What the data for this category looks like
Data related to biologics yield rates comes from two main sources: public batch issuance interfaces of the National Medical Products Administration, and third-party pharmaceutical industry data service platforms. There are two update frequency schedules: batch issuance data is updated once monthly, and terminal sales data is updated once weekly. Each individual data document includes core fields such as product generic name, batch issuance batch number, reporting period, unit production volume, winning bid unit price, and cumulative revenue. Unit production volume is measured in vials or bottles, winning bid unit price is measured in yuan per unit dose, and cumulative revenue is measured in ten thousand yuan. Data is organized into groups by reporting period, with each group containing full batch data for the corresponding period.
Constraints imposed by these characteristics in the model integration and configuration workflow
Data sources with multiple update frequencies require configuration of multi-dimensional scheduled pull rules, with separate synchronization intervals for monthly batch issuance data and weekly sales data. Fields include multiple unit types, so precise field mapping and unit conversion rules must be configured to avoid unit mismatch issues during calculations. Differences in reporting periods across data sources require configuration of data alignment logic to ensure cross-source yield rate calculations are based on a unified time window. Additionally, individual biologics data documents have a large length, so sufficient parsing and processing time must be reserved to avoid timeout errors.
How to set configurations
| Configuration Item | Recommended Value | Rationale |
|---|---|---|
model_provider | ollama or xinference | Supports locally deployed open-source models, adapts to the user's local computing power environment |
prompt_template | Please generate a yield rate and market trend summary for the specified period based on the following biologics data: {context} | Clearly defines the model output topic, binds data source field variables, and controls the scope of output content |
field_mapping_rule | Strictly match field names + unit conversion | Biologics fields include multiple unit types, so clear mapping rules are required to avoid calculation deviations |
data_sync_cron | 0 0 2 * * * | Triggers synchronization at 2:00 AM daily, covering the update windows for monthly and weekly data |
similarity_threshold | 0.75–0.85 | Filters low-correlation market data, retaining information strongly associated with the target biologics |
parse_timeout | 300 seconds | Biologics data documents have a large length, so sufficient parsing and processing time must be reserved |
The parameter values provided on this page are common starting points for configuration setup. Actual values are affected by material form, data volume, and business rules. Specific issues require case-by-case analysis, and it is recommended to test on your own samples before finalizing settings.
Three common configuration errors
- Issue: When using a model deployed via
ollama, custom prompt templates do not take effect, and model output deviates from the preset topic. Cause: The prompt template is not attached to the model's conversation context entry, and only adding it in global configuration does not trigger it. - Issue: When accessing the platform via a domain name, model configuration cannot be saved, and the interface prompts that configuration fields are empty. Cause: Cross-domain access rules for the domain name are not properly configured, preventing the frontend from loading model configuration options.
- Issue: After configuring
data_sync_cron, data does not update as scheduled. Cause: The Cron expression format for the scheduled task is incorrect, and it does not match the system's default time zone.
How to confirm the configuration is complete
- Access the model configuration page, verify that the
model_providerfield has selected the corresponding service provider, and that themodel_namefield matches the name of the locally deployed model. - Initiate a test query, enter keywords for the target biologics, and confirm that returned content includes data items specified in the configured field mapping rules.
- Review data synchronization logs to confirm that scheduled tasks have triggered per the configured
data_sync_cronexpression, with no error records. - Access the platform via a domain name, attempt to add a new model configuration, and confirm that the interface loads all configuration options normally, with no field missing prompts.
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.