What the data for this category looks like
Baijiu market data comes from the China Securities Baijiu Index market data. Updates occur after market close on each trading day. No data is available on non-trading days. Data is delivered in structured format. It includes these fields: trade date, index point value, price change range, number of constituent stocks, and daily trading volume. Trade date uses the YYYY-MM-DD format. Index point value uses points as its unit. Price change range uses percentage as its unit. Daily trading volume uses ten thousand yuan as its unit. Number of constituent stocks is an integer value. Updated data is synced to the official market API. No extra fields are added.
Constraints imposed by these characteristics on model access and configuration
The data source only updates on trading days. Model invocation triggers must align with the update completion time on trading days. This avoids fetching invalid data after invocation. Data fields include multiple numeric and integer value types. Field mapping rules must be configured during model access. This ensures parsed fields match their correct type and unit, preventing parsing errors. The constituent stocks of the Baijiu Index may be adjusted. Data validation rules must be configured. This triggers alerts when fields are missing or abnormal. The data source update has a fixed delay. The model invocation time window must be set to 15 minutes to 1 hour after market close. This avoids fetching outdated, unupdated data. Model invocations triggered on non-trading days return empty data. Request interception rules for non-trading days must be configured. This prevents invalid invocation requests.
Configuration Settings
| Configuration Item | Recommended Value | Rationale |
|---|---|---|
request_trigger_cron | 0 19 * * 1-5 | Aligns with the market data update completion time 1 hour after market close on trading days, ensuring access to the latest daily data |
stream_mode | true | Adapts to large language models that only support stream mode, ensuring normal receipt of streaming return results |
field_mapping_rule | trade_date: Trade Date, index_value: Index Level, change_range: Price Change Range, daily_volume: Daily Transaction Volume | Matches the actual field names and display names of Baijiu market data, ensuring the model correctly parses data fields |
api_timeout | 600 seconds | Accounts for temporary delays in the data source, preventing premature request timeouts |
invalid_data_alarm | enabled | Triggers an alert when empty data from non-trading days or abnormal fields are retrieved |
max_retry_count | 3 times | Handles retries when the data source is temporarily unavailable, reducing the impact of single request failures |
The parameter values provided 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 settings.
Three Common Misconfigurations
- Setting
stream_modetofalsewhen invoking large models that only support stream mode results in empty return values or errors. This occurs because the model's output format requirements are not matched, and streaming data cannot be received normally. - Switching model channels after locally deploying the frontend results in a 304 status code returned in the browser console, and model channel configuration fails to load. This occurs because the frontend interface's caching policy is not correctly configured, leading to request interception.
- Using traditional SK/AK format keys when configuring model channels results in an
errormessage displayed in the interface. Invocation logs showdata acquisition exception. This occurs because the authorization format requirements of the current channel are not met, and identity verification cannot be completed.
How to Confirm Successful Configuration
- Manually trigger a model invocation. Check that the fields returned in the invocation log match those configured in
field_mapping_rule, with no empty fields or format errors. - Wait for the close of a trading day. Check that the scheduled task triggers normally, and complete market data is retrieved.
- Navigate to the model channel configuration page. Confirm that the authorization status is normal, with no error messages. Check that invocation logs show no
data acquisition exceptionorerrorentries. - Test invocation of a large language model that only supports stream mode. Confirm that the returned streaming results are complete, with no empty data or format abnormalities.
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.