adapter:openai-node

`adapter:openai-node`

Field	Value
Prefix	`adapter:openai-node`
Namespace constant	`OpenAINodeConnectorNamespace`
Subjects constant	`OpenAINodeConnectorSubjects`
Kind	adapter
Schema record	`<inline>`
Tier	framework
Package	`@makaio/adapter-openai-node`
Defined in	`adapters/implementations/openai-node/src/namespaces/index.ts`

Subjects

Key	Wire	Type	Schema
`agent_complete`	`adapter:openai-node.agent_complete`	event	`lifecycle.ts`
`agent_started`	`adapter:openai-node.agent_started`	event	`lifecycle.ts`
`chunk`	`adapter:openai-node.chunk`	event	`chunk.ts`
`error`	`adapter:openai-node.error`	event	`lifecycle.ts`
`message_complete`	`adapter:openai-node.message_complete`	event	`message.ts`
`reasoning_complete`	`adapter:openai-node.reasoning_complete`	event	`reasoning.ts`
`reasoning_delta`	`adapter:openai-node.reasoning_delta`	event	`reasoning.ts`
`sdk.event`	`adapter:openai-node.sdk.event`	event	—
`sdk.raw`	`adapter:openai-node.sdk.raw`	event	—
`tool_approval`	`adapter:openai-node.tool_approval`	rpc	`tool-approval.ts`
`tool_calls`	`adapter:openai-node.tool_calls`	event	`tool-calls.ts`
`tool_completed`	`adapter:openai-node.tool_completed`	event	`tool-lifecycle.ts`
`tool_started`	`adapter:openai-node.tool_started`	event	`tool-lifecycle.ts`
`turn.state_changed`	`adapter:openai-node.turn.state_changed`	event	`turn-state.ts`
`turn.step_finished`	`adapter:openai-node.turn.step_finished`	event	`turn-state.ts`
`turn.step_started`	`adapter:openai-node.turn.step_started`	event	`turn-state.ts`
`turn.turn_finished`	`adapter:openai-node.turn.turn_finished`	event	`turn-state.ts`
`turn.turn_started`	`adapter:openai-node.turn.turn_started`	event	`turn-state.ts`
`usage`	`adapter:openai-node.usage`	event	`usage.ts`

Subject Details

`adapter:openai-node.agent_complete` (event)

Schema for agent complete event. Emitted when agent finishes processing (success or error).

Subject: adapter:openai-node.agent_complete Type: Event

Field	Type	Required
`error`	`string \| undefined`	no
`eventType`	`"agent_complete"`	yes
`message`	`string \| undefined`	no

`adapter:openai-node.agent_started` (event)

Schema for agent started event. Emitted when agent begins processing a user message.

Subject: adapter:openai-node.agent_started Type: Event

Field	Type	Required
`eventType`	`"agent_started"`	yes
`model`	`string \| undefined`	no

`adapter:openai-node.chunk` (event)

Schema for streaming chunk event. Based on OpenAI ChatCompletionChunk structure.

Subject: adapter:openai-node.chunk Type: Event

Field	Type	Required
`choices`	`{ delta: { content?: string \| null \| undefined; role?: "user" \| "assistant" \| "system" \| "developer" \| "tool" \| undefined; refusal?: string \| null \| undefined; tool_calls?: { index: number; id?: string \| undefined; type?: "function" \| undefined; function?: { name?: string \| undefined; arguments?: string \| undefined; } \| undefined; }[] \| undefined; }; index: number; finish_reason: "length" \| "stop" \| "tool_calls" \| "content_filter" \| "function_call" \| null; logprobs?: unknown; }[]`	yes
`created`	`number`	yes
`eventType`	`"chunk"`	yes
`id`	`string`	yes
`model`	`string`	yes
`object`	`"chat.completion.chunk"`	yes
`service_tier`	`"default" \| "auto" \| "priority" \| "flex" \| "scale" \| null \| undefined`	no
`system_fingerprint`	`string \| undefined`	no

`adapter:openai-node.error` (event)

Schema for error event. Emitted when an error occurs during processing.

