What the documents look like
For product questions in Franchise fee settlement, this page examines FastGPT retrieval. A service line and equipment references may be retrieved while the equipment subjects still require the actual supply list. The examples are hypothetical.
Constraints at the retrieval step
A product-selection question inside franchise settlement material first needs to distinguish a service line from equipment detail. In this proposed invoice, the entry says franchise service item and refers equipment to a supply list. A payment-plan note says equipment matters are listed separately. These are two documentary references, with no equipment entries themselves. Searching for equipment names may still return the invoice and note. Identify their actual contribution: the project identity and a pointer to another source. Repeated references to the same missing list do not establish a configuration or add an equipment item to the supplied wording.
Configuration and implementation checks
| Retrieval subject | Specific wording relationship | Reading outcome |
|---|---|---|
| Service line | Invoice identifies franchise service item | Establish what this item refers to. |
| Document pointer | Equipment refers to a supply list | Locate required text without inserting a machine name. |
| Delegated scope | Payment note separates equipment matters | Two references do not supply configuration entries. |
| Rewriting and trimming | Machine follow-up and see-list condition | Locate added entities or lost qualifications. |
| Supply-list retest | Applicability to the same service project | Trace equipment subjects, configuration and exclusions. |
Documented conditions
| Documented item / implementation check | Condition |
|---|---|
Associated knowledge bases | Knowledge bases associated with one search node must use the same embedding model. |
Search mode | Semantic, full-text and hybrid modes are available. |
Citation limit | Citation limits are measured in tokens. |
Minimum relevance | Range 0–1; active for semantic search or reranking. |
Question optimization | Resolves references and expands questions. |
Three common errors
For a follow-up asking which machine this item includes, this item initially refers to the known service line. Question optimization that inserts a machine name changes that subject. Retrieval of similar equipment vocabulary would still need an applicable supply source. A separate failure occurs when passage trimming retains equipment but drops see the supply list and listed separately, allowing the answer to describe equipment as already included. Compare the original question, optimized wording, complete candidate and actual citation. This locates whether an unsupported entity entered during rewriting, retrieval interpretation or answer composition, without assuming that every matching noun describes a supplied machine.
How to verify the setup
Acceptance should inspect three distinct reading results: the kind of item the invoice establishes, the material to which the payment note delegates equipment details and whether the available text names any equipment with its configuration. The first two can trace to their respective statements; the third still needs the list text. If the actual supply list is later obtained, establish its relationship to the same service project before reading equipment subjects, configuration and exclusions, then trace its presence in candidates and citations. The earlier invoice remains useful for project identity. Adjoining old and new passages cannot establish that payment or equipment delivery actually occurred.
| Material context | Scope |
|---|---|
| Document types | Franchise Fee Invoice, Deposit Ledger, Deduction Notice and Contract Payment Plan |
| Candidate sources | Financial System, Franchise Contract, Bank Statement and Store Violation Records |
| Technical baseline | FastGPT v4.17.2 |