HTTP Interface and External Systems for Solid Tumor Quality

Data Characteristics

Solid tumor-related quality documentation typically comes from diverse sources. These include clinical trial protocols, investigator brochures, ethics approval documents, drug manufacturing batch records, quality standards, and inspection reports. The update frequency of these documents varies based on the clinical trial phase, drug development progress, and regulatory requirements. For example, clinical protocol revisions may occur monthly, while batch records are generated with each batch.

Document structures often combine scanned PDFs and structured data (e.g., CSV, JSON). PDFs contain extensive unstructured text, charts, and signature pages. Structured data primarily holds trial data and quality control results. Specific fields include RECIST 1.1 assessment results, PD-L1 expression levels, TMB values, and KRAS mutation status. Units involved are mm (millimeters), % (percentage), and ng/mL (nanograms per milliliter).

Constraints from "HTTP Interface and External Systems"

The complexity of solid tumor quality documentation imposes specific requirements on HTTP interfaces and external system integration.

First, large volumes of unstructured PDF documents require efficient text extraction and image recognition capabilities. This means interfaces must handle large file uploads and integrate with OCR services.

Second, varying document update frequencies necessitate flexible scheduling mechanisms. Examples include real-time updates via Webhooks or periodic polling of specific directories.

Third, when transmitting specific fields from structured data, such as tumor size or gene mutation information, through HTTP interfaces, a clear JSON Schema is essential. This ensures data type and unit accuracy. For assessment standards like RECIST 1.1, interface design must account for its multi-layered structured representation to avoid information loss.

Finally, robust error handling and retry mechanisms are necessary to address various error codes from external systems. These can include "file too large," "unsupported format," or "data validation failed."

Configuration Guidelines

Configuration Item	Recommended Value	Rationale
`UPLOAD_FILE_MAX_SIZE`	`500 MB`	Solid tumor documents often contain many images and scanned pages, resulting in large file sizes. Support for large file uploads is necessary.
`PARSE_FILE_TIMEOUT_SECONDS`	`600 seconds`	Parsing large files, especially during the OCR process, can be time-consuming. This prevents parsing failures due to timeouts.
`maxContext`	`3000 characters`	Document paragraphs are often long. Sufficient context is needed to understand complex medical descriptions and experimental results.
`Chunk size`	`800–1200 characters`	Ensures each segment contains enough clinical context while avoiding excessive length that could impact retrieval efficiency.
`Recall count`	`Top 8 entries`	Quality documentation queries often require cross-referencing information from multiple angles. Increasing the number of retrieved items improves coverage.
`Similarity threshold`	`0.75`	For domains with many specialized terms, a higher threshold ensures the precision of retrieval results.

Common Pitfalls

HTTP requests return a 404 error. This is often due to incorrect external system interface address configuration or an improperly passed API Key.
Voice input functionality fails to enable, appearing as an unresponsive interface or an error. This is typically because the Whisper speech model is not correctly deployed or the TTS_SERVER environment variable in FastGPT does not point to the correct model service address.
After uploading a large PDF document, content parsing is empty. This can occur if the document contains many scanned images that have not undergone OCR processing, or if PARSE_FILE_TIMEOUT_SECONDS is set too short, causing parsing to abort.

Verification Steps

Upload a PDF document containing RECIST 1.1 assessment results. Check if the knowledge base correctly extracts and indexes key information such as tumor size and target lesion count.
Submit a simulated quality control report to an external system via an HTTP interface. Verify that the external system successfully receives and processes it. Also, confirm that FastGPT can receive processing results from the external system via Webhook.
Use a query containing a specific gene mutation (e.g., KRAS G12C). Verify that the system can retrieve relevant clinical trial protocols or inspection reports from the knowledge base. Check if the Similarity threshold effectively filters out irrelevant results.

The values provided are common starting points. Measure 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.