Use IngaDB from an AI agent
AI agents can run causal analyses through IngaDB's OpenAPI. The agent interprets natural language and selects an operation; IngaDB performs the probability and What-if computations. Results retain the data and revision used for the calculation.
How an agent runs an analysis
The agent selects the operation and inputs. IngaDB computes the same result from the same inputs.
If we halve the seal-leak probability, how does line-stop risk change?
- 01DISCOVEROpenAPI
workspace_pipeline_what_if_post - 02CHECKinput and permission checkeddry-run ✓
- 03COMPUTEdeterministic computationrev 42
- 1AI AGENT
Interpret the request
Read the natural-language goal and select an operation and its inputs.
- 2OPENAPI / INGACTL
Validate the request
Check the input schema and required permission, then preview the request with dry-run.
- 3INGADB ENGINE
Compute with IngaDB
Compute from the causal model and evidence, and record the revision with the result.
Available analysis operations
Agents and ingactl use the same OpenAPI contract. Input schemas and required permissions come from the connected instance's catalog.
Trace causes and evidence
Retrieve search results, causal paths, evidence, importance, and cut sets.
workspace_data_causal_path_getCompare changed conditions
Run deterministic What-if analysis without changing stored data.
workspace_pipeline_what_if_postCheck whether a result is still valid
Read the calculation revision and freshness, along with changes that affected the result.
workspace_data_analysis_getConnect with ingactl
Generate SKILL.md or AGENTS.md for an agent from the connected instance's catalog. The guide includes available routes, input examples, and required permissions.
$ ingactl sync
$ ingactl skill --format agents-md --out .
wrote ./AGENTS.md
$ ingactl api show workspace_data_causal_path_get
permission analysis.read ✓
$ ingactl api call workspace_data_causal_path_get \
--project packaging-line --dry-run