HTTP API and External Systems for Stability Study

Data Structure for This Category

Stability study data originates from long-term laboratory observation and analysis. It includes batch information, storage conditions (temperature, humidity, light), sampling time points, and various physicochemical indicators (e.g., content, dissolution, pH, moisture) and microbial limits. This data typically exists in structured tabular formats like Excel, CSV files, or directly within Laboratory Information Management Systems (LIMS). Data update frequency is low, generally following a predefined sampling schedule such as 0, 3, 6, 9, 12, 18, 24, 36, 48, 60 months. Document structures are rigorous, often including study protocols, raw data records, and analysis reports. Field names are standardized, and units are explicit, for example, Temperature (°C) (Temperature (°C)), Relative Humidity, in %RH (%) (Relative Humidity (%)), Content (%) (Content (%)).

Constraints Imposed by These Characteristics on "HTTP API and External Systems"

The low update frequency of stability study data means external systems calling FastGPT APIs for incremental synchronization do not require frequent full updates. Instead, scheduled tasks can handle updates. The highly structured nature of the data requires APIs to accurately parse tabular data and identify key fields such as batch_id, sampling_time, and metric_name. The multi-batch, multi-time point data characteristics mean a single file can contain a large amount of information, necessitating a high UPLOAD_FILE_MAX_SIZE parameter for file upload APIs. Additionally, raw data records and analysis reports may be in PDF format, requiring FastGPT's file parsing capabilities to ensure critical information is correctly extracted. Standardized field units help build more accurate knowledge graphs and question-answering models, preventing misinterpretations due to unit confusion.

Configuration Settings

Configuration Item	Recommended Value	Rationale
`UPLOAD_FILE_MAX_SIZE`	`500 MB`	Individual stability study reports or LIMS export data files can be large; large file uploads must be supported.
`PARSE_FILE_TIMEOUT_SECONDS`	`600 seconds`	Parsing large PDF reports is time-consuming and requires a longer timeout.
`chunk_size`	`800 characters`	Ensures each chunk contains sufficient contextual information while avoiding redundancy from excessive length.
`chunk_overlap`	`100 characters`	Maintains connections between chunks, reducing the risk of information loss.
`maxContext`	`4000 token`	Stability study reports are often lengthy, requiring a larger context window to understand complete information.
`embedding_model`	`text-embedding-ada-002`	Suitable for semantic understanding of text in the biomedical domain, improving recall accuracy.

Common Pitfalls

After calling the file upload API, a successful response is not received for a long time, or a 504 Gateway Timeout is returned. This occurs if the uploaded file size exceeds the UPLOAD_FILE_MAX_SIZE limit or if file parsing time exceeds PARSE_FILE_TIMEOUT_SECONDS.
Querying stability data via API returns results missing key indicators or batch information. This may happen if FastGPT fails to correctly parse the tabular structure in the source file, leading to unextracted specific fields.
When an external system calls the GET /openapi/v1/knowledge_base/{kb_id}/files API, the file list is incomplete or an empty list is returned. This can occur if the knowledge base ID kb_id was not correctly specified during file upload, causing the file to be uploaded to the wrong knowledge base.

How to Verify Configuration

After uploading a typical stability study PDF report, check if FastGPT has successfully parsed the file in the knowledge base and if key information from the report can be found via keyword search.
Upload a CSV file containing multi-batch, multi-time point data via FastGPT's API. Then, use the query interface to verify if specific indicator data for a particular batch at a specific time point can be accurately retrieved.
Use FastGPT's question-answering feature to ask questions about a stability batch's data. Check if the returned results accurately reference relevant data from the knowledge base and can recognize units like Content (%) (Content (%)) and dissolution (%) (Dissolution (%)).

The values provided are common starting points and should be measured against the reader's own 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.