What the documents look like
For product questions in MCN operations, this page examines FastGPT vector indexes. Compare the material received with the sources used in the answer. The examples are hypothetical.
Constraints at the vector indexes step
An MCN product-index test can start with new proposed contract text: “Only listed skincare products are demonstrated in this stream; gifts follow the current schedule.” Proposed search entries may address demonstration scope and the source for gift information, but the original names no gift. For each index, identify what the original supports and which source the second question still needs. Several indexes pointing to one original do not establish that it contains every product fact requested. Inspect the statement supported by each search entry, not only whether both reach the same contract. Shared retrieval of an original does not establish that a gift is named there; the schedule reference still needs actual schedule content.
Configuration and implementation checks
| Item or implementation check | Original or index content | Traceable evidence |
|---|---|---|
| Associated models | Contract and schedule bases associated with the same retrieval node | Check the same embedding model across bases associated with that node. |
| Original mapping | Contract states scope and a schedule reference | Trace each index to its actual original. |
| Index content | Demonstration scope and gift-evidence entries | Do not rewrite demonstration as giveaway. |
| Repeated questions | Separate demonstrated-item and gift questions | Identify which question the available source supports. |
| Image ingestion | Missing schedule list supplied as an image | Establish the actual processing path, then inspect its output and citation. |
Documented conditions
| Documented item / implementation check | Condition |
|---|---|
Embedding model | Knowledge bases associated with one search node must use the same embedding model. |
Original data and indexes | One original data block can have multiple indexes. |
Index content | Index wording should be complete and focused. |
Image indexes | Where image retrieval is used, inspect descriptions or image vectors. |
Version and result check | After a change, reuse questions to compare originals. |
Three common errors
The first error labels a proposed entry skincare giveaways for this stream while its original covers demonstration only. Compare index and contract wording for an unsupported giveaway claim. The second keeps the gift entry targeting that older demonstration passage after the schedule list is obtained. Inspect the new original, entry target and actual citation; file acquisition does not establish correct mapping. The third assumes an image-origin list is already searchable. Establish the actual processing path: compare extracted wording for text processing, or inspect descriptions or image vectors only for actual image retrieval, then test the gift citation. Source appearance cannot establish an effective processing chain.
How to verify the setup
Without the schedule list, gift information lacks an original source. After it is supplied, inspect correspondence between the added original and its indexes, then compare citation changes for the same question; counting more indexes is insufficient. For an image-origin list, establish the processing path actually used. A text-parsing path requires comparison of extracted text with the list; descriptions or image vectors need checking only when image retrieval is actually used. Then trace the question to its supporting passage. Verify coverage separately from authorization, without treating product relevance as permission to promote. If the gift question still cites only the old contract after a list is added, inspect the added original, index target and actual candidates before assuming the update took effect. Compare the new list wording with the response to distinguish an unread source, incorrect mapping and omitted evidence.
| Material context | Scope |
|---|---|
| Document types | Influencer Contract, Training Records, Violation Handling Form and Account Scheduling |
| Candidate sources | MCN Backend, Platform Violation Notice, Influencer Account Data and Contract Ledger |
| Technical baseline | FastGPT v4.17.2 |