range of
quoteMaxToken. - Test the knowledge base call module in the workflow to confirm that the returned results include complete citation markup and source fields.
- Simulate a high-traffic retrieval scenario to check that the system response speed does not experience obvious lag.
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.