What the data for this category looks like
This category covers documents including bidding announcements, qualification pre-review documents, bidding documents, tender documents, bid evaluation reports, successful bid candidate announcements, successful bid result announcements, and similar materials. Data sources include the national public resource trading platform, local public resource trading platforms, engineering approval systems, and bidding and tendering supervision systems. Updates occur daily. Documents contain structured fields such as section division, qualification requirements, performance lists, and bid evaluation rules. These fields must match the specifications for the corresponding business scenario, with no unified fixed unit.
Constraints on citation source and traceability
Multi-source, heterogeneous data sources require the traceability link to support document formats and metadata structures from different platforms. Daily updates require configurations to support real-time synchronization of source document changes. Structured fields such as section division and qualification requirements must accurately match the business meaning of cited fragments to avoid traceability misalignment. The strict nature of bidding and tendering processes requires traceability to cover all business links, ensure cited content matches source documents exactly, and prevent compliance risks from missing or incorrect citations.
Configuration settings
Under FastGPT v4.17.0, the core configuration items and recommended settings for citation traceability in this category are as follows:
| Configuration Item | Recommended Setting | Rationale |
|---|---|---|
quoteMaxToken | 800–1200 token | Bidding document fragments often contain long texts such as qualifications and performance records, and need to cover complete citation units |
responseData | Return only hit fragment metadata | Need to accurately locate specific paragraphs of source documents such as bidding announcements and bid evaluation reports, and reduce redundant content |
detail | Enabled | Need to display necessary traceability information such as the source document's publishing organization and update time |
| Single recall count | Top 6 entries | Bidding document volumes are large, avoiding context overflow that affects generation quality |
| Source document synchronization cycle | Daily | Match the daily update frequency of data in this category, ensuring cited content is the latest version |
Three common misconfigurations
- Enabling the rerank function leads to empty citation fragments. This manifests as no content in the citation fields of search results. The cause is that the rerank process does not retain metadata association with originally recalled fragments, leading to loss of traceability information.
- Exporting a workflow and importing it into another environment causes plugins associated with citation configurations to become unavailable. This manifests as workflow node errors showing missing configurations. The cause is that associated configuration files for the plugin were not synchronized during cross-environment import.
- The message "No permission to operate this conversation record" appears at the end of responses. This manifests as a permission denied prompt returned in citation fields. The cause is that access permissions for the source document were not enabled in the knowledge base configuration, making traceability information unreadable.
How to confirm correct configuration
- Trigger a bidding document drafting request, and check if the citation fields in returned results include the specific name and paragraph information of the source document.
- Adjust the
quoteMaxTokenparameter value, and verify that long text fragment citations are fully included without truncation. - Disable the rerank function and initiate another test, confirming that citation fragments are not lost or misaligned.
- Export the current workflow and import it into a test environment, checking that all citation configuration items are fully retained without missing content.
The values provided are common starting points and should be evaluated against your 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 2026-09-28.