Data Characteristics for This Category
Data for culture media and consumables primarily comes from supplier technical documentation, internal quality control reports, and product specifications compliant with ISO 13485 standards. These documents are typically in PDF, Word, or structured database formats. Update frequency is relatively stable, primarily occurring during new product releases, formulation or material changes, and regulatory adjustments. Document structures for culture media often include formulation ingredients, preparation procedures, batch analysis reports (e.g., endotoxin, sterility testing), shelf life, and storage conditions. Consumables documentation focuses on material composition, manufacturing processes, biocompatibility test reports, sterilization validation reports (e.g., SAL values), and packaging information. Fields and units for culture media often involve ingredient concentrations in grams per liter (g/L), pH values, and osmolality (mOsm/kg). Consumables often include polymer types, dimensions (mm), pore sizes (µm), and specific biological performance indicators.
Constraints from These Characteristics on "Citing Sources and Traceability"
The characteristics of culture media and consumables data impose several requirements on source citation and traceability. Diverse document formats and structures demand powerful multi-format file parsing capabilities from the knowledge base to ensure accurate extraction of all key information. The relatively stable update frequency makes historical version management critical, ensuring citations can trace back to specific batches or versions. Specialized terminology, chemical names, and material science parameters within documents require the model to accurately understand them, preventing citation errors due to semantic deviations. Numerical information like batch reports and test data requires the citation system to identify and link to specific testing methods and units, ensuring traceability accuracy. For instance, when citing an endotoxin level for a culture medium, the system must trace back to the corresponding batch number and test report number, ensuring information authenticity and uniqueness.
Configuration Settings
| Configuration Item | Recommended Value | Rationale |
|---|---|---|
maxContext | 800–1200 characters | Technical descriptions for culture media and consumables are often detailed. Sufficient context is needed to capture complete product parameters and test results, preventing information truncation. |
Recall Count | Top 10–15 items | Ensures coverage of relevant document segments from different suppliers, batches, or test projects, improving citation comprehensiveness. |
Similarity Threshold | 0.75–0.85 | Accurately matches key technical parameters and report numbers, filtering out irrelevant generic descriptions, improving recall accuracy. |
Reranked Return Count | Top 5 items | Further optimizes ranking, prioritizing batch reports, test data, or product specifications most relevant to the current query, reducing manual screening effort. |
Segment Length | 300–500 characters | Balances the completeness of information within a single segment with model processing efficiency, ensuring each segment contains a complete technical point or test conclusion. |
PARSE_FILE_TIMEOUT_SECONDS | 600 seconds | Addresses the parsing needs for large PDF files (e.g., supplier product catalogs or multi-batch quality reports), preventing partial file ingestion due to timeouts. |
Three Common Mistakes
- The number of citation results does not match expectations, for example, only a few items are returned. This happens when the knowledge base segment length is set too small, causing key information to be fragmented, and individual segments cannot fully convey the intent.
- The numerical values cited by the model differ from the original document, for example, incorrect endotoxin units or mismatched values. This occurs when the knowledge base fails to correctly identify units or data formats when processing numerical data, leading to model misinterpretation.
- When retrieving batch reports, it is impossible to trace back to specific batch numbers. This happens when batch numbers are not extracted as independent and searchable entities during document parsing, or the batch number naming rules are not understood by the system.
How to Confirm Correct Configuration
- Select a culture medium technical document containing a specific batch number, test data, and units. Query the knowledge base and verify if the model's citation results accurately include the document's batch number, test values, and units, and can trace back to the original document link.
- Select a consumables product specification containing a biocompatibility test report. Query to verify if the model can accurately cite the report number, test method, and key conclusions. Check if the recall count covers all relevant test details.
- Simulate a query about a product update or change. Check if the knowledge base prioritizes citing the latest version of product specifications or change notifications, and can differentiate between old and new version information, confirming the effectiveness of the version management strategy.
The values provided are common starting points and should be measured against specific 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.