Subject: adapter:openai-node.error Type: Event

Field	Type	Required
`code`	`string \| number \| undefined`	no
`eventType`	`"error"`	yes
`message`	`string`	yes
`type`	`string \| undefined`	no

`adapter:openai-node.message_complete` (event)

Schema for message complete event. Emitted when a full assistant message has been assembled from streaming chunks. Extends the base schema with OpenAI-specific finish_reason values.

Subject: adapter:openai-node.message_complete Type: Event

Field	Type	Required
`content`	`string \| null`	yes
`eventType`	`"message_complete"`	yes
`finish_reason`	`"length" \| "stop" \| "tool_calls" \| "content_filter" \| "function_call" \| null`	yes
`reasoning`	`string \| undefined`	no
`tool_calls`	`{ id: string; type: "function"; function: { name: string; arguments: string; }; }[] \| undefined`	no

`adapter:openai-node.reasoning_complete` (event)

Base schema for a reasoning complete event.

Emitted when the full reasoning / thinking content has been assembled from all streaming deltas. Does NOT include the Anthropic-specific signature field — the Anthropic adapter extends this schema with .extend().

Adapter-specific schemas may extend this via .extend().

Subject: adapter:openai-node.reasoning_complete Type: Event

Field	Type	Required
`content`	`string`	yes
`eventType`	`"reasoning_complete"`	yes

`adapter:openai-node.reasoning_delta` (event)

Base schema for a reasoning delta event.

Emitted during streaming when the model outputs incremental reasoning / thinking content. Both Anthropic (extended thinking) and OpenAI-compatible reasoning models emit this event.

Adapter-specific schemas may extend this via .extend().

Subject: adapter:openai-node.reasoning_delta Type: Event

Field	Type	Required
`content`	`string`	yes
`eventType`	`"reasoning_delta"`	yes

`adapter:openai-node.sdk.event` (event)

Envelope schema for sdk.event catch-all subject. Wraps the discriminated union in an envelope for proper TypeScript narrowing at emit time. Pattern matches codex-mcp’s { id, msg } envelope.

Subject: adapter:openai-node.sdk.event Type: Event

Field	Type	Required
`adapterId`	`string \| undefined`	no
`adapterName`	`string \| undefined`	no
`adapterSessionId`	`string \| undefined`	no
`agentId`	`string \| undefined`	no
`event`	{ eventType: "reasoning_delta"; content: string; } \| { eventType: "agent_started"; model?: string \| undefined; } \| { eventType: "agent_complete"; message?: string \| undefined; error?: string \| undefined; } \| { eventType: "error"; message: string; code?: string \| number \| undefined; type?: string \| undefined; } \| { eventType: "tool_started"; toolName: string; toolCallId: string; } \| { eventType: "tool_completed"; toolName: string; toolCallId: string; result: string; success: boolean; } \| { eventType: "tool_calls"; toolCalls: { id: string; type: "function"; function: { name: string; arguments: string; }; }[]; } \| { eventType: "chunk"; id: string; choices: { delta: { content?: string \| null \| undefined; role?: "user" \| "assistant" \| "system" \| "developer" \| "tool" \| undefined; refusal?: string \| null \| undefined; tool_calls?: { index: number; id?: string \| undefined; type?: "function" \| undefined; function?: { name?: string \| undefined; arguments?: string \| undefined; } \| undefined; }[] \| undefined; }; index: number; finish_reason: "length" \| "stop" \| "tool_calls" \| "content_filter" \| "function_call" \| null; logprobs?: unknown; }[]; created: number; model: string; object: "chat.completion.chunk"; service_tier?: "default" \| "auto" \| "priority" \| "flex" \| "scale" \| null \| undefined; system_fingerprint?: string \| undefined; } \| { eventType: "usage"; prompt_tokens: number; completion_tokens: number; total_tokens: number; prompt_tokens_details?: { audio_tokens?: number \| undefined; cached_tokens?: number \| undefined; } \| undefined; completion_tokens_details?: { accepted_prediction_tokens?: number \| undefined; audio_tokens?: number \| undefined; reasoning_tokens?: number \| undefined; rejected_prediction_tokens?: number \| undefined; } \| undefined; } \| { eventType: "message_complete"; content: string \| null; finish_reason: "length" \| "stop" \| "tool_calls" \| "content_filter" \| "function_call" \| null; reasoning?: string \| undefined; tool_calls?: { id: string; type: "function"; function: { name: string; arguments: string; }; }[] \| undefined; } \| { eventType: "reasoning_complete"; content: string; }	yes
`sessionId`	`string \| undefined`	no

