@firebase/ai 2.1.0 → 2.2.0-canary.095c098de

package/dist/esm/src/types/content.d.ts

@@ -38,6 +38,11 @@ export interface TextPart {
     inlineData?: never;
     functionCall?: never;
     functionResponse?: never;
+    thought?: boolean;
+    /**
+     * @internal
+     */
+    thoughtSignature?: string;
 }
 /**
  * Content part interface if the part represents an image.
@@ -52,6 +57,11 @@ export interface InlineDataPart {
      * Applicable if `inlineData` is a video.
      */
     videoMetadata?: VideoMetadata;
+    thought?: boolean;
+    /**
+     * @internal
+     */
+    thoughtSignature?: never;
 }
 /**
  * Describes the input video content.
@@ -78,6 +88,11 @@ export interface FunctionCallPart {
     inlineData?: never;
     functionCall: FunctionCall;
     functionResponse?: never;
+    thought?: boolean;
+    /**
+     * @internal
+     */
+    thoughtSignature?: never;
 }
 /**
  * Content part interface if the part represents {@link FunctionResponse}.
@@ -88,6 +103,11 @@ export interface FunctionResponsePart {
     inlineData?: never;
     functionCall?: never;
     functionResponse: FunctionResponse;
+    thought?: boolean;
+    /**
+     * @internal
+     */
+    thoughtSignature?: never;
 }
 /**
  * Content part interface if the part represents {@link FileData}
@@ -99,6 +119,11 @@ export interface FileDataPart {
     functionCall?: never;
     functionResponse?: never;
     fileData: FileData;
+    thought?: boolean;
+    /**
+     * @internal
+     */
+    thoughtSignature?: never;
 }
 /**
  * A predicted {@link FunctionCall} returned from the model
@@ -107,6 +132,15 @@ export interface FileDataPart {
  * @public
  */
 export interface FunctionCall {
+    /**
+     * The id of the function call. This must be sent back in the associated {@link FunctionResponse}.
+     *
+     *
+     * @remarks This property is only supported in the Gemini Developer API ({@link GoogleAIBackend}).
+     * When using the Gemini Developer API ({@link GoogleAIBackend}), this property will be
+     * `undefined`.
+     */
+    id?: string;
     name: string;
     args: object;
 }
@@ -120,6 +154,14 @@ export interface FunctionCall {
  * @public
  */
 export interface FunctionResponse {
+    /**
+     * The id of the {@link FunctionCall}.
+     *
+     * @remarks This property is only supported in the Gemini Developer API ({@link GoogleAIBackend}).
+     * When using the Gemini Developer API ({@link GoogleAIBackend}), this property will be
+     * `undefined`.
+     */
+    id?: string;
     name: string;
     response: object;
 }

package/dist/esm/src/types/enums.d.ts

@@ -304,6 +304,11 @@ export declare const ResponseModality: {
      * @beta
      */
     readonly IMAGE: "IMAGE";
+    /**
+     * Audio.
+     * @beta
+     */
+    readonly AUDIO: "AUDIO";
 };
 /**
  * Generation modalities to be returned in generation responses.

package/dist/esm/src/types/error.d.ts

@@ -60,6 +60,8 @@ export declare const AIErrorCode: {
     readonly RESPONSE_ERROR: "response-error";
     /** An error occurred while performing a fetch. */
     readonly FETCH_ERROR: "fetch-error";
+    /** An error occurred because an operation was attempted on a closed session. */
+    readonly SESSION_CLOSED: "session-closed";
     /** An error associated with a Content object.  */
     readonly INVALID_CONTENT: "invalid-content";
     /** An error due to the Firebase API not being enabled in the Console. */

