Reference only; verify against the target FastGPT release and deployment documentation.
Data Characteristics for This Category
Oil and gas exploration and development business data comes from drilling logs, 3D seismic exploration reports, trial production analysis documents, oilfield development plans, and similar sources. Different data types follow distinct update schedules: Teams update daily operational drilling logs each day. Periodic exploration reports receive updates during their respective exploration cycles. Long-term development plans undergo quarterly revisions. Most documents include standardized fields such as well ID, depth, porosity, and permeability. Some professional documents include charts and sampling tables. Supported units include meters, millidarcy (mD), percentage, and other industry standard measurements. A single large-scale exploration report can contain tens of thousands of characters of text.
Constraints for Knowledge Base Retrieval and Recall
Professional fields and units require retrieval systems to match parameter and unit associations. This prevents false recall of irrelevant parameters. Long text documents limit context coverage for single-block vector embeddings. Adjust retrieval context matching logic to address this. Business data isolated across multiple blocks requires retrieval to bind datasets to oilfield blocks. This stops cross-block irrelevant content from being recalled. Documents with varying update frequencies require vector database sync cycles to align with data update rhythms. This avoids retrieving expired exploration results.
Configuration Recommendations
| Configuration Item | Recommended Range | Rationale |
|---|---|---|
similarity | 0.72–0.85 | Oil and gas exploration documents have high professionality. This range filters low-relevance results while balancing recall precision and coverage. |
limit | Top 8–12 results | Single exploration documents have large content volumes. Excessive results exceed context window limits. This range balances recall coverage and context length. |
searchMode | Semantic retrieval + keyword hybrid | Oil and gas documents have both professional semantic links and clear keywords such as well IDs and parameters. Hybrid mode improves recall accuracy. |
usingReRank | Enabled | Long, professional documents may return semantically similar but non-matching results. Reranking optimizes result sorting precision. |
datasetSearchUsingExtensionQuery | Enabled | Exploration queries often use abbreviated terms, such as "poro-perm" for porosity and permeability. Extended queries cover more term variants. |
collectionIds | Configured by oilfield block | Oil and gas exploration data is isolated by block. Binding collectionIds for the target block prevents irrelevant cross-block recall. |
Three Common Configuration Errors
- Issue: Returned results lack document source information, or source fields are empty. Cause: Knowledge base document source configuration is not enabled, or retrieval does not activate document metadata association logic.
- Issue: Retrieved results do not include images from the original document. Cause: The default vector library only generates embeddings for text. No image-to-text or image-associated retrieval logic is configured, and no image reference processing is added during generation.
- Issue: Calling the
/api/v1/chat/completionsinterface with a specifiedappIdfails to associate uploaded knowledge base content. Cause: Target datasetcollectionIdsare not bound to the correspondingappId, or dataset-related configurations are not correctly passed in interface request parameters.
How to Verify Correct Configuration
- Submit a query with professional terms, such as "Porosity parameters for Well XX". Check if the number of returned recall results matches the range set in the
limitconfiguration. - Review returned result metadata fields. Confirm they include document names, well IDs, and other source information.
- Check interface request parameters. Confirm
collectionIdsbinds to the target oil and gas exploration dataset, andsimilarityvalues fall within the preset range. - Submit a query with abbreviated terms, such as "Poro-perm data for Block XX". Confirm extended queries cover relevant variant terms, and recalled results match business requirements.
The values provided here are common starting points, and should be evaluated against the specific samples of the target deployment environment.
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.