Data Characteristics
Stem cell therapy pharmacovigilance data originates from clinical trial reports, real-world evidence (RWE) studies, post-market surveillance, and various medical literature. This data typically exists as unstructured text, semi-structured tables, or structured database records. Update frequency varies; clinical trial data usually updates in batches when phase reports are released, while post-market surveillance data may flow in continuously, daily or weekly. Document structures are diverse, including study protocols, case report forms (CRFs), serious adverse event (SAE) reports, medical journal articles, and conference abstracts. Fields involved include patient demographics, treatment regimens, stem cell product batch information, adverse event descriptions, occurrence times, severity, outcomes, and causality assessments. Units cover dosage (e.g., cells/kg), time (e.g., days, weeks), and metrics (e.g., mmHg, ng/mL).
Constraints on Deployment and Upgrade
The complexity and diversity of stem cell therapy data sources require FastGPT to have robust multi-format parsing capabilities during data ingestion. Unstructured text (e.g., adverse event descriptions) needs efficient natural language processing (NLP) for information extraction, while structured or semi-structured data requires precise field mapping. Continuous data inflow, especially from post-market surveillance, demands real-time processing and incremental update mechanisms. The complex document structure impacts knowledge base segmentation strategies, requiring careful consideration of how to effectively chunk data at different granularities while maintaining contextual integrity. Specific dosage and time units, along with medical terminology, necessitate specialized domain understanding from the model to avoid discrepancies during information extraction and question answering. Accurate identification of critical fields like batch information is also crucial for traceability and risk assessment.
Configuration Settings
| Configuration Item | Recommended Value | Rationale |
|---|---|---|
UPLOAD_FILE_MAX_SIZE | 1000 MB | Clinical trial reports and medical literature often contain many images and charts, leading to large file sizes. Sufficient upload limits are necessary. |
PARSE_FILE_TIMEOUT_SECONDS | 600 seconds | Processing large PDF reports or multi-page CRF files can be time-consuming. Increasing the timeout prevents parsing failures. |
Chunk Length | 800–1200 characters | Adverse event descriptions in stem cell therapy are often detailed. A longer chunk length helps preserve complete contextual information. |
Recall Count | Top 10 | This ensures enough relevant adverse event reports or clinical details are recalled for complex queries, improving retrieval accuracy. |
Similarity Threshold | 0.75 | A higher similarity threshold helps achieve precise matches, especially given subtle differences in medical terminology and symptom descriptions. |
maxContext | 32000 | Complex medical case analysis and multi-dimensional adverse reaction correlations require the model to process longer contextual information. |
Common Pitfalls
- After uploading knowledge base documents, adverse event descriptions or treatment regimen fields are empty. This usually occurs because PDF or image format documents were not processed with Optical Character Recognition (OCR), or OCR errors led to incomplete text extraction.
- When processing daily new adverse event reports, data occasionally fails to update promptly or is duplicated. This typically results from improper API call frequency settings for the data source or flaws in the incremental synchronization logic.
- FastGPT calls a locally deployed MinerU for file parsing, but a file size limit error appears. This happens because the
UPLOAD_FILE_MAX_SIZEparameter was not configured correctly when starting the MinerU container, leading to an excessively small default limit.
Verification Steps
- Upload a clinical trial report PDF containing complex charts and extensive medical terminology. Check if the knowledge base fully extracts critical fields like adverse event descriptions and stem cell product batch information. Compare the extracted content with the original to confirm no omissions or errors.
- Simulate an influx of new adverse event report data. Observe the FastGPT knowledge base update status to confirm that new data is correctly indexed, does not duplicate or overwrite existing data, and that the update frequency meets expectations.
- Query for potential adverse reactions related to a specific stem cell therapy regimen. Check if the recall results include relevant information from different document types (e.g., clinical reports, RWE data). Evaluate the accuracy and comprehensiveness of the answers, and confirm if the recall count meets requirements.
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.