Workflows
Workflows
Akurion workflows turn repeated knowledge tasks into reusable processes. A workflow can define inputs, retrieval steps, AI instructions, tool usage, sharing rules, and run history.
Workflow Use Cases
Good workflow candidates include:
- Compliance evidence review.
- Vendor risk summary.
- Incident postmortem draft.
- Customer account brief.
- Product release summary.
- Contract obligation extraction.
- Onboarding knowledge pack.
- Weekly support trend report.
Workflow Lifecycle
| Stage | Description |
|---|---|
| Draft | Build or edit workflow definition. |
| Publish | Make the workflow available to users or teams. |
| Assign | Grant access to users or projects. |
| Run | Execute the workflow with inputs. |
| Review | Inspect output, citations, trace, credits, and errors. |
| Retry | Rerun failed or incomplete runs. |
| Share | Share workflow or run result where allowed. |
Control-Plane Operations
Akurion app operations include:
createWorkflowupdateWorkflowdeleteWorkflowpublishWorkflowduplicateWorkflowsetWorkflowShareassignWorkflowUsersrunWorkflowretryWorkflowRundeleteWorkflowRuns
Workflow queries include:
getWorkflowsgetWorkflowDetailsgetWorkflowRunsgetWorkflowRunDetailsgetWorkflowAssignableUsersgetWorkflowToolCatalogresolveWorkflowByShareToken
Design Recommendations
For reliable workflows:
- Keep the goal specific.
- Define expected inputs clearly.
- Use metadata filters where possible.
- Require citations for factual claims.
- Estimate credit usage before high-volume runs.
- Use project instructions to constrain answer style.
- Use approval gates for sensitive or high-cost workflows.
Developer Integration
Developers can recreate workflow-style behavior using MCP or REST tools:
- Use
discover_metadata_schemato understand the corpus. - Use
generate_multi_queriesto create retrieval variants. - Use
vector_search_with_metadata_filterto get evidence. - Use
collect_chunks_from_multiple_filesfor full-document context. - Use
answer_queryfor final cited synthesis.