Data Characteristics for This Category
Real-World Evidence (RWE) quality documentation data originates from diverse sources, including Electronic Health Records (EHR), insurance claims databases, registries, and Patient-Reported Outcomes (PRO) data. Update frequencies vary. EHR data may update in real-time or daily, while claims data typically updates in monthly or quarterly batches. Document structures for RWE quality documentation are often semi-structured or unstructured, containing extensive clinical narratives, diagnostic codes, medication information, laboratory results, and adverse event reports. Fields include standard patient identifiers and timestamps, as well as disease-specific indicators, treatment plan details, and efficacy assessment scale scores. Units range from milligrams and milliliters to international units and percentages, with potential unit inconsistencies across different data sources.
Constraints Imposed by These Characteristics on "HTTP API and External Systems"
The multi-source nature and varied update frequencies of RWE quality documentation necessitate highly flexible and configurable HTTP APIs. These APIs must accommodate different data retrieval strategies. For example, EHR data may require incremental synchronization support, while claims data needs batch import capabilities. Semi-structured and unstructured document characteristics mean that APIs cannot rely solely on strict JSON or XML schemas for data reception. They must support various content types, such as file uploads and text stream transfers, and demand more sophisticated internal parsing logic. The complexity of fields and units requires external systems to perform rigorous preprocessing and standardization before sending data to FastGPT. This includes unifying medical terminology codes or converting units to ensure accuracy and consistency in knowledge base ingestion, preventing ambiguity during vectorization and retrieval.
Configuration Guidelines
| Configuration Item | Recommended Value | Rationale |
|---|---|---|
UPLOAD_FILE_MAX_SIZE | 500 MB | RWE documents often include large PDF reports or imaging data, requiring large file upload support. |
PARSE_FILE_TIMEOUT_SECONDS | 600 seconds | Complex RWE document parsing can be time-consuming; this prevents parsing interruptions. |
Chunk size | 800-1200 characters | Balances context completeness and retrieval efficiency, suitable for clinical narrative texts. |
Similarity threshold | Calibrated by actual measurement | Ensures retrieval of highly relevant document segments for RWE questions, avoiding irrelevant information. |
Rerank result count | Top 5 entries | Improves the relevance of the final RWE answers presented to the user, optimizing query results. |
SSE_CONNECTION_TIMEOUT | 300 seconds | Addresses intermittent delays in external system SSE connections, maintaining connection stability. |
Common Pitfalls
- An external system uploads a file to FastGPT and receives the error "Key is error. You need to use the app key rather than the account key." This indicates an incorrect API key type was used. An application-level API Key is required for authentication.
- Connecting to an SSE interface with Header authentication results in an error "Authentication Not Supported vials Supported" (authentication not supported). This may occur if FastGPT's HTTP API configuration does not correctly recognize or pass authentication information within custom HTTP Headers.
- After calling an external HTTP API in a workflow, a specific field in the returned result is empty. This happens when the data structure returned by the external system does not match the field path defined in the FastGPT workflow.
Verification Steps
- Upload a typical RWE document via the FastGPT administration interface. Check that corresponding segments are successfully created in the knowledge base and that content is complete and readable.
- Simulate an HTTP API call from an external system. Observe FastGPT logs to confirm that request parameters, authentication information, and return status codes are as expected.
- Configure a simple FastGPT application. Ask questions related to the uploaded RWE document to evaluate answer accuracy and relevance, confirming that knowledge retrieval and generation logic function correctly.
The values provided are common starting points. Measure performance against specific samples to determine optimal configurations.
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.