Tool Calling and Plugins for Hematology-Oncology Regulations

Hematology-oncology regulations and SOP documents primarily source data from National Medical Products Administration (NMPA) regulations, industry

Data Characteristics in this Domain

Hematology-oncology regulations and SOP documents primarily source data from National Medical Products Administration (NMPA) regulations, industry association treatment guidelines, and hospital internal management rules and clinical pathways. These documents update frequently, especially with new drug approvals, treatment protocol iterations, or regulatory adjustments. Updates can occur quarterly or even monthly. Document structures typically include standardized section titles such as "Scope of Application," "Definitions," "Responsibilities," "Operating Procedures," "Risk Control," and "Attachments." Common fields include drug name, dosage, administration route, treatment course, adverse reactions, monitoring indicators, and follow-up cycles. They also involve International Classification of Diseases (ICD) codes, International Nonproprietary Names (INN), and various biomarker abbreviations. Units strictly adhere to the International System of Units (SI), such as mg/kg, IU/mL, and %.

Constraints Imposed by These Characteristics on Tool Calling and Plugins

The high update frequency of hematology-oncology regulatory documents requires tool calling and plugins to have flexible data source synchronization mechanisms. This ensures the timeliness of retrieval results. Documents contain numerous specialized terms, abbreviations, and codes. This necessitates synonym expansion and code mapping when constructing queries to prevent retrieval failures due to terminology mismatches. Furthermore, the standardized document structure facilitates structured information extraction. However, it also requires plugins to identify and parse specific section content for precise targeting. For example, when a user asks about a drug's "administration route," the tool must extract this field from the "Operating Procedures" or "Drug Instructions" sections. Unit strictness demands unit conversion or validation in numerical comparison and calculation tools to avoid erroneous results from inconsistent units, such as conversions between mg and g.

Configuration Guidelines

Configuration ItemRecommended ValueRationale
chunkOverlapRatio0.15Ensures contextual continuity, addressing logical connections across paragraphs in regulatory documents.
top_k8Balances recall scope and relevance, suitable for potentially dispersed key information in hematology-oncology regulatory documents.
maxContext3000 TokensAccommodates the lengthy and information-dense nature of regulatory documents, ensuring the model receives sufficient context.
similarity_threshold0.78Ensures the precision of retrieved content, filtering out results unrelated to specialized terminology.
query_expansion_factor3Addresses the diversity of specialized terms and abbreviations in documents, improving recall coverage.
http_timeout_seconds60 secondsAccounts for potential network latency of external tools (e.g., drug database APIs), allowing sufficient response time.

Three Common Pitfalls

  • Calling external drug database APIs results in an HTTP 400 Bad Request error. A common cause is the drug name in the request parameters not being URI-encoded.
  • When chaining RAG knowledge bases and external tools in a workflow, retrieved document fragments from the RAG stage do not correctly map to the input fields required by the tool. This leads to tool call failures or empty results.
  • The model fails to standardize or expand disease names or drug abbreviations contained in user queries before calling the tool. This prevents the tool from recognizing them.

Configuration Validation

  • Simulate typical user questions. Check the retrieved content from the RAG knowledge base in the workflow. Ensure key information (e.g., drug names, dosages) is accurately extracted.
  • For scenarios involving external tool calls, review tool logs. Confirm the format and content of request parameters and return results meet expectations.
  • Design test cases with various specialized terms and abbreviations. Verify effective query expansion or terminology standardization occurs before tool calls.
  • Test scenarios involving numerical calculations or unit conversions. Ensure the tool's numerical output maintains unit consistency and accuracy.

The values given 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.