Data Characteristics for This Category
CAR-T cell therapy registration dossiers involve multi-modal data. Core data types include clinical trial reports, manufacturing process documents, quality control records, and non-clinical study data. Data sources are decentralized, encompassing Case Report Forms (CRFs) from clinical research organizations, Manufacturing Execution Systems (MES) and Quality Management Systems (QMS) within pharmaceutical companies, and non-clinical study reports from Contract Research Organizations (CROs). Data updates are frequent, especially during clinical trial phases, with continuous generation and revision. Document structures are complex, containing extensive unstructured text, tables, graphs, and images, such as flow cytometry data and ELISA results. Fields and units are highly specialized, for example, cell expansion (X-fold), viral vector copy number (VCG/cell), cell viability (%), and cytokine concentration (pg/mL). There are also numerous specific medical terms and abbreviations.
Constraints Imposed by These Characteristics on "Deployment and Upgrade"
The complex data characteristics of CAR-T cell therapy registration dossiers impose specific constraints on deployment and upgrade. Multi-modal data processing requires FastGPT deployments to have robust file parsing capabilities, particularly for recognizing embedded tables and image content within various formats like PDF, Word, and Excel. High-frequency data updates necessitate incremental synchronization mechanisms and version management features to ensure knowledge base timeliness and accuracy. Decentralized data sources and complex document structures make data cleaning and preprocessing critical, requiring customized data import scripts and preprocessing rules. Specialized fields, units, and extensive medical terminology demand high accuracy in model understanding and extraction. This may require fine-tuning domain-specific language models and continuously optimizing vocabulary and entity recognition capabilities during upgrades. Furthermore, data sensitivity mandates that deployment environments meet strict compliance and security standards, including data encryption and access control.
Configuration Settings
| Configuration Item | Recommended Value | Rationale for Recommendation |
|---|---|---|
UPLOAD_FILE_MAX_SIZE | 500 MB | CAR-T clinical trial reports and manufacturing batch records often contain numerous images and tables, leading to large individual file sizes. |
PARSE_FILE_TIMEOUT_SECONDS | 600 seconds | Processing complex PDF and Word documents, especially those with intricate charts, requires longer parsing times. |
maxContext | 8192 token | Paragraphs in CAR-T dossiers typically contain dense information and specialized terminology, requiring a longer context window for comprehension. |
Chunk size | 800–1200 characters | Ensures each text chunk retains sufficient professional context, preventing semantic fragmentation. |
Recall count | Top 10 entries | Increases the probability of retrieving relevant key information from a vast volume of specialized documents. |
Similarity threshold | Calibrate by actual measurement | Requires fine-tuning based on domain-specific corpora for specialized terms and abbreviations to avoid false positives or negatives. |
Three Common Mistakes
- Query results do not reflect the latest data after a knowledge base update. This usually occurs because incremental synchronization mechanisms are not configured or triggered correctly, leading to unrefreshed old data caches.
- Uploading a large PDF report results in a
File Parsing Timeout(file parsing timeout) error. This likely happens because thePARSE_FILE_TIMEOUT_SECONDSparameter is set too low, failing to accommodate the parsing demands of complex CAR-T documents. - Queries for specific CAR-T cell parameters (e.g.,
VCG/cell) return results missing relevant numerical values. This may be due to the document parser inadequately recognizing specialized fields embedded in charts or non-standard tables, or the model's extraction capability for specific units needs improvement.
How to Confirm Correct Configuration
- Upload a CAR-T clinical trial report PDF containing complex tables and charts. Verify that it is fully parsed and that key data from the charts can be retrieved from the knowledge base.
- Make minor revisions to an existing CAR-T manufacturing process document in the knowledge base, then perform an incremental update. Verify that query results immediately reflect the latest revisions.
- Use a series of query statements containing CAR-T domain-specific terms, abbreviations, and specific units. Verify that FastGPT accurately recalls relevant document snippets and extracts correct numerical values.
The values given are common starting points and should be measured against your 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.