Data Characteristics in This Category
Data for hematology-oncology product and reagent consultations primarily originates from clinical trial reports, drug monographs, medical research literature, and pharmaceutical company product manuals. This data updates frequently, especially when new drugs launch or clinical guidelines revise, which changes core information. Document structures typically include precise medical terminology, dosage instructions, indications, contraindications, adverse reactions, and pharmacological mechanisms of action. Common fields include generic drug name, brand name, batch number, manufacturer, active ingredient, content, dosage form, storage conditions, expiry date, and medical insurance code. Units strictly follow international standards, such as milligrams (mg), milliliters (mL), international units (IU), and moles per liter (mol/L), with extremely high precision requirements.
Constraints on Forms and Interaction from These Characteristics
Frequent data updates require an efficient knowledge base synchronization mechanism to avoid providing outdated information. Strict medical terminology and complex document structures necessitate precise and standardized user input in form design to prevent ambiguity. For example, confusion in dosage units can lead to severe consequences. The multi-source nature of the data requires integrating information from different sources during knowledge retrieval and clearly displaying its origin. The standardized nature of fields dictates strong validation for form inputs, such as format and logical validation for drug batch numbers and expiry dates. High-precision unit requirements mean that for sensitive information like dosage, interaction must provide clear unit selection or prompts to prevent user input errors.
Configuration Settings
| Configuration Item | Recommended Value | Rationale |
|---|---|---|
Chunk size (Segment Length) | 800–1200 characters (characters) | Clinical trial reports and literature paragraphs are often long, ensuring complete semantic units. |
Recall count (Recall Count) | Top 8 entries (top 8) | Ensures coverage of various aspects, such as indications, side effects, and dosage. |
Similarity threshold (Similarity Threshold) | 0.78–0.85 | Medical terminology is highly precise, requiring a higher threshold to reduce irrelevant information. |
maxContext | 4096 | Accommodates the context length required for complex medical queries, ensuring complete information. |
UPLOAD_FILE_MAX_SIZE | 500 MB | Handles large clinical trial reports or drug monograph PDF files. |
PARSE_FILE_TIMEOUT_SECONDS | 600 seconds (seconds) | Processes large file parsing, preventing failures due to timeouts. |
Three Common Mistakes
- Users enter incomplete or misspelled drug names, resulting in empty retrieval results. This occurs because the knowledge base index lacks sufficient alias processing and fuzzy matching optimization.
- Users enter numerical values in a form without specifying units, leading to system misunderstanding. This happens when the form design does not provide mandatory unit selection or clear prompts.
- After minimizing and then maximizing the FastGPT page, input content is lost. This may be due to improper handling of front-end page state management or caching mechanisms, failing to persist input field content.
Verification Steps
- Test queries with different drug names (including generic, brand, and aliases) to check if retrieval results are accurate and include core information.
- Submit queries containing different dosage units (e.g., mg, g, mL) to check if the system correctly identifies and provides feedback with the corresponding units.
- Upload multiple large PDF documents (e.g., clinical reports over 100 MB) to check if parsing is successful and content is complete.
- Simulate network fluctuations or prolonged operations to check if form input content persists, for example, if content remains after page scaling.
The values provided are common starting points. Measure them against your 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.