Data Characteristics for This Category
Culture media and consumables data primarily originates from supplier product catalogs, technical specifications, Safety Data Sheets (SDS), and experimental protocols. These documents are typically in PDF, Excel, or online database formats. Data update frequency is relatively stable, with updates occurring when new products are released or formulations change, but generally not too often. Structurally, product catalogs usually contain structured information such as product name, catalog number, specifications, lot number, shelf life, and intended use. Technical specifications provide in-depth descriptions of product components (e.g., specific culture media formulations), physicochemical indicators, and quality control standards, which are semi-structured or unstructured content. Fields may include specialized metrics like "osmolarity," "pH value," and "endotoxin level," with units such as "mOsm/kg," "EU/mL," and "μg/mL." Some data may be embedded in tables within PDFs, requiring additional parsing.
Constraints Imposed by These Characteristics on Tool Calling and Plugins
The diverse data sources for culture media and consumables, particularly tables and unstructured descriptions embedded in PDF documents, demand robust parsing capabilities from the tool calling module. For example, when querying using a product catalog number, if the product name has multiple representations in the document, the tool needs semantic understanding. The non-real-time nature of data updates means additional steps may be required to confirm data source timeliness before tool invocation. The presence of specialized fields and units requires the tool to accurately identify and process these specific formats during data extraction and comparison, preventing errors due to unit confusion. Furthermore, product specifications and lot information are often used for traceability and compliance checks. Tool calling must ensure these critical details are accurately associated and extracted to support complex inquiry scenarios, such as querying quality reports for specific product lots.
Configuration Recommendations
| Configuration Item | Recommended Value | Rationale for Recommendation |
|---|---|---|
maxContext | 3000 Tokens | Technical documents for culture media and consumables contain extensive content, requiring a longer context window for complete understanding of product descriptions and technical details. |
Recall count (Recall Count) | Top 5 entries (Top 5) | Experience indicates that recalling the top 5 most relevant document snippets is sufficient to cover information needed for common inquiry questions. |
Similarity threshold (Similarity Threshold) | 0.75 | Setting a higher similarity threshold ensures that recalled documents are highly relevant to the user query, reducing interference from irrelevant information. |
PARSE_FILE_TIMEOUT_SECONDS | 600 seconds (600 seconds) | Parsing large PDF technical specifications can take a long time; sufficient timeout prevents task interruption. |
Chunk size (Segment Length) | 500 characters (500 characters) | Balances information completeness and recall efficiency. Prevents overly long segments from introducing too much noise, and overly short segments from losing context. |
Rerank result count (Rerank Return Count) | 3 entries (3 items) | Reranking based on recall results further improves the ordering of the most relevant information, ensuring core answers are prioritized. |
Three Common Mistakes
- Symptom: The tool calling module fails to extract product parameters from the knowledge base, and the answer lacks specific values. Reason: Specialized fields in knowledge base documents (e.g., "endotoxin level") are not correctly identified and mapped to the tool's parameter list.
- Symptom: The tool calling module in the workflow does not trigger the tool, instead providing a generic answer. Reason: The user query's keyword or intent match with the tool definition is insufficient, leading the system to determine that no specific tool needs to be called.
- Symptom: When querying a quality inspection report for a specific product lot, the tool returns "No relevant information found." Reason: Table data within PDF documents is not effectively parsed and indexed, preventing the tool from performing precise queries using the lot number.
How to Confirm Correct Configuration
- For typical product parameter queries (e.g., pH range of a specific culture medium), verify that the parameter values returned after tool calling match the data in the original technical specifications.
- Simulate various phrasing to test whether the tool consistently identifies user intent and correctly invokes the corresponding tool, for example, querying "shelf life of product A" and "expiration date of product A."
- Examine tool call logs to confirm that in multi-document scenarios, the knowledge base's recalled document snippets cover all critical information needed to solve the problem, without significant irrelevant content.
Note: The values provided are common starting points. Measure 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.