What the data for this category looks like
Data sources for this category include the Chinese Center for Disease Control and Prevention Information System, hospital HIS systems, infectious disease reporting systems, and laboratory testing systems. Daily processed documents include infectious disease reporting cards, epidemic analysis reports, case correction records, clustered epidemic reports, and epidemiological investigation reports. Data is updated daily. Each single record contains fields such as basic patient information, diagnosis results, reporting time limit, and correction status. All fields must comply with statutory notifiable infectious disease reporting specifications, with no custom units.
Constraints Imposed on Multi-turn Dialogue and Prompt Engineering
The daily data update requirement means multi-turn dialogue must limit the time range of the current session to avoid mixing data across cycles. Differences in fields across multiple types of structured documents require prompts to clearly distinguish extraction rules for different documents such as infectious disease reporting cards and epidemiological investigation reports. Data from multiple system sources may have inconsistent fields. Multi-turn dialogue must gradually guide users to confirm field mapping relationships. For time-consuming steps such as verifying patient diagnoses and correcting duplicate or incorrect cards, multi-turn dialogue must support task breakdown into steps, gradually obtaining information about records to be verified.
Configuration Settings
Reference only; verify against the target FastGPT release and deployment documentation.
| Configuration Item | Recommended Setting | Rationale |
|---|---|---|
| Maximum Context | First 2000 tokens | Adapts to the field volume of a single epidemiological investigation report or infectious disease reporting card, avoiding exceeding model context limits |
messages | Retain the first 10 turns of conversation history | Matches the interactive logic of multi-step verification in disease control scenarios, reducing redundant context usage |
variables | Preset statutory reporting field variables | Quickly calls standardized fields such as basic patient information and reporting time limits, improving prompt reusability |
chatId | Bind to the current day's session ID | Distinguishes ledger data across different dates, avoiding mixing data across days |
| Single Interaction Response Limit | First 3 records | Corresponds to the daily processing scale for summary work, avoiding excessive data returned in a single request that causes timeouts |
| Session Timeout | 300 seconds | Adapts to the thinking time required for complex verification tasks, avoiding forced interruption of unfinished interactions |
Three Common Misconfigurations
- Uploading a ledger file triggers an error, while text dialogue functions normally. The cause is that a parsing template for the corresponding file format is not configured, so structured fields of infectious disease reporting cards cannot be recognized.
- The Unexpected end of JSON input error occurs during dialogue, and is easily triggered when using the chatglm2 model. The cause is that the model's returned content is not fully generated, leading to broken JSON formatting, with no model output truncation and retry logic configured.
- Multi-turn dialogue response speed is slow. The cause is that the number of recalled records per single request is not limited, causing the system to load excessive historical ledger data and occupy too many computing resources.
How to Verify Correct Configuration
- Upload a single infectious disease reporting card file, check whether preset statutory reporting fields can be correctly extracted, and confirm no parsing errors occur.
- Initiate a multi-turn verification request, gradually input information about cases to be corrected, check whether the session retains context from previous interactions and correctly associates previously entered content.
- Wait for the session to complete, check whether results are returned within the preset timeout period, with no formatting break errors.
- Adjust the record count limit for a single interaction, verify that the number of returned results matches the preset value range.
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.