Multi-turn Dialogue and Prompting for Feed Intelligent Due Diligence Reports

Feed due diligence data sources include feed ingredient purchase ledgers, farm feeding records, third-party ingredient test reports, and industry

What the Data for This Category Looks Like

Feed due diligence data sources include feed ingredient purchase ledgers, farm feeding records, third-party ingredient test reports, and industry circulation monitoring data. Purchase ledgers and feeding records are structured Excel files, with fields including ingredient name, batch number, purchase/feeding date, and supplier information. Test reports are mostly multi-page PDF files, containing measured values for indicators such as crude protein, moisture, and heavy metals. Data update cycles vary: purchase and feeding data is updated daily, test reports are issued with each ingredient batch, and industry circulation data is updated weekly. Indicator value units include grams per kilogram and milligrams per kilogram, and some ledgers use simplified labeling.

Constraints Imposed on Multi-turn Dialogue and Prompting

The multi-source and decentralized nature of feed data requires multi-turn dialogue to first guide users to clarify data source types. This avoids field misalignment caused by mixed parsing of structured and unstructured files. The diverse unit characteristics of the data require prompts to explicitly define unified unit conversion rules. Without this, the model may mix up units for indicators across different batches. Differing update cycles require dialogue to first confirm the data time range. This prevents mixing expired test reports with real-time purchase data when generating due diligence content. Diverse document formats require specifying parsing rules for different file types in the configuration. This ensures accurate column mapping for Excel ledgers and correct keyword extraction for PDF test reports.

Configuration Settings

Configuration ItemRecommended ValueRationale
maxContext8000–12000 charactersFeed due diligence requires associating multiple batches of ingredient data and multiple test reports. An overly long context may cause the model to forget key batch parameters
systemPromptRequire all values to be uniformly converted to grams per kilogram or milligrams per kilogram, and verify field and unit matching before outputFeed data has diverse units, so unified formatting is needed to avoid unit confusion in due diligence reports
UPLOAD_FILE_MAX_SIZE500 MBSingle-batch test reports and ledger files typically do not exceed 200 MB. Reserve redundant space to support large multi-page reports
PARSE_FILE_TIMEOUT_SECONDS300 secondsParsing large multi-page feed test PDFs takes a long time. Avoid interrupting the parsing process due to timeout
recallTopNTop 6 entriesFeed due diligence requires associating multi-dimensional data including ingredient purchases, tests, and feeding records. Too few recalled entries will not cover all relevant items
similarityThreshold0.75–0.85A relatively high threshold is required for similarity between feed ingredient names and test items, to avoid matching data from unrelated categories with similar names

The parameter values provided on this page are common recommended starting points for configuration. Actual values are affected by material form, data volume, and business rules. Specific issues require case-by-case analysis, and testing on local samples is recommended before finalizing settings.

Common Configuration Mistakes

  • Symptom: After calling the dialogue interface and uploading a feed test report, the returned result does not include parsed file content. Cause: The UPLOAD_FILE_MAX_SIZE parameter is not configured correctly, or the required file field is not included in the request body.
  • Symptom: Unit confusion appears in the due diligence report returned by the dialogue, such as both grams per kilogram and simplified labeled indicator values being present. Cause: The unit unification rule is not clearly defined in systemPrompt, or the prompt does not cover unit verification for multi-batch data.
  • Symptom: A 401 status code is returned when calling the dialogue interface. Cause: The application's appId and apiKey are not correctly obtained and included in the request, or the secret key has expired and become invalid.

How to Verify Correct Configuration

  • Upload a single feed test report PDF, view the parsed field list, and confirm that it includes ingredient name, batch number, values of each indicator, and corresponding units.
  • Initiate two rounds of dialogue: first ask for the crude protein content of a specific batch, then ask for the moisture content of another batch, and confirm that the model can distinguish data from different batches without unit confusion.
  • Call the dialogue interface, include the correct appId and apiKey, upload a feed ledger Excel file, and check whether the returned result includes key data from the file.
  • Adjust the recallTopN parameter, test recalling different numbers of related data, and confirm that all dimensions required for due diligence are covered.

Question material comes from public community discussions. Configuration values are common starting points and should be measured against your own samples. Verified on 2026-09-14.