What data for this category looks like
Data for this category comes from the Chinese Center for Disease Control and Prevention Information System, medical institution HIS and infectious disease reporting systems, and laboratory testing systems. The update frequency is daily. Daily documents include infectious disease reporting cards, epidemic analysis reports, case correction records, clustered epidemic reports, and epidemiological investigation reports. Core fields cover basic case information, clinical diagnosis results, reporting time limits, correction notes, clustered epidemic association information, etc. Data must strictly comply with legal infectious disease reporting rules.
What constraints these characteristics impose on the model access and configuration link
This category of data is updated daily and has dispersed sources. Model access configuration must adapt to the daily knowledge base refresh mechanism to avoid using expired data. Daily documents include structured reporting cards and semi-structured epidemiological investigation reports. The configuration link must support accurate parsing of multiple document types and field extraction. Data must strictly follow legal infectious disease reporting rules. Model output must retain the integrity and accuracy of original data fields, and must not arbitrarily alter or omit key information. The requirement for multi-source data docking also means the configuration link must support unified access and verification of multi-system interfaces to ensure consistency of data synchronization.
How to set the configuration
Reference only; verify against the target FastGPT release and deployment documentation.
| Configuration Item | Recommended Value | Rationale |
|---|---|---|
maxContext | 8000–12000 token | Adapts to the total length of multiple daily epidemic reports and epidemiological investigation reports, avoiding truncation of key case and epidemic information |
maxTokens | 4000 token | Matches the output length requirement of a single infectious disease reporting card or epidemic analysis report, ensuring complete presentation of report content |
quoteMaxToken | 2000 token | Limits the single segment length of quoted documents, avoiding context overflow caused by overly long quotes |
temperature | 0.2–0.4 | Reduces random generation probability, ensuring the output of party building learning Q&A and infectious disease monitoring content is rigorous and accurate, and complies with legal reporting rules |
AIPROXY_API_ENDPOINT | Fill in the internal network interface address according to the deployment environment | Adapts to the deployment needs of disease control institutions without public network access, ensuring data interaction is completed within the internal network |
AIPROXY_API_TOKEN | Set based on actual testing | Ensures the security of interface access permissions, only allowing internal system calls |
Three common configuration mistakes
- The symptom is a
[FATAL] failed to get-type error when starting the oneapi container, with logs pointing to lines related tobuild/model/main.go. The cause is that the deployment environment has no network access, and no internal proxy or offline deployment solution is configured, resulting in failure to pull models or interface resources. - The symptom is that after connecting external conversation channels, the output of party building learning Q&A and disease control content contains a large number of meaningless punctuation marks such as # and *, and the interface display format is abnormal. The cause is that no cleaning rules are configured for residual marks after document parsing, causing format errors to be carried into the output content.
- The symptom is that after the rerank model is deployed, the matching degree between the recall results and case reports and party building learning content does not meet expectations. The cause is that no matching weights are configured for the rerank model for the core fields of this category, such as case diagnosis and reporting time limit, resulting in deviation of recall priority.
How to confirm the configuration is complete
- Perform a parsing test on a single infectious disease reporting card, and check whether the output fields cover core content such as basic case information, diagnosis results, and reporting time limits.
- Trigger a daily knowledge base refresh task, and confirm that data synchronization and model update can be completed normally in the deployment environment.
- Launch a Q&A test related to party building learning, and check whether the output content complies with specifications and does not carry redundant punctuation marks.
- Call the model interface to view the return logs, and confirm that the
AIPROXY_API_TOKENverification is passed and the interface response meets expectations.
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.