DGI ​
| Workflow | Type | Description |
|---|---|---|
| dgi.action.create-connections | Linear | Create connections (uses query-database) |
| dgi.action.query-connections | Linear | Query active connections (read-only; scoped to organization) |
| dgi.action.query-database | Linear | Execute a database query (requires organization_uuid) |
| dgi.action.show-dag | Linear | Extract and return DAG structure |
| dgi.action.think | Linear | Call the org's LLM for a one-shot reasoning turn (via ai.chat) |
| dgi.action.update-connections | Linear | Rename a connection within the caller's organization scope |
| dgi.active-turn.find | Atomic | Atomic adapter for dgi.active-turn.find; required by parent workflow specs: dgi.session.restore-active-turn. |
| dgi.ai_chat | Atomic | AI-powered chat for DGI conversations |
| dgi.composition.validate | Atomic | Atomic adapter for dgi.composition.validate; required by parent workflow specs: dgi.workflow.publish-version, dgi.workflow.validate-and-save. |
| dgi.intent_recognition | Atomic | AI-powered intent recognition |
| dgi.prompt_generation | Atomic | AI-powered prompt generation |
| dgi.reasoning-dag | Dag | Queue-based reasoning workflow with iterative synapse processing |
| dgi.reasoning-finalize | Linear | Finalize reasoning DAG - extract final answer and aggregate state |
| dgi.reasoning-handle-output-step | Linear | Handle output - route and handle synapse output |
| dgi.reasoning-init | Linear | Initialize reasoning DAG - create root synapse and initialize state |
| dgi.reasoning-process-synapse | Linear | Process a single synapse - prepare, call LLM, return output |
| dgi.reasoning-process-synapse-step | Linear | Process synapse - prepare, call LLM, return output |
| dgi.session.get | Atomic | Atomic adapter for dgi.session.get; required by parent workflow specs: dgi.session.restore-active-turn. |
| dgi.session.restore-view | Atomic | Atomic adapter for dgi.session.restore-view; required by parent workflow specs: dgi.session.restore-active-turn. |
| dgi.session.websocket-resubscribe | Atomic | Atomic adapter for dgi.session.websocket-resubscribe; required by parent workflow specs: dgi.session.restore-active-turn. |
| dgi.synapse.build_dag | Atomic | Construct layered DAG via topological sort |
| dgi.synapse.decompose | Atomic | Break goal into atomic operations with dependencies |
| dgi.synapse.map_capabilities | Atomic | Map operations to library_ids, identify missing params |
| dgi.synapse.understand | Atomic | Deep intent understanding from user message + history |
| dgi.synapse.validate | Atomic | Self-correct: review DAG for consistency |
| dgi.thought-processing | Dag | Processes user messages using synthetic-thought with skills (DAG) |
| dgi.thought-processing-analyze-intent | Linear | Analyze user intent - initial synapse processing |
| dgi.thought-processing-apply-skills | Linear | Apply skills - execute action workflows |
| dgi.thought-processing-finalize | Linear | Finalize thought processing - generate final answer and update lock |
| dgi.thought-processing-process-reasoning | Linear | Process reasoning - synapse processing |
| dgi.thought-processing-synthesize | Linear | Synthesize results — calls the org's LLM via ai.chat to produce the final answer |
| dgi.thought-processing-synthesize-runner | Linear | Internal runner for dgi.thought-processing-synthesize |
| dgi.workflow-request.create | Atomic | Capture a user's request for a workflow that does not exist yet |
| dgi.workflow.promote | Atomic | Promote a DGI-compiled plan to a durable, versioned composition via the existing composition.save primitive (member build-time; no end-user path). |
| dgi.workflow_builder | Atomic | AI-powered workflow building |
