Data Characteristics
Deviation and CAPA (Corrective and Preventive Action) system data originates from internal quality management systems, Manufacturing Execution Systems (MES), Laboratory Information Management Systems (LIMS), and audit reports. This data typically exists as structured or semi-structured documents, including deviation reports, investigation reports, CAPA plans, implementation records, and effectiveness verification reports. Documents contain fields like event descriptions, root cause analyses, action plans, responsible parties, completion dates, and status. The update frequency depends on deviation events and CAPA execution cycles, usually involving continuous updates. Key status fields (e.g., "Closed," "Delayed") may trigger immediate updates. Data units include time (e.g., "hours," "days"), quantity (e.g., "batches," "items"), and temperature (e.g., "℃"). Different fields may have different units.
Constraints on Deployment and Upgrade
The continuous update nature of deviation and CAPA data requires incremental synchronization mechanisms during knowledge base deployment. Files uploaded may need periodic or on-demand re-indexing to ensure information timeliness. Diverse document structures (e.g., varying report templates at different stages) challenge text segmentation and embedding model selection, requiring semantic integrity across different document types. The specificity of fields and units, especially in queries involving numerical comparisons or unit conversions, demands high recall accuracy and model comprehension from the knowledge base. Compliance requirements for this data also make data security and access permission configuration critical deployment considerations. During upgrades, ensure compatibility with the existing knowledge base and smooth integration of historical and new data formats.
Configuration Settings
| Configuration Item | Suggested Value | Rationale |
|---|---|---|
UPLOAD_FILE_MAX_SIZE | 100 MB | Deviation and CAPA reports may contain many images or attachments, requiring sufficient file upload capacity. |
Chunk size (Segment Length) | 500–800 characters (characters) | Balances logical integrity of reports and semantic correlation between paragraphs, avoiding excessive fragmentation. |
Recall count (Recall Count) | Top 8 entries (top 8) | Ensures critical information from multiple relevant reports is covered during queries, improving recall rate. |
Similarity threshold (Similarity Threshold) | 0.75 | Balances recall precision and breadth, filtering out irrelevant low-similarity content. |
PARSE_FILE_TIMEOUT_SECONDS | 300 seconds (seconds) | Accommodates parsing time for large or complex documents, preventing task failure due to parsing timeouts. |
Rerank result count (Reranked Return Count) | Top 5 entries (top 5) | Further refines recall results, improving the quality and efficiency of information ultimately provided to the user. |
Common Mistakes
- JavaScript code in a workflow fails with dependency or permission errors. This is due to Docker container environments potentially lacking required Node.js modules or the container user lacking permissions for specific operations.
- Knowledge base content uploaded via files does not update after source file modifications. File uploads do not automatically synchronize by default; manual re-indexing or external scripts are required.
- Query results miss critical deviation or CAPA report field information. This occurs when the knowledge base fails to correctly identify and extract all custom fields during import, or segmentation strategies truncate key fields.
Verification Steps
- Upload a deviation report with complex tables and multiple pages. Verify the knowledge base fully parses and displays all text content.
- Query an updated CAPA plan using keywords. Verify the knowledge base returns the latest version of the content.
- Perform queries including specific fields (e.g., "root cause," "responsible department"). Verify results accurately hit and display information for these fields.
- Simulate multiple concurrent queries. Observe system response speed and resource utilization to ensure stability under high concurrency.
The values given 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.