package/dist/esm/src/types/imagen/internal.d.ts

@@ -59,6 +59,14 @@ export interface ImagenResponseInternal {
          * The reason why the image was filtered.
          */
         raiFilteredReason?: string;
+        /**
+         * The safety attributes.
+         *
+         * This type is currently unused in the SDK. It is sent back because our requests set
+         * `includeSafetyAttributes`. This property is currently only used to avoid throwing an error
+         * when encountering this unsupported prediction type.
+         */
+        safetyAttributes?: unknown;
     }>;
 }
 /**
@@ -81,6 +89,7 @@ export interface ImagenResponseInternal {
  *     "personGeneration": "allow_all",
  *     "sampleCount": 2,
  *     "includeRaiReason": true,
+ *     "includeSafetyAttributes": true,
  *     "aspectRatio": "9:16"
  *   }
  * }
@@ -108,6 +117,7 @@ export interface PredictRequestBody {
         safetyFilterLevel?: string;
         personGeneration?: string;
         includeRaiReason: boolean;
+        includeSafetyAttributes: boolean;
     };
 }
 /**

package/dist/esm/src/types/live-responses.d.ts

@@ -0,0 +1,53 @@
+/**
+ * @license
+ * Copyright 2025 Google LLC
+ *
+ * Licensed under the Apache License, Version 2.0 (the "License");
+ * you may not use this file except in compliance with the License.
+ * You may obtain a copy of the License at
+ *
+ *   http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed to in writing, software
+ * distributed under the License is distributed on an "AS IS" BASIS,
+ * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ * See the License for the specific language governing permissions and
+ * limitations under the License.
+ */
+import { Content, GenerativeContentBlob, Part } from './content';
+import { LiveGenerationConfig, Tool, ToolConfig } from './requests';
+/**
+ * User input that is sent to the model.
+ *
+ * @internal
+ */
+export interface _LiveClientContent {
+    clientContent: {
+        turns: [Content];
+        turnComplete: boolean;
+    };
+}
+/**
+ * User input that is sent to the model in real time.
+ *
+ * @internal
+ */
+export interface _LiveClientRealtimeInput {
+    realtimeInput: {
+        mediaChunks: GenerativeContentBlob[];
+    };
+}
+/**
+ * The first message in a Live session, used to configure generation options.
+ *
+ * @internal
+ */
+export interface _LiveClientSetup {
+    setup: {
+        model: string;
+        generationConfig?: LiveGenerationConfig;
+        tools?: Tool[];
+        toolConfig?: ToolConfig;
+        systemInstruction?: string | Part | Content;
+    };
+}