`adapter:openai-node.sdk.raw` (event)

Schema for raw OpenAI streaming chunks. Passthrough schema for observability - captures raw API response. Used by stream-bridge to emit unprocessed chunks before normalization.

Subject: adapter:openai-node.sdk.raw Type: Event

Field	Type	Required
`choices`	`Choice[]`	yes
`created`	`number`	yes
`id`	`string`	yes
`model`	`string`	yes
`object`	`"chat.completion.chunk"`	yes
`service_tier`	`"default" \| "auto" \| "priority" \| "flex" \| "scale" \| null \| undefined`	no
`system_fingerprint`	`string \| undefined`	no
`usage`	`CompletionUsage \| null \| undefined`	no

`adapter:openai-node.tool_approval` (rpc)

Scoped tool approval schema for adapter connector buses.

sessionId is optional here because the connector emits the approval request before the agent layer has enriched it. The agent’s wireToolApprovalRpc (or equivalent) injects sessionId from its own context before forwarding to the global AgentSubjects.toolApprove subject, where sessionId is required.

Adapters with a genuinely different wire format (e.g., gemini-sdk’s callId/name) should define their own schema rather than extending this one.

Subject: adapter:openai-node.tool_approval Type: Request (RPC)

Request:

Field	Type	Required
`adapterId`	`string`	yes
`adapterName`	`string`	yes
`adapterSessionId`	`string`	yes
`agentId`	`string`	yes
`args`	`Record<string, unknown> \| undefined`	no
`clientId`	`string \| undefined`	no
`messageId`	`string \| undefined`	no
`occurredAt`	`number \| undefined`	no
`providerConfigId`	`string \| undefined`	no
`reasoning`	`string \| undefined`	no
`sessionId`	`string \| undefined`	no
`toolCallId`	`string`	yes
`toolName`	`string \| undefined`	no
`turnId`	`string \| undefined`	no

Response:

Field	Type	Required
`action`	`"allow" \| "deny"`	yes

`adapter:openai-node.tool_calls` (event)

Schema for a tool calls event.

Emitted when the model requests tool / function execution. Both adapters emit this event with the same structure, except the Anthropic adapter’s individual ToolCall entries also carry blockIndex.

Subject: adapter:openai-node.tool_calls Type: Event

Field	Type	Required
`eventType`	`"tool_calls"`	yes
`toolCalls`	`{ id: string; type: "function"; function: { name: string; arguments: string; }; }[]`	yes

`adapter:openai-node.tool_completed` (event)

Schema for tool completed event. Emitted when tool execution finishes (success or failure).

Subject: adapter:openai-node.tool_completed Type: Event

Field	Type	Required
`eventType`	`"tool_completed"`	yes
`result`	`string`	yes
`success`	`boolean`	yes
`toolCallId`	`string`	yes
`toolName`	`string`	yes

`adapter:openai-node.tool_started` (event)

Schema for tool started event. Emitted when tool execution begins.

Subject: adapter:openai-node.tool_started Type: Event

Field	Type	Required
`eventType`	`"tool_started"`	yes
`toolCallId`	`string`	yes
`toolName`	`string`	yes