ChromeAIBackendAdapter

Defined in: backend-browser/src/chrome-ai.ts:65

Backend adapter interface.

Backend adapters handle actual API calls to AI providers. They transform universal IR into provider-specific API requests and normalize responses back to IR.

Implements

• BackendAdapter

Constructors

Constructor

new ChromeAIBackendAdapter(config?): ChromeAIBackendAdapter

Defined in: backend-browser/src/chrome-ai.ts:69

Parameters

config?

Partial<BackendAdapterConfig>

Returns

ChromeAIBackendAdapter

Properties

metadata

readonly metadata: AdapterMetadata

Defined in: backend-browser/src/chrome-ai.ts:66

Adapter metadata for identification and capabilities.

Implementation of

BackendAdapter.metadata

Methods

estimateCost()

estimateCost(_request): Promise<number | null>

Defined in: backend-browser/src/chrome-ai.ts:371

Estimate the cost of a Chrome AI request.

Chrome AI is completely free as it runs locally on-device, so this always returns 0. No API keys or payment required.

Parameters

_request

IRChatRequest

IR chat request (unused)

Returns

Promise<number | null>

Promise resolving to 0 (Chrome AI is free)

Example

const adapter = new ChromeAIBackendAdapter();
const cost = await adapter.estimateCost(request);
console.log(cost); // Always 0

Implementation of

BackendAdapter.estimateCost

---

execute()

execute(request, signal?): Promise<IRChatResponse>

Defined in: backend-browser/src/chrome-ai.ts:136

Execute a chat request using Chrome’s built-in AI.

This method processes an IR chat request through Chrome AI’s streaming API, collecting all chunks into a complete response. Chrome AI always streams internally, so this method consumes the stream and returns the final result. Requires Chrome 129+ with AI features enabled.

Parameters

request

IRChatRequest

Universal IR chat request

signal?

AbortSignal

Optional AbortSignal for cancellation

Returns

Promise<IRChatResponse>

Promise resolving to IR chat response

Throws

If Chrome AI is unavailable or request fails

Example

const adapter = new ChromeAIBackendAdapter();
const response = await adapter.execute({
  messages: [{ role: 'user', content: 'Hello!' }],
  parameters: { temperature: 0.7, topK: 40 },
  metadata: { requestId: 'req_123', timestamp: Date.now(), provenance: {} }
});
console.log(response.message.content);

Implementation of

BackendAdapter.execute

---

executeStream()

executeStream(request, signal?): IRChatStream

Defined in: backend-browser/src/chrome-ai.ts:211

Execute a streaming chat request using Chrome’s built-in AI.

This async generator method processes an IR chat request through Chrome AI’s streaming interface, yielding IR stream chunks as they arrive. It checks for Chrome AI availability, creates a session with the specified parameters, combines all messages into a single prompt, and streams the response. Supports abort signals for cancellation.

Parameters

request

IRChatRequest

Universal IR chat request

signal?

AbortSignal

Optional AbortSignal for cancellation

Returns

IRChatStream

Yields

IR stream chunks including start, content, done, and error events

Throws

Yields error chunk if Chrome AI unavailable or stream fails

Example

const adapter = new ChromeAIBackendAdapter();
for await (const chunk of adapter.executeStream(request)) {
  if (chunk.type === 'content') {
    console.log('Delta:', chunk.delta);
  } else if (chunk.type === 'done') {
    console.log('Complete:', chunk.message.content);
  }
}

Implementation of

BackendAdapter.executeStream

---

fromIR()

fromIR(request): IRChatRequest

Defined in: backend-browser/src/chrome-ai.ts:97

Convert IR request to provider format (passthrough - uses IR internally).

Parameters

request

IRChatRequest

Returns

IRChatRequest

Implementation of

BackendAdapter.fromIR

---

healthCheck()

healthCheck(): Promise<boolean>

Defined in: backend-browser/src/chrome-ai.ts:342

Check if Chrome AI is available and healthy.

This method verifies that Chrome AI is present in the browser environment and reports being available. Returns false if Chrome AI is not supported, not enabled, or if the device doesn’t support on-device AI.

Returns

Promise<boolean>

Promise resolving to true if Chrome AI is available, false otherwise

Example

const adapter = new ChromeAIBackendAdapter();
if (await adapter.healthCheck()) {
  console.log('Chrome AI is ready!');
} else {
  console.log('Chrome AI not available - use a different backend');
}

Implementation of

BackendAdapter.healthCheck

---

toIR()

toIR(response, _originalRequest, _latencyMs): IRChatResponse

Defined in: backend-browser/src/chrome-ai.ts:104

Convert provider response to IR format (passthrough - uses IR internally).

Parameters

response

IRChatResponse

_originalRequest

IRChatRequest

_latencyMs

number

Returns

IRChatResponse

Implementation of

BackendAdapter.toIR