Data Characteristics
Deviation and CAPA (Corrective and Preventive Action) data in clinical trial pre-screening originate from abnormal event reports, quality audit records, instrument calibration reports, and internal review documents. Data updates are irregular, typically recorded after a deviation event or during CAPA implementation and verification. Document structures are often semi-structured, including free-text descriptions, structured fields (e.g., deviation type, impact, root cause, CAPA description, responsible person, completion date, verification results), and attachments (e.g., photos, test reports). Key fields include event number, deviation level, CAPA status, and impact assessment results. Date fields are usually precise to the hour, and some measurement results may include specific units, such as drug concentration (μg/mL) or device parameters (mV).
Constraints Imposed by These Characteristics on "HTTP Interface and External Systems"
The semi-structured nature of deviation and CAPA data requires the HTTP interface to have flexible data parsing capabilities to accommodate mixed input of free text and structured fields. Irregular update frequency means external systems need to support both event-driven or scheduled polling data synchronization mechanisms to ensure timely information. The presence of attachments demands file upload and download functionality from the interface, requiring handling of large file transfers and various file formats. Additionally, the precision of date fields requires data transfer protocols to support standard timestamp formats, avoiding timezone or format conversion errors. The existence of specific units requires the interface to retain original unit information during data transfer or provide clear unit conversion rules to prevent data misinterpretation.
Configuration Guidelines
| Configuration Item | Recommended Value | Rationale |
|---|---|---|
maxContext | 4000–8000 Token | Deviation and CAPA reports often contain detailed descriptions and analyses, requiring a larger context window to capture complete information. |
PARSE_FILE_TIMEOUT_SECONDS | 300 seconds | File parsing can be time-consuming when processing CAPA reports with attachments, requiring sufficient timeout duration. |
similarityThreshold | 0.75 | A high similarity threshold helps accurately match documents with similar deviation types or CAPA plans, reducing irrelevant results. |
recallCount | 10–15 items | Ensures that enough potentially relevant deviation and CAPA records are covered during the initial recall phase, providing sufficient candidates for subsequent ranking. |
reRankCount | 5 items | After re-ranking, filters out the most relevant deviation and CAPA records, improving the accuracy and usability of the final results. |
HTTP_REQUEST_TIMEOUT_MS | 60000 milliseconds | External system interfaces may respond slowly when handling complex queries or large file uploads; a longer timeout prevents request interruptions. |
Common Pitfalls
- The HTTP interface returns a 500 error with an "out of memory" log message. This may be due to an excessively large deviation report attachment exceeding server processing capacity.
- Date fields in retrieval results display as garbled characters or in an incorrect format. This occurs when the timestamp format returned by the external system does not match the format expected by FastGPT, and proper time format conversion is not performed.
- Frequent API calls lead to high server load and significantly reduced response speed. This may be due to not setting reasonable API call frequency limits or not batching data during bulk synchronization.
Verification Steps
- Upload CAPA reports containing different file types (e.g., PDF, Word documents, images) via the API and confirm that all attachments are correctly stored and parsed.
- Synchronize a batch of deviation records with detailed dates and specific units from an external system. Check that the corresponding fields in FastGPT display accurately, without format errors or unit loss.
- Simulate high-concurrency requests by continuously calling the knowledge base retrieval interface. Use system monitoring tools to observe server CPU, memory usage, and response time to confirm they are within acceptable limits.
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 2026-09-21.