Data Characteristics for this Category
Gene therapy AAV (adeno-associated virus) quality documents include production batch records, test reports, stability study data, raw and auxiliary material release forms, equipment calibration records, and deviation handling reports. Data sources are diverse, involving internal LIMS systems, ELN electronic lab notebooks, Quality Management Systems (QMS), and Certificates of Analysis (CoA) from suppliers. Document formats vary, including PDF, Word, Excel, and scanned images. The update frequency is relatively stable: batch records are generated with each production batch, test reports are updated upon completion of testing projects, and stability data is generated periodically at preset time points. Fields and units are highly specialized; for example, virus titer units are vg/mL, host cell DNA residue units are ng/mg, and endotoxin units are EU/mL. Document structures typically follow GMP guidelines, comprising fixed templates and variable data areas.
Constraints on Deployment and Upgrade
The specialized nature and diversity of AAV quality documents impose specific requirements on FastGPT deployment. Many documents are in PDF format, containing numerous tables and images, necessitating efficient OCR and table parsing capabilities. The periodic nature of data updates requires the knowledge base to support incremental updates, avoiding full re-imports. The presence of specific fields and units means that during vectorization and retrieval, accurate understanding and matching of these specialized terms are crucial for precise question answering. For instance, when querying "virus titer for batch number AAV2023001", the system must accurately identify the batch number and associate it with the corresponding titer value. The sensitive nature of the documents means the deployment environment must meet strict data security and access control requirements. Private deployment is a common choice to address these constraints, ensuring data remains within the local network and can integrate with existing QMS systems.
Configuration Guidelines
| Configuration Item | Recommended Value | Rationale |
|---|---|---|
UPLOAD_FILE_MAX_SIZE | 500 MB | AAV quality documents often include many charts and attachments, leading to potentially large individual file sizes. |
PARSE_FILE_TIMEOUT_SECONDS | 600 seconds | OCR and table parsing for large PDF files are time-consuming, requiring longer processing times. |
Chunk Length | 800–1200 characters | Balances semantic completeness and vector recall efficiency, preventing long paragraphs from diluting key information. |
Recall Count | Top 8 | Ensures coverage of potentially dispersed key information within documents, improving relevance. |
Similarity Threshold | 0.75–0.85 | Addresses the precise matching requirements for specialized terminology, setting a higher threshold to reduce false recalls. |
Rerank Return Count | Top 4 | Selects the most relevant segments through reranking, building on a high recall rate. |
Common Pitfalls
- Knowledge base click error, displaying
Failed to load knowledge base: This usually indicates incorrectSTORAGE_DRIVERorDB_URLconfiguration during private deployment, preventing FastGPT from connecting to storage or database services. - Specific fields (e.g., batch number, titer) in Q&A results are empty or inaccurate: This occurs when OCR recognition or table parsing during file preprocessing poorly handles special table formats and specialized terminology in AAV documents, leading to failed key information extraction.
- Uncaught exceptions in simple applications after an upgrade: This is typically due to uncleared Docker container caches, or incompatibility between new and old data models, resulting in incomplete upgrade script execution or service startup errors.
Verification Steps
- Upload an AAV batch record PDF file containing tables and specialized units. Verify that the knowledge base correctly parses table data and specialized fields like virus titer and host cell DNA residue.
- Test with a query containing an AAV batch number, test item, and result, for example, "Query endotoxin test result for batch number
AAV2023001". Verify that the system accurately extracts the corresponding information from relevant documents. - Simulate concurrent user queries. Observe if the system response time is within an acceptable range and check logs for excessive memory overflows or database connection errors.
- Perform a minor version upgrade. Confirm that the upgrade script runs without errors and that all existing knowledge bases and application functions operate normally after the upgrade.
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.