Production Ledger Summary and Report Generation for Incoming

Page Overview

Client industry: Incoming material inspection and supplier quality management
Business direction: Production ledger summary and report generation
Capability area: Knowledge base retrieval and recall

What data looks like for this use case

Data sources for this use case include purchase orders, drawing specifications, supplier quality certificates, inspection equipment records, historical quality data, and supplier portal documents. Daily generated documents cover inspection specifications, sampling plans, incoming material inspection reports, supplier audit reports, exception contact sheets, 8D correction reports, and concession acceptance approval forms. Data updates occur synchronously with each batch of incoming material to generate inspection records. Supplier overall evaluations are completed monthly or quarterly. Special audits are conducted on an irregular basis. Document fields include incoming material batch number, sampling quantity, inspection items, non-conformity judgment results, supplier qualification number, concession acceptance order number, and others. Units include pieces, batches, days, and others.

What constraints these characteristics impose on knowledge base retrieval and recall

Multi-source and multi-type documents require retrieval to cover associated content across datasets, to avoid missing matches between supplier quality certificates and historical inspection records. Frequently updated incoming material inspection data requires retrieval to support incremental synchronization, ensuring that the latest batch inspection specifications and reports can be quickly recalled. Multi-dimensional fields such as batch and supplier require retrieval to accurately match business dimensions, avoiding generalized recall of irrelevant content. Mixed input of different document structures requires retrieval to adapt to both structured content in inspection specifications and unstructured content in correction reports, ensuring the relevance of recalled results.

How to configure settings

Configuration Item	Recommended Setting	Rationale
`similarity`	0.75–0.85	Matches the precise document needs of incoming material inspection, avoids recalling irrelevant historical quality records
`limit`	Top 10 entries	Adapts to the fast retrieval scenario for each batch of incoming material, controls context length
`searchMode`	`mixedRecall`	Balances keyword matching and semantic recall, covers multiple document types such as inspection specifications and historical reports
`usingReRank`	Enabled	Filters low-relevance results, improves recall accuracy for non-conformity items and supplier audit reports
`datasetSearchUsingExtensionQuery`	Enabled	Supplements associated retrieval for fields such as batch and supplier name, meets the multi-dimensional needs of ledger summary
`collectionIds`	Incoming material inspection exclusive dataset ID	Isolates supplier quality related documents, avoids recalling irrelevant production process data

These configurations are based on the interface and API logic of FastGPT v4.17.0.

Three common configuration errors

Phenomenon: Semantic retrieval score displays a value over 4000, exceeding the standard 0-1 range. Cause: The normalization logic for the similarity parameter is not correctly configured, leading to return of raw embedding model scores without normalization processing.
Phenomenon: Some incoming material inspection specifications and supplier audit reports cannot be retrieved. Cause: The collectionIds parameter does not include the exclusive dataset for the corresponding documents, or searchMode is only configured for semantic retrieval and does not cover keyword matching scenarios.
Phenomenon: Source file information used for current retrieval cannot be obtained via API calls. Cause: The knowledge base file provenance configuration is not enabled, or the parameter for returning file metadata is not specified in the retrieval request.

How to confirm correct configuration

A retrieval targeting a specific incoming material batch number is initiated. The relevance of returned results is checked against business scenario requirements, and the similarity parameter is adjusted to the range suitable for the business.
Retrieval logs are obtained via API. Returned results are confirmed to only include documents under the specified collectionIds, with no cross-category irrelevant data.
After usingReRank is enabled, the reordered results are checked to confirm that core business documents such as non-conformity items and supplier audits are prioritized.
The status of the extended query switch in the retrieval request is checked. Fields such as batch and supplier name are confirmed to be automatically included in the retrieval scope.

The values provided here are common starting points, and should be evaluated against the samples of the actual deployment.