What Data Looks Like for This Category
This category’s data primarily comes from raw material and auxiliary material qualification certificates, ERP/MES systems, weighing systems, oven or frying line equipment, laboratory testing, and food trace ledgers. Daily documents include formulas and feeding lists, production records, baking or steaming curves, metal detection records, cleaning and disinfection records, and factory inspection reports. Data is updated by production batch, shift, or daily. Document fields include feeding quantity, baking temperature, proofing duration, allergen labeling, shelf life, etc. Units include kilograms, degrees Celsius, minutes, days, and more.
Constraints Imposed by These Characteristics on the Knowledge Base Retrieval and Recall Link
Data for this category is updated by production batch, shift, or daily, and frequent line changes lead to frequent shifts in formulas and allergen information. This requires knowledge base retrieval and recall to quickly match the latest relevant data for current production. Document types cover structured feeding lists, unstructured baking curves and test records, so the system must support accurate recall of multi-format content. Data related to allergens and foreign matter control has high sensitivity, so recall results must strictly match the current production’s cleaning and material requirements to avoid mismatches. Single-batch data volume fluctuates with production rhythm, so the number of retrieval results must be reasonably controlled to avoid redundant information interfering with the drafting of process documents and exception reports.
Configuration Settings
These configurations are based on FastGPT v4.17.0.
| Configuration Item | Recommended Value | Rationale |
|---|---|---|
similarity | 0.75–0.85 | Matches the mixed structured and unstructured document scenario of this category, balancing recall accuracy and coverage |
limit | Top 6 entries | Controls the number of single-batch retrieval results to avoid redundant information interfering with process document drafting, aligning with the fluctuating single-batch data volume characteristics of this category |
searchMode | mixedRecall | Balances keyword matching and semantic recall, adapting to document types that have both structured fields and unstructured curves for this category |
usingReRank | Enabled | Re-ranks recall results to improve the recall priority of critical information such as allergens and feeding parameters |
datasetSearchUsingExtensionQuery | Enabled | Expands retrieval terms based on the original question, adapting to the recall of professional terms such as formulas and process parameters for this category |
collectionIds | Bound by production workshop/product line | Precise retrieval scope limitation, avoiding interference from irrelevant data across workshops/product lines, matching the frequent line change business scenario of this category |
Three Common Misconfigurations
- Phenomenon: Markdown format content cannot be rendered in context references, and original markup symbols such as
|are displayed. Cause: The original document format was not preserved during knowledge base chunking, or relevant format retention settings were not configured in the retrieval and recall process. - Phenomenon: After submitting a query, specified files in the knowledge base cannot be referenced, and the returned results are empty or irrelevant. Cause: The
collectionIdsconfiguration is incorrect, the corresponding document collection is not bound, or thesimilaritythreshold is set too high, leading to excessive precision filtering. - Phenomenon: A 60-second timeout occurs when switching the knowledge base index, or the status remains "training in progress" or "rebuilding in progress" even when the knowledge base has little content. Cause: A reasonable index update trigger mechanism was not configured for the frequently updated data of this category, or insufficient system resources lead to index operation delays.
How to Verify the Configuration Is Correct
- Upload a feeding list document that contains Markdown formatting, initiate a retrieval, and check the context references to confirm that the format renders correctly.
- Enter the allergen material keyword for the current production batch, and verify that the retrieval results only return cleaning and feeding documents for the corresponding workshop or product line.
- Submit a query that includes professional process parameters, and check whether the number of returned results matches the preset
limitconfiguration. - Manually trigger a knowledge base index rebuild, confirm that the operation completes within a reasonable time frame, with no persistent timeouts or abnormal statuses.
The values provided are common starting points and should be measured against the reader's own samples.