What the data for this category looks like
Survey and design data primarily originates from project survey reports, professional design drawings, approved construction drawing documents, and phased change records. Data update rhythm follows project phase progression, with independent versions generated during the preliminary design, construction drawing, and completion stages respectively. Document structure is divided into modules by discipline, including architecture, structure, water supply and drainage, electrical, and others. Fields include project unique ID, discipline name, drawing size, version number, approval status, and more. Units mostly adopt general engineering standard units such as millimeters, megapascals, and square meters.
What constraints these characteristics impose on tool calling and plugins
The multi-disciplinary classification and long-document nature of survey and design data requires tool calling to support filtering retrieval scope by discipline, to avoid interference from irrelevant data. Frequent version updates require plugins to support scheduled pulling of the latest drawing and report versions, to ensure the timeliness of called data. The structured requirements of multiple fields require that exclusive parameters be clearly passed during tool calling; otherwise, parameter verification will fail. Additionally, individual documents contain substantial content, so the number of results returned by the tool and the context length must be limited to avoid exceeding the model's processing limits.
How to set configurations
| Configuration Item | Recommended Setting | Rationale |
|---|---|---|
functionCall | Specify survey and design-specific field sets | Survey and design data includes exclusive fields such as project ID, discipline type, drawing version number, so the calling parameter scope must be clearly defined |
toolChoice | auto | Adapt to multi-disciplinary tool linkage needs in survey and design scenarios, automatically match corresponding processing tools |
| Tool call timeout duration | 120–180 seconds | Parsing survey and design reports and drawings takes a long time, to avoid interrupting the calling process due to timeout |
| Tool return result limit | Top 8 entries | Individual survey and design documents contain substantial content, limiting the number of returned entries to avoid context overflow |
| Context window length | 8000–12000 characters | Match the parsing needs of long survey and design documents, avoid truncating critical content |
| Knowledge base reranking rule | Sort by discipline priority | Survey and design data is classified by disciplines such as architecture and structure, reranking can improve retrieval matching accuracy |
Three Common Errors
- An
invalid function parameterserror is returned during calling: Cause: Parameters are not passed in the survey and design-specific field format, such as omitting the drawing version number or discipline type field. - No engineering tool options appear in the tool call panel: Cause: The professional tool extension switch for FastGPT v4.17.0 is not enabled; survey and design scenarios require enabling engineering parsing plugins.
- A high memory usage alert appears during tool calling: Cause: The number of returned results and context length are not limited, resulting in loading excessive survey and design document data.
How to Confirm the Configuration is Properly Set
- Initiate a simulated call, check if the request parameters include survey and design-specific fields such as project ID and discipline name.
- View the tool call log, confirm that the
toolChoiceparameter is correctly set toautoor the specified discipline tool. - Trigger the tool call process, verify that the response time falls within the 120–180 second range.
- Retrieve test data, confirm that the returned results are sorted by discipline priority and the number of entries matches the preset limit.
- Verify that the large language model can correctly call the specified tool, confirm that the tool call process has no abnormal interruptions.
The values provided here are common starting points and should be evaluated against the user’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.