Data Characteristics for This Category
Culture media and consumable product data comes from various sources. These typically include product manuals, technical specifications, safety data sheets (SDS) from suppliers, and internal lab validation reports. Data update frequency is stable; updates occur when new products launch or existing products improve, but not frequently. Document structures are often PDF or Word formats. They contain fields such as product name, catalog number, batch, specification, components, storage conditions, shelf life, and application range. Some data may be embedded as images, for example, microscopic images of culture media. Field units typically involve physical quantities (e.g., mL, g, ℃), concentrations (e.g., %, mM), and packaging quantities (e.g., packs, boxes).
Constraints Imposed by These Characteristics on "Forms and Interaction"
The discrete nature of data sources requires the system to integrate data from multiple sources, especially for parsing unstructured documents. The stable update frequency allows for periodic bulk update strategies, reducing real-time synchronization pressure. Image information in documents, such as microscopic images, requires conversion into queryable metadata through visual recognition or manual annotation. The standardized nature of field units requires form designs to provide clear unit selection or automatic unit conversion functions to prevent data misinterpretation due to inconsistent units. Product components are complex, requiring interactive forms to support multi-level linked selection and input. For example, selecting a culture medium type should automatically load the corresponding component list. This also requires real-time validation during user input, such as judging the reasonableness of storage temperature ranges, to ensure input data validity.
Configuration Settings
| Configuration Item | Recommended Value | Rationale for Recommendation |
|---|---|---|
MAX_DOC_SIZE_MB | 20 MB | Ensures sufficient file upload capacity, considering that a single product technical document often includes images and detailed descriptions. |
FIELD_EXTRACTION_MODEL | OCR + LLM | Addresses embedded text in images and complex table structures common in documents, improving field extraction accuracy. |
FORM_VALIDATION_RULES | JSON FormatRule Set | Flexibly defines field type, range, and unit validation logic, adapting to unique requirements of different products. |
CONTEXT_WINDOW_SIZE | 4096 tokens | Guarantees sufficient context information when processing product inquiries involving multiple components and parameters. |
RESPONSE_RELEVANCE_THRESHOLD | 0.75 | Ensures returned results are highly relevant to the user query, filtering out low-confidence information. |
FALLBACK_STRATEGY | Human Intervention Queue | Provides a path for human expert intervention for complex queries that the AI cannot accurately answer. |
Three Common Mistakes
- After a user submits a form, the system displays "Cannot convert undefined or null to object." This usually indicates a mismatch between backend validation rules for form fields and the data type passed from the frontend, or a required field is empty.
- AI responses contain missing product parameter information or incorrect units. This occurs because the original document's field extraction was incomplete, or the unit recognition and conversion logic has flaws.
- In multi-turn interactions, the AI fails to correctly associate product information mentioned in the previous question. This usually happens when
CONTEXT_WINDOW_SIZEis set too small, causing historical conversation information to be truncated.
How to Verify Configuration
- Upload typical culture media and consumable product manuals. Check if the system accurately extracts all key fields (e.g., catalog number, components, shelf life) and verify correct unit recognition.
- Simulate user input in forms with various data ranges (including boundary and abnormal values). Confirm that all
FORM_VALIDATION_RULEScorrectly trigger validation prompts and that error messages are clear. - Conduct multi-turn conversation tests. Ask questions about product components, storage conditions, and application compatibility. Observe if the AI maintains conversational context and provides accurate answers.
Note: The values provided are common starting points. Measure them against specific samples to determine optimal settings.
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.