What data in this category looks like
Oil and gas exploration and development data comes from field seismic acquisition data, drilling operation logs, trial production condition records, and reserve assessment documents. Update frequency: Seismic exploration data is updated in batches corresponding to acquisition operations. Drilling logs are updated per single well operation cycle. Trial production data synchronizes operating conditions daily. Most documents use structured formats. Core metadata includes well ID, operation time, and operation type. Fields include well depth, permeability, porosity, and daily gas production. Their units are meters, millidarcys, percent, and cubic meters per day, respectively.
What constraints these characteristics impose on tool calling and plugins
Seismic exploration data has a large volume, with a single batch reaching the GB level. Limit the amount of data loaded in a single tool call to avoid excessive resource usage. Drilling logs are updated periodically. Tools must support incremental synchronization or resumable uploads to ensure data timeliness. Trial production data and reserve assessment documents are mostly long texts. Split long text blocks during tool calls to avoid context overflow. Professional fields have diverse units. Configure a unified unit conversion logic in plugins to avoid parsing errors. Many professional terms exist in exploration data. Match professional term mapping rules during tool calls to ensure accurate data parsing.
How to set configurations
Reference only; verify against the target FastGPT release and deployment documentation.
| Configuration item | Recommended setting | Rationale |
|---|---|---|
toolChoice | auto | Adapts to scenarios with many oil and gas exploration professional terms, allowing the model to automatically select matching tools |
| Workflow call timeout | 600 seconds | Oil and gas exploration data processing takes a long time, prevents tool calls from being interrupted due to timeout |
| Long text split length | 800–1200 characters | Adapts to the characteristic that exploration documents are mostly long texts, avoids context overflow |
| Plugin parameter mapping rules | Calibrated based on actual measurements | Matches professional fields such as well ID and operation type to ensure accurate data parsing |
| Plugin cache validity period | 24 hours | Adapts to the daily update rhythm of drilling logs, avoids using expired data |
| Number of items returned per tool call | First 3 items | Controls the number of exploration data results returned in a single call, avoids excessive results causing context overflow |
Three common errors
- Phenomenon: After calling a workflow, running data in the conversation log is empty. Cause: No field mapping rules between plugins and exploration data are configured, so valid data cannot be extracted.
- Phenomenon: After a custom plugin runs, the output download address flickers continuously before a result is provided. Cause: The plugin triggered multiple incremental pull requests for drilling logs, and the number of pull batches per request was not limited.
- Phenomenon: Nested workflows cannot execute completely, only some tool call links are completed. Cause: The number of nested workflow call layers exceeds the configured threshold, and no parameters for context transfer are configured.
How to confirm the configuration is properly set
- Manually trigger a tool call, check if running data fields in the conversation log include professional exploration fields such as well ID and operation time.
- Run the custom plugin, verify that the download address of the output result is stable and does not flicker continuously.
- Test the nested workflow, confirm that all tool call links in the sub-workflow have completed execution.
- Check that within the plugin cache validity period, calling the tool again can obtain the latest trial production data.
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.