package/dist/esm/src/types/requests.d.ts

@@ -37,6 +37,17 @@ export interface ModelParams extends BaseParams {
     toolConfig?: ToolConfig;
     systemInstruction?: string | Part | Content;
 }
+/**
+ * Params passed to {@link getLiveGenerativeModel}.
+ * @beta
+ */
+export interface LiveModelParams {
+    model: string;
+    generationConfig?: LiveGenerationConfig;
+    tools?: Tool[];
+    toolConfig?: ToolConfig;
+    systemInstruction?: string | Part | Content;
+}
 /**
  * Request sent through {@link GenerativeModel.generateContent}
  * @public
@@ -107,6 +118,56 @@ export interface GenerationConfig {
      */
     thinkingConfig?: ThinkingConfig;
 }
+/**
+ * Configuration parameters used by {@link LiveGenerativeModel} to control live content generation.
+ *
+ * @beta
+ */
+export interface LiveGenerationConfig {
+    /**
+     * Configuration for speech synthesis.
+     */
+    speechConfig?: SpeechConfig;
+    /**
+     * Specifies the maximum number of tokens that can be generated in the response. The number of
+     * tokens per word varies depending on the language outputted. Is unbounded by default.
+     */
+    maxOutputTokens?: number;
+    /**
+     * Controls the degree of randomness in token selection. A `temperature` value of 0 means that the highest
+     * probability tokens are always selected. In this case, responses for a given prompt are mostly
+     * deterministic, but a small amount of variation is still possible.
+     */
+    temperature?: number;
+    /**
+     * Changes how the model selects tokens for output. Tokens are
+     * selected from the most to least probable until the sum of their probabilities equals the `topP`
+     * value. For example, if tokens A, B, and C have probabilities of 0.3, 0.2, and 0.1 respectively
+     * and the `topP` value is 0.5, then the model will select either A or B as the next token by using
+     * the `temperature` and exclude C as a candidate. Defaults to 0.95 if unset.
+     */
+    topP?: number;
+    /**
+     * Changes how the model selects token for output. A `topK` value of 1 means the select token is
+     * the most probable among all tokens in the model's vocabulary, while a `topK` value 3 means that
+     * the next token is selected from among the 3 most probably using probabilities sampled. Tokens
+     * are then further filtered with the highest selected `temperature` sampling. Defaults to 40
+     * if unspecified.
+     */
+    topK?: number;
+    /**
+     * Positive penalties.
+     */
+    presencePenalty?: number;
+    /**
+     * Frequency penalties.
+     */
+    frequencyPenalty?: number;
+    /**
+     * The modalities of the response.
+     */
+    responseModalities?: ResponseModality[];
+}
 /**
  * Params for {@link GenerativeModel.startChat}.
  * @public
@@ -146,7 +207,10 @@ export interface RequestOptions {
      */
     timeout?: number;
     /**
-     * Base url for endpoint. Defaults to https://firebasevertexai.googleapis.com
+     * Base url for endpoint. Defaults to
+     * https://firebasevertexai.googleapis.com, which is the
+     * {@link https://console.cloud.google.com/apis/library/firebasevertexai.googleapis.com?project=_ | Firebase AI Logic API}
+     * (used regardless of your chosen Gemini API provider).
      */
     baseUrl?: string;
 }
@@ -302,4 +366,48 @@ export interface ThinkingConfig {
      * feature or if the specified budget is not within the model's supported range.
      */
     thinkingBudget?: number;
+    /**
+     * Whether to include "thought summaries" in the model's response.
+     *
+     * @remarks
+     * Thought summaries provide a brief overview of the model's internal thinking process,
+     * offering insight into how it arrived at the final answer. This can be useful for
+     * debugging, understanding the model's reasoning, and verifying its accuracy.
+     */
+    includeThoughts?: boolean;
+}
+/**
+ * Configuration for a pre-built voice.
+ *
+ * @beta
+ */
+export interface PrebuiltVoiceConfig {
+    /**
+     * The voice name to use for speech synthesis.
+     *
+     * For a full list of names and demos of what each voice sounds like, see {@link https://cloud.google.com/text-to-speech/docs/chirp3-hd | Chirp 3: HD Voices}.
+     */
+    voiceName?: string;
+}
+/**
+ * Configuration for the voice to used in speech synthesis.
+ *
+ * @beta
+ */
+export interface VoiceConfig {
+    /**
+     * Configures the voice using a pre-built voice configuration.
+     */
+    prebuiltVoiceConfig?: PrebuiltVoiceConfig;
+}
+/**
+ * Configures speech synthesis.
+ *
+ * @beta
+ */
+export interface SpeechConfig {
+    /**
+     * Configures the voice to be used in speech synthesis.
+     */
+    voiceConfig?: VoiceConfig;
 }

