Data Characteristics in This Domain
Hematologic oncology pharmacovigilance data originates from various sources. These include clinical trial reports, real-world evidence (RWE) studies, adverse event reporting systems (e.g., FAERS, EudraVigilance), and medical literature. Data update frequencies vary; clinical trial data publishes as research progresses, while adverse event reporting systems provide continuous data streams. Document structures are diverse, encompassing both structured database records and unstructured clinical notes and literature abstracts. Key fields include patient demographics, diagnosis, medication regimens (drug name, dosage, duration), adverse event (AE) descriptions, severity, onset time, outcome, and causality assessment. Some data involves International Classification of Diseases (ICD) or Medical Dictionary for Regulatory Activities (MedDRA) coding. Units are common for dosage (milligrams, units), time (days, months), and laboratory indicators.
Constraints Imposed by These Characteristics on Tool Calling and Plugins
The multi-source and heterogeneous nature of hematologic oncology pharmacovigilance data requires robust data integration and standardization capabilities from tool calling and plugins. The prevalence of unstructured text necessitates natural language processing (NLP) plugins for information extraction and entity recognition. High-frequency data sources, such as adverse event reporting systems, demand real-time or near real-time data synchronization from plugins to ensure timely alerts. The inclusion of specialized coding systems like MedDRA means tools must be able to invoke terminology mapping or coding plugins. These plugins convert free text into standardized terms, enabling unified analysis across data sources. Furthermore, the presence of key numerical fields like drug dosage and treatment duration requires plugins to accurately parse and quantitatively compare these values.
Configuration Guidelines
| Configuration Item | Recommended Value | Rationale |
|---|---|---|
maxContext | 1000–1500 characters | Balances the completeness of complex medical text with model processing efficiency. |
Recall Count | Top 8–12 entries | Covers multi-source data while preventing interference from excessive irrelevant information. |
Similarity Threshold | 0.75–0.85 | Ensures relevance, reduces false positives, and meets the precision requirements of medical terminology. |
PARSE_FILE_TIMEOUT_SECONDS | 600 seconds | Addresses the parsing time needed for large clinical reports and literature. |
http_timeout_seconds | 300 seconds | Accommodates potential delays from external adverse event database interfaces. |
extract_entity_types | Drug Name, Adverse Event, Dosage, Treatment Duration | Precisely extracts core information for hematologic oncology pharmacovigilance. |
Common Pitfalls
- An external API call returning a 502 error may indicate that the FastGPT environment cannot directly access the target API address. Check network configurations or proxy settings.
- Search plugins returning links whose content cannot be automatically read leads to incomplete information. This occurs because most plugins only return search result summaries, lacking deep parsing capabilities for link content.
- Not all deployed models appearing in the tool list typically results from incorrect model interface configurations or restricted permission settings, preventing the system from recognizing or loading all available models.
Verification Steps
- Use FastGPT's debugging interface to perform information extraction on a clinical report containing complex medical terminology. Verify that key fields (e.g., drugs, adverse events, dosages) are accurately identified.
- Simulate a query to an external adverse event reporting system. Observe if the HTTP plugin successfully retrieves data and if the returned JSON structure matches expectations.
- Use a query containing MedDRA codes. Verify that the terminology mapping plugin correctly converts free text descriptions into standardized codes and observe the accuracy of the coding results.
- Conduct multi-turn dialogue tests. Ask questions about the side effects of specific hematologic oncology drugs. Evaluate whether tool calling and plugins can integrate multi-source information and provide logically clear answers.
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.