AI SDK v6 Protocol

The AI SDK v6 adapter translates Awaken’s internal AgentEvent stream into the Vercel AI SDK v6 UI Message Stream format. This allows any AI SDK-compatible frontend (useChat, useAssistant) to consume agent output without custom parsing or a custom frontend transport adapter. Awaken-specific data-* parts are optional metadata for richer consoles; ordinary chat UIs may ignore them.

Endpoint

POST /v1/ai-sdk/chat

Request Body

{
  "messages": [
    {
      "id": "msg-1",
      "role": "user",
      "parts": [{ "type": "text", "text": "Hello" }]
    }
  ],
  "threadId": "optional-thread-id",
  "agentId": "optional-agent-id"
}

Field	Type	Required	Description
messages	UIMessage[]	yes	AI SDK v6 UI messages. text and file parts become runtime content; tool/data/source parts are UI state and are not sent to the model.
threadId	string	no	Existing thread to continue. Omit to create a new thread.
agentId	string	no	Target agent. Uses the default agent when omitted.

file parts may use hosted URLs or Data URLs. Images, audio, video, PDFs, and text-like documents are converted to runtime ContentBlocks. The selected model must advertise compatible input modalities when modalities are enforced.

Response

SSE stream (text/event-stream). Each line is a JSON-encoded UIStreamEvent.

Auxiliary Routes

Route	Method	Description
/v1/ai-sdk/threads/:thread_id/runs	POST	Start a thread-scoped run.
/v1/ai-sdk/agents/:agent_id/runs	POST	Start an agent-scoped run.
/v1/ai-sdk/agent-previews/runs	POST	Run a draft AgentSpec against the current registries without persisting it.
/v1/ai-sdk/chat/:thread_id/stream	GET	Reconnect to an active SSE stream by thread ID.
/v1/ai-sdk/threads/:thread_id/stream	GET	Alias for thread stream reconnect.
/v1/ai-sdk/threads/:thread_id/replay	GET	Replay durable protocol frames for a thread.
/v1/ai-sdk/chat/:thread_id/replay	GET	Alias for durable protocol replay.
/v1/ai-sdk/threads/:thread_id/messages	GET	Retrieve thread message history.
/v1/ai-sdk/threads/:thread_id/cancel	POST	Cancel the active or queued run on a thread.
/v1/ai-sdk/threads/:thread_id/interrupt	POST	Interrupt a thread, supersede queued work, and cancel the active run.

Replay routes require a configured ProtocolReplayLog. They accept ?cursor= or Last-Event-ID, plus ?limit= (default 100, max 500), and return SSE frames with the protocol replay cursor as the SSE id. Missing replay storage returns 503; malformed cursors return 400; expired cursors return 410 Gone.

Event Mapping

The AiSdkEncoder is a stateful transcoder that converts AgentEvent variants into UIStreamEvent variants. It tracks open text blocks and reasoning blocks across tool-call boundaries.

AgentEvent	UIStreamEvent(s)
RunStart	MessageStart + Data("run-info", ...)
TextDelta	TextStart (if block not open) + TextDelta
ReasoningDelta	ReasoningStart (if block not open) + ReasoningDelta
ReasoningEncryptedValue	ReasoningStart (if not open) + ReasoningDelta
ToolCallStart	Close open text/reasoning blocks, then ToolCallStart
ToolCallDelta	ToolCallDelta
ToolCallDone	ToolCallEnd
StepStart	(no direct mapping)
StepEnd	(no direct mapping)
InferenceComplete	Data("inference-complete", ...)
MessagesSnapshot	Data("messages-snapshot", ...)
StateSnapshot	Data("state-snapshot", ...)
StateDelta	Data("state-delta", ...)
ActivitySnapshot	Data("activity-snapshot", ...)
ActivityDelta	Data("activity-delta", ...)
RunFinish	Close open blocks, Data("finish", ...), Finish

UIStreamEvent Types

The wire format uses the type field as a discriminant, serialized in kebab-case:

• start — message lifecycle start, carries optional messageId and messageMetadata
• text-start, text-delta, text-end — text block lifecycle with content ID
• reasoning-start, reasoning-delta, reasoning-end — reasoning block lifecycle
• tool-call-start, tool-call-delta, tool-call-end — tool call lifecycle
• data — arbitrary named data events (state snapshots, activity, inference metadata)
• finish — terminal event with finish reason and usage summary

Text Block Lifecycle

The encoder automatically manages text block open/close boundaries:

1. First TextDelta opens a text block (TextStart).
2. Subsequent deltas append to the open block.
3. When a ToolCallStart arrives, the encoder closes any open text or reasoning block before emitting tool events.
4. After tool execution completes, new text deltas open a fresh block with an incremented ID.

This ensures the frontend receives well-formed block boundaries even though the runtime emits flat delta events.

Related

• Events — full AgentEvent enum
• HTTP API — server configuration
• Integrate AI SDK Frontend — frontend integration guide