Data Characteristics for This Category
Pharmacovigilance data for high-value consumables primarily originates from healthcare institution reports, manufacturer-collected data, and the National Center for Adverse Drug Reaction Monitoring. Data updates frequently, especially for post-market surveillance, typically summarized and released weekly or monthly. Data often appears in structured table formats like CSV, Excel, or HL7 messages. Fields include product batch number, manufacturing date, expiration date, implantation/usage date, adverse event description, basic patient information, medical institution code, and reporter information. Adverse event descriptions may contain extensive unstructured text. Units for physical parameters such as size, weight, and volume use standard units like millimeters (mm), grams (g), and milliliters (ml). Event timestamps use date-time formats.
Constraints Imposed by These Characteristics on "HTTP API and External Systems"
Frequent updates to high-value consumable data require HTTP APIs to support efficient data synchronization for near real-time data pulling or pushing. The coexistence of structured data and unstructured text means APIs must map structured fields and effectively parse and index unstructured text, such as extracting keywords from adverse event descriptions. Diverse data sources lead to format variations, requiring APIs to offer flexible data conversion and cleansing capabilities to standardize input. Precise matching of critical fields like batch numbers and manufacturing dates demands high data transmission integrity and accuracy, requiring strict validation of API request parameters. Additionally, the specialized nature of high-value consumables may result in adverse event descriptions containing numerous medical terms, challenging the semantic understanding capabilities of the knowledge base.
Configuration Guidelines
| Configuration Item | Recommended Value | Rationale for This Value |
|---|---|---|
maxContext | 4000 | Ensures accommodation of a complete adverse event description and related product information. |
Chunk size (Segment Length) | 500 characters (characters) | Balances text block completeness and retrieval efficiency, preventing excessively long text from affecting similarity calculations. |
Recall count (Recall Count) | Top 10 entries (top 10) | Given the complexity of high-value consumable adverse events, increasing the recall count improves relevance coverage. |
Similarity threshold (Similarity Threshold) | 0.75 | Balances precise matching and semantic generalization, filtering out irrelevant or weakly relevant results. |
PARSE_FILE_TIMEOUT_SECONDS | 600 seconds (seconds) | Provides sufficient time to process report files containing large amounts of unstructured text. |
UPLOAD_FILE_MAX_SIZE | 100 MB | Accounts for uploading batch data files that may contain multiple adverse events. |
Three Common Mistakes
- Slow indexing speed after calling the API to create a file collection. This occurs when uploaded raw files are not pre-processed, containing redundant information or inconsistent formats, leading to inefficient tokenization and vectorization.
- Knowledge base assistant returns null values when calling a workflow via API. This can happen if a nested knowledge base assistant in the workflow is not correctly configured with the associated knowledge base ID, or if knowledge base versions are incompatible, for example, a knowledge base created in versions below v4.8.10 being called in a higher version.
- Inability to correctly identify question and answer columns after uploading a CSV template file via API. This is due to incorrect CSV file encoding or column order not matching expectations, preventing the system from parsing according to the convention of
first column as question, second column as answer.
How to Verify Configuration
- Upload a typical high-value consumable adverse event report file (e.g., CSV format) via API. Check the knowledge base index status to confirm all documents are successfully processed and indexed.
- Use the API to call the knowledge base retrieval function. Input a query containing high-value consumable product models and adverse reaction symptoms. Verify the relevance and accuracy of the returned results against expected answers.
- Simulate an external system pushing a new adverse event data record via HTTP API. Observe if the system correctly receives, parses, and stores the data. Verify data retrievability through a query interface.
- Check system logs to confirm no HTTP 5xx error codes or
Connection Timeoutexceptions occurred during data synchronization and knowledge base updates.
The values provided are common starting points and should be measured against specific 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.