What this category’s data looks like
TCM marketing content data comes from internal enterprise product archives, official public pharmacopoeia materials, clinical usage feedback documents, and customized marketing materials. Update rhythm adjusts based on marketing nodes and product iterations. Batch update copy materials during seasonal wellness campaigns, free clinic events, and similar nodes. Sync product details and compliance documents when new products launch.
Document types include structured product parameter tables, long-text processing specification descriptions, and unstructured image-text materials and short video scripts. Fields include professional TCM terminology. Units are mostly grams, milliliters, and treatment course days. Some materials include compliance identifiers such as national drug approval numbers.
What constraints these characteristics impose on deployment and upgrade
Mixed structured and unstructured document types require deployment to support both text parsing and image OCR. Otherwise, valid information in marketing materials cannot be fully extracted.
Frequently updated marketing materials require deployment configurations to support scheduled sync tasks, avoiding delays from manual updates.
Professional terminology and compliance fields require retaining existing metadata indexing logic during upgrades, preventing damage to pre-configured compliance validation rules.
Some small and medium-sized pharmaceutical companies use Arm architecture servers. Deployment packages must support this architecture, otherwise container startup failures will occur.
How to set configurations
| Configuration Item | Recommended Value | Rationale |
|---|---|---|
PARSE_FILE_TIMEOUT_SECONDS | 600 seconds | TCM marketing content includes long-text processing specifications and high-resolution photos of medicinal materials, so parsing time exceeds that of general product categories |
UPLOAD_FILE_MAX_SIZE | 1000 MB | Some marketing material packages include batches of high-resolution formula diagrams and long video scripts, so this upper limit covers single-file requirements |
Chunk Length | 800–1200 characters | Most TCM terms are coherent combinations; this range avoids splitting that disrupts the integrity of professional expressions |
Recall Count | Top 6–8 results | TCM marketing requires a balance between professional accuracy and reading fluency; too many recalled results will cause content redundancy |
ENABLE_IMAGE_PARSE | Enabled | TCM marketing materials include many photos of medicinal materials and formula diagrams, so OCR is needed to extract text content |
ARM64_DEPLOY_SUPPORT | Enabled | Adapts to Arm architecture servers commonly used by small and medium-sized pharmaceutical companies, avoiding deployment compatibility issues |
The parameter values listed on this page are common starting points for configuration. Actual values are affected by material form, data volume, and business rules. Specific issues require individual analysis. It is recommended to test on your own samples before finalizing.
Three common errors
- Issue: Slow download speed when pulling the
fastgpt-minioimage, causing the deployment process to stall. Cause: No domestic mirror acceleration source configured; the official image repository loads slowly in overseas network environments. - Issue: Unable to add image datasets after private deployment, and no image content is extracted during knowledge base parsing. Cause: The
ENABLE_IMAGE_PARSEconfiguration item is not enabled, or OCR dependency components are not loaded during deployment. - Issue: Parsing errors occur for existing TCM marketing CSV files after upgrading the version. Cause: The new version adjusted the default value of the
CSV_CHUNK_OVERLAPparameter; the original chunking logic does not match the context continuity requirements of long TCM texts.
How to confirm the configuration is complete
- Access the system configuration page of the deployment backend, check the file parsing module, and confirm that the
ENABLE_IMAGE_PARSEswitch is enabled. - Upload a single high-resolution photo of a medicinal material, wait for parsing to complete, and verify that extracted text content is complete to confirm OCR function operates normally.
- Upload a test TCM marketing CSV file, review parsed chunking results, and confirm that chunk length matches the preset range.
- Run the
docker pscommand, confirm that the architecture of all containers matches the current server architecture, and no failed startup services are present.
Question material comes from public community discussions. Configuration values are common starting points and should be measured against your own samples. Verified on 2026-09-14.