Data Characteristics in This Category
DTP pharmacy pharmacovigilance data originates from pharmacy sales systems, patient feedback records, and adverse drug reaction (ADR) reports from pharmaceutical companies or regulatory bodies. Data updates are frequent, typically in daily or weekly batches. High-risk drug adverse event reports may be entered in real-time. Data documents are often structured or semi-structured, such as CSV, JSON, or XML files. Content includes patient basic information (anonymized), drug names, batch numbers, dosages, administration methods, adverse reaction descriptions (free text), occurrence times, and treatment measures. Adverse reaction descriptions frequently involve medical terminology and non-standard colloquial expressions. Dosage units commonly include milligrams (mg), grams (g), and milliliters (ml). Time units are days, weeks, and months.
Constraints Imposed by These Characteristics on Tool Calling and Plugins
High-frequency data updates require efficient data retrieval and processing capabilities for tool calling. This prevents data lag from affecting the timeliness of pharmacovigilance. The coexistence of structured and semi-structured data necessitates tools that can flexibly adapt to different data format parsers. Free text and medical terminology in adverse reaction descriptions demand high natural language processing capabilities from the model. Plugins are needed for terminology standardization and entity recognition to accurately extract key information. During multi-source data integration, data cleaning and deduplication are crucial to prevent duplicate reports or erroneous data from interfering with analysis. Additionally, critical identification information like drug batch numbers requires high precision from tools during data matching and association.
Configuration Guidelines
| Configuration Item | Recommended Value | Rationale |
|---|---|---|
maxContext | 32000 token | Ensures accommodation of detailed descriptions and relevant medical background information from DTP pharmacy adverse reaction reports. |
PARSE_FILE_TIMEOUT_SECONDS | 600 seconds | Handles batch import of adverse drug reaction report files, addressing large file sizes or network transmission delays. |
reranker_top_n | Top 5 entries | In pharmacovigilance scenarios, recalls a small number of highly relevant key information items, reducing false positives. |
similarity_threshold | 0.78 | Filters knowledge base entries that highly match patient adverse reaction descriptions, ensuring result accuracy. |
chunk_overlap_ratio | 0.15 | Maintains contextual coherence when processing long free-text adverse reaction descriptions, preventing information fragmentation. |
tool_call_retries | 3 times | Addresses occasional transient network fluctuations or service unavailability of external drug databases or regulatory platform interfaces. |
Three Common Mistakes
- External drug knowledge base calls return empty or incomplete results. This typically occurs when tool call parameters do not match interface requirements. Examples include unstandardized drug name fields or missing authentication information.
- Multimodal models report parameter errors when processing adverse reaction images. This may relate to incompatible file formats, image sizes exceeding limits, or incorrect population of the
image_urlfield for models likeglm-4v-plus. - Plugin edits show "Unsaved" even after clicking the save button. This often indicates the backend service did not correctly receive or persist the configuration. Check FastGPT logs for configuration write failure error messages.
How to Verify Correct Configuration
- Use the FastGPT interface. Select a typical adverse reaction report text and trigger a tool call. Observe the logs to confirm successful calls to external drug database interfaces and check if the returned results contain the expected drug information.
- Upload an image containing drug packaging or symptoms. Use a multimodal model for analysis. Verify if the model correctly identifies image content and returns relevant descriptions. Compare the accuracy of model output with expected results.
- Modify a plugin configuration item and save it. Then refresh the page or re-enter the editing interface. Confirm that the modified configuration item persists to verify proper configuration persistence.
- Simulate a drug batch update operation. Observe if FastGPT retrieves and processes new data promptly. Verify that no errors or warnings occur during the data processing flow.
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.