What the data for this category looks like
Reference only; verify against the target FastGPT release and deployment documentation.
What constraints these characteristics impose on workflow orchestration
Workflows must be configured with multiple document access nodes that support batch import of construction documents in different formats, to accommodate multi-source and heterogeneous ledger data sources. Data updates have no fixed cycle, so workflows must be bound to construction node events as trigger conditions to meet non-fixed cycle update requirements. Wide variation in individual document length requires text parsing that adapts to different lengths, to prevent loss of key information due to long document truncation. Numeric content with multiple unit types is included, so unit recognition rules must be configured during the variable extraction stage to ensure correct matching of values and units during ledger summary.
Configuration Settings
| Configuration Item | Recommended Setting | Rationale |
|---|---|---|
| Single search limit | Top 30 entries | Infrastructure construction ledgers have a large number of documents accessed per batch, and excessive single recall will cause context overflow |
| Maximum context length | 8000–12000 characters | Parsed text from long individual special progress reports is lengthy, so long context processing must be supported |
variables | Extract three core fields: project number, material quantity, acceptance date | Ledger summary must focus on core business fields to avoid interference from redundant information |
nodeId | Bind the ID of the construction node trigger node | The workflow must trigger upon completion of construction nodes to adapt to non-fixed cycle update requirements |
| Question optimization | Enabled | Queries for ledger summary must clearly point to field extraction and value statistics; optimization improves recall accuracy |
| Reranking | Adjust based on actual testing | Document structures vary greatly across different projects, so adjustments must be made based on actual recall results |
Three Common Mistakes
- The single reference limit for knowledge base search in the workflow is set to 3000, which does not match the 13000 limit configured for standalone applications. The cause is that the search context quota rules for workflow nodes are separate from those for standalone knowledge base applications, and no separate adjustments are made for the long-document ledger scenario.
- After orchestrating the workflow and calling it via API, the returned ledger data fields are empty. The cause is that the core fields to be extracted are not declared in the
variablesconfiguration, so the workflow does not return corresponding data. - Workflow execution times out, with the interface prompting that the execution duration exceeds 600 seconds. The cause is that the upper limit for the number of documents accessed in a single batch is not configured, and simultaneous parsing of multiple long progress reports exceeds the system's default timeout threshold.
How to Confirm the Configuration Is Correct
- Enter the workflow editing page, check the
variablesconfiguration item, and confirm that the core ledger fields to be extracted have been declared. - Trigger a simulated construction node event, view the workflow execution log, and confirm that all nodes execute in the expected order with no errors.
- Call the test interface, verify that the returned workflow results include the preset ledger summary fields and correct units.
- Adjust the single search limit parameter, compare recall results across different values, and confirm that they meet the recall requirements of the business scenario.
The values provided here are common starting points, and should be evaluated 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 2026-09-28.