Blood Glucose Data Characteristics
Blood glucose data primarily originates from blood glucose meters, continuous glucose monitoring (CGM) devices, and patient self-recorded logs. Data characteristics include timestamps, blood glucose values (typically in mmol/L or mg/dL), and event markers (e.g., pre-meal, post-meal, exercise, medication). CGM devices update frequently, up to every 5 minutes, generating time-series data streams. Blood glucose meter data is often discrete points. Patient self-recorded logs may include richer contextual information, such as dietary content, exercise intensity, and emotional states. Document structures are typically structured or semi-structured, such as CSV, JSON files, or tabular data embedded in electronic health record systems. Field names are relatively standardized, but unit conversions and missing data are common.
Constraints Imposed by Data Characteristics on Deployment and Upgrades
High-frequency CGM data streams require real-time updates and indexing capabilities for the knowledge base, necessitating rapid ingestion and processing of time-series data. The diversity of blood glucose units (mmol/L and mg/dL) demands standardization during data preprocessing or flexible unit conversion mechanisms. Unstructured information from patient self-recorded logs, such as dietary descriptions, increases the complexity of text understanding, requiring more powerful semantic analysis capabilities. During deployment, the security and privacy of sensitive health data are core considerations; data transmission and storage must comply with relevant regulations. During upgrades, introducing new models or algorithms requires re-indexing or re-vectorizing historical data to ensure data consistency and query accuracy.
Configuration Guidelines
| Configuration Item | Recommended Value | Rationale |
|---|---|---|
UPLOAD_FILE_MAX_SIZE | 500 MB | Supports uploading structured files containing large amounts of historical CGM data. |
PARSE_FILE_TIMEOUT_SECONDS | 600 seconds | Allows sufficient time to parse large CSV or JSON blood glucose data files. |
maxContext | 3000 characters | Ensures inclusion of patient questions and multiple relevant blood glucose records, providing ample context. |
Chunk size | 800–1200 characters | Balances the information density per segment with the model's processing capacity, ensuring each segment contains a complete blood glucose event description. |
Recall count | Top 10 entries | Covers blood glucose data for multiple time periods or related events potentially involved in patient questions. |
Similarity threshold | 0.75 | Filters out irrelevant blood glucose data, improving the accuracy of retrieval results. |
Common Pitfalls
- Knowledge base index creation fails, with messages indicating the file is too large or a timeout occurred. This happens when default file size or parsing time limits are insufficient for report files containing months or years of blood glucose data.
- The model's responses mix or incorrectly convert blood glucose units. This occurs when blood glucose data from all sources is not uniformly standardized during preprocessing, or unit representations are inconsistent in the model's training data.
- The AI assistant cannot accurately interpret the correlation between patient-described dietary content and blood glucose fluctuations. This is due to a lack of sufficient knowledge in the knowledge base regarding diet-blood glucose correlations, or the model's limited ability to understand unstructured dietary text.
Verification Steps
- Upload a test blood glucose data file containing both mmol/L and mg/dL units. Check if the data in the knowledge base is standardized to the specified unit.
- Simulate a patient asking, "Has my blood glucose fluctuated much in the past week?" Verify that the assistant's response cites complete blood glucose data covering the specified time range.
- Upload a patient log containing typical dietary records. Ask, "How did my blood glucose change after I ate this meal?" Evaluate the assistant's accuracy in interpreting the correlation between diet and blood glucose.
- Through API calls, check if large files can be parsed and successfully ingested within the specified time after setting the
PARSE_FILE_TIMEOUT_SECONDSparameter.
The values provided are common starting points and should be measured against specific 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.