What the Data for This Category Looks Like
Data sources for pre-sales technical proposal materials include customer interview records, tender documents, vendor product materials, and historical project proposal documents. The data update cycle is per order or weekly, with updates aligned with project progress or industry adjustments. Document types cover technical proposal books, customer requirement survey forms, quotation lists, and demo scripts. Fields include technical routes, implementation boundaries, quotation items, solutions corresponding to customer pain points, and more. The length of individual documents varies widely, ranging from requirement lists of a few hundred characters to complete proposal books of tens of thousands of characters.
What Constraints These Characteristics Impose on Knowledge Base Retrieval and Recall
Pre-sales technical proposal data comes from scattered sources with significant format differences. Term density and field structures vary across different document types. This requires retrieval logic to adapt to multi-format content. Data updates occur at a high frequency, so support for fast retraining of abnormal documents is needed to avoid outdated content affecting recall results. Individual proposals are tied to exclusive project data, so retrieval scope must isolate knowledge base content across different projects to prevent cross-project data interference. Document length varies widely, so the number of recall results and context length must be controlled to avoid exceeding model processing limits.
How to Configure the Settings
This document is written based on FastGPT v4.17.0. The following configuration items are tailored for the pre-sales technical proposal scenario:
| Configuration Item | Recommended Value | Rationale |
|---|---|---|
similarity | 0.72–0.80 | Adapts to the technical term density of pre-sales proposal documents, filters irrelevant historical proposal content with low matching scores |
limit | Top 6 entries | Matches the volume of documents for a single pre-sales proposal, prevents excessive recall results from exceeding model context processing limits |
searchMode | Full-text search | Adapts to long paragraph content in technical proposal books, covers core fields such as technical routes and implementation boundaries |
usingReRank | Enabled | Performs secondary ranking on initial recall results to improve the priority of high-matching documents |
datasetSearchUsingExtensionQuery | Enabled | Performs term expansion on pre-sales-related questions to supplement professional expressions from vendor product materials |
collectionIds | Bind the knowledge base collection ID for the current project | Isolates proposal data across different projects to prevent cross-project content from interfering with current project retrieval results |
Three Common Misconfigurations
- Symptom: Some questions fail to recall knowledge base content during API calls, but normal results are returned on the debug page. Cause: The API request does not carry the
collectionIdsparameter. By default, it retrieves the full knowledge base without limiting to the exclusive collection for the current pre-sales project, leading to cross-project data interfering with recall results. - Symptom: After abnormal training of knowledge base documents, one-click batch retraining of abnormal documents is not available. Cause: The retraining task is not executed on the targeted abnormal document collection for pre-sales proposals. Only single-document or full knowledge base retraining is supported, with no fast processing for abnormal content of the current project.
- Symptom: The knowledge base export task only returns 50,000 records, and cannot fully export all pre-sales proposal documents. Cause: Full export configuration is not enabled. The default setting enables a single-batch export quantity limit, and the mode is not adjusted to full export.
How to Confirm Configurations Are Correct
- Initiate an API call with the specified
collectionIdsparameter, and verify that the returned knowledge base recall results match those from the debug page. - Mark 3 to 5 abnormal documents of pre-sales proposals, execute a batch retraining task, and verify that the training status of the abnormal documents updates to completed.
- Execute a knowledge base export task, confirm that the export scope includes all pre-sales proposal documents with no preset quantity limit.
- Check the system storage interface, confirm that the uploaded pre-sales proposal documents are associated with the specified knowledge base collection.
The values provided here are common starting points, and should be measured against the reader's own samples.
Question material comes from public community discussions. Configuration values are common starting points and should be measured against your own samples. Verified on 2026-10-19.