Skip to content
For the complete documentation index, see llms.txt. This page is also available as Markdown.

WorkflowContext reference

WorkflowContext (ctx) is passed to the workflow entry function and each agent's run handler. Use it for input, shared state, calling other agents, and SDK methods (AI, search, documents).

Properties

Member
Description

ctx.input

Workflow input from runWorkflow, or overridden when you call runAgent with input

ctx.state

Shared mutable object for this run β€” use it to pass data between agents

ctx.agentName

Name of the agent currently running (inside an agent run handler)

ctx.senderId

Sender id used for SDK calls inside the workflow

ctx.workflow

Parent workflow definition

ctx.notifier

Underlying notifier instance (advanced use)

Methods

Method
Description

ctx.runAgent(name, { input? })

Run one registered agent. Optional input overrides ctx.input for that agent.

ctx.runAgents(spec)

Run multiple agents in parallel. Returns { results }.

ctx.waitForMessage({ timeoutMs? }) / wait_for_message

Wait for the next inbound realtime message (WebSocket).

ctx.sendAI({ message, saveHistory?, sessionId?, provider?, model?, label? })

Call the AI API. Optional provider / model override project defaults (or agent defaults). Optional label names the step in the dashboard.

ctx.fetchAIHistory({ label? })

Fetch AI history for senderId.

ctx.search({ query, limit?, minSimilarity?, filterbySource?, label?, vectorStore?, embedding?, serverConfig? })

Semantic search over the knowledge base. Store and embedding follow vector stores.

ctx.addDocument(opts) / updateDocument / deleteDocument / listDocuments / getDocument

Knowledge-base document helpers. The same vectorStore, embedding, and serverConfig fields apply.

ctx.recordStep({ label, input?, output?, status? })

Record a custom step when observability is enabled.

Python names use snake_case: run_agent, run_agents, send_ai, fetch_ai_history, etc.

Call one agent

Call agents in parallel

More patterns: Parallel agents & shared state.

Shared state

Observability

When observability: true on the workflow, calls like sendAI, search, runAgent, and recordStep show up as steps in the dashboard. You do not need extra setup beyond enabling the flag and connecting the client.

Errors

Language notes

Language
Import

JS / TS

@dnotifier-realtime/dnotifier β€” runAgent, runAgents

Dart

package:dnotifier/dnotifier.dart β€” same camelCase API

Python

from dnotifier import Workflow β€” run_agent, run_agents

Next steps

Last updated