package/dist/esm/src/types/responses.d.ts

@@ -48,15 +48,34 @@ export interface EnhancedGenerateContentResponse extends GenerateContentResponse
      */
     text: () => string;
     /**
-     * Aggregates and returns all {@link InlineDataPart}s from the {@link GenerateContentResponse}'s
-     * first candidate.
-     *
-     * @returns An array of {@link InlineDataPart}s containing data from the response, if available.
+     * Aggregates and returns every {@link InlineDataPart} from the first candidate of
+     * {@link GenerateContentResponse}.
      *
      * @throws If the prompt or candidate was blocked.
      */
     inlineDataParts: () => InlineDataPart[] | undefined;
+    /**
+     * Aggregates and returns every {@link FunctionCall} from the first candidate of
+     * {@link GenerateContentResponse}.
+     *
+     * @throws If the prompt or candidate was blocked.
+     */
     functionCalls: () => FunctionCall[] | undefined;
+    /**
+     * Aggregates and returns every {@link TextPart} with their `thought` property set
+     * to `true` from the first candidate of {@link GenerateContentResponse}.
+     *
+     * @throws If the prompt or candidate was blocked.
+     *
+     * @remarks
+     * Thought summaries provide a brief overview of the model's internal thinking process,
+     * offering insight into how it arrived at the final answer. This can be useful for
+     * debugging, understanding the model's reasoning, and verifying its accuracy.
+     *
+     * Thoughts will only be included if {@link ThinkingConfig.includeThoughts} is
+     * set to `true`.
+     */
+    thoughtSummary: () => string | undefined;
 }
 /**
  * Individual response from {@link GenerativeModel.generateContent} and
@@ -378,3 +397,67 @@ export interface CountTokensResponse {
      */
     promptTokensDetails?: ModalityTokenCount[];
 }
+/**
+ * An incremental content update from the model.
+ *
+ * @beta
+ */
+export interface LiveServerContent {
+    type: 'serverContent';
+    /**
+     * The content that the model has generated as part of the current conversation with the user.
+     */
+    modelTurn?: Content;
+    /**
+     * Indicates whether the turn is complete. This is `undefined` if the turn is not complete.
+     */
+    turnComplete?: boolean;
+    /**
+     * Indicates whether the model was interrupted by the client. An interruption occurs when
+     * the client sends a message before the model finishes it's turn. This is `undefined` if the
+     * model was not interrupted.
+     */
+    interrupted?: boolean;
+}
+/**
+ * A request from the model for the client to execute one or more functions.
+ *
+ * @beta
+ */
+export interface LiveServerToolCall {
+    type: 'toolCall';
+    /**
+     * An array of function calls to run.
+     */
+    functionCalls: FunctionCall[];
+}
+/**
+ * Notification to cancel a previous function call triggered by {@link LiveServerToolCall}.
+ *
+ * @beta
+ */
+export interface LiveServerToolCallCancellation {
+    type: 'toolCallCancellation';
+    /**
+     * IDs of function calls that were cancelled. These refer to the `id` property of a {@link FunctionCall}.
+     */
+    functionIds: string[];
+}
+/**
+ * The types of responses that can be returned by {@link LiveSession.receive}.
+ *
+ * @beta
+ */
+export declare const LiveResponseType: {
+    SERVER_CONTENT: string;
+    TOOL_CALL: string;
+    TOOL_CALL_CANCELLATION: string;
+};
+/**
+ * The types of responses that can be returned by {@link LiveSession.receive}.
+ * This is a property on all messages that can be used for type narrowing. This property is not
+ * returned by the server, it is assigned to a server message object once it's parsed.
+ *
+ * @beta
+ */
+export type LiveResponseType = (typeof LiveResponseType)[keyof typeof LiveResponseType];

package/dist/esm/src/websocket.d.ts

@@ -0,0 +1,67 @@
+/**
+ * @license
+ * Copyright 2025 Google LLC
+ *
+ * Licensed under the Apache License, Version 2.0 (the "License");
+ * you may not use this file except in compliance with the License.
+ * You may obtain a copy of the License at
+ *
+ *   http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed to in writing, software
+ * distributed under the License is distributed on an "AS IS" BASIS,
+ * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ * See the License for the specific language governing permissions and
+ * limitations under the License.
+ */
+/**
+ * A standardized interface for interacting with a WebSocket connection.
+ * This abstraction allows the SDK to use the appropriate WebSocket implementation
+ * for the current JS environment (Browser vs. Node) without
+ * changing the core logic of the `LiveSession`.
+ * @internal
+ */
+export interface WebSocketHandler {
+    /**
+     * Establishes a connection to the given URL.
+     *
+     * @param url The WebSocket URL (e.g., wss://...).
+     * @returns A promise that resolves on successful connection or rejects on failure.
+     */
+    connect(url: string): Promise<void>;
+    /**
+     * Sends data over the WebSocket.
+     *
+     * @param data The string or binary data to send.
+     */
+    send(data: string | ArrayBuffer): void;
+    /**
+     * Returns an async generator that yields parsed JSON objects from the server.
+     * The yielded type is `unknown` because the handler cannot guarantee the shape of the data.
+     * The consumer is responsible for type validation.
+     * The generator terminates when the connection is closed.
+     *
+     * @returns A generator that allows consumers to pull messages using a `for await...of` loop.
+     */
+    listen(): AsyncGenerator<unknown>;
+    /**
+     * Closes the WebSocket connection.
+     *
+     * @param code - A numeric status code explaining why the connection is closing.
+     * @param reason - A human-readable string explaining why the connection is closing.
+     */
+    close(code?: number, reason?: string): Promise<void>;
+}
+/**
+ * A wrapper for the native `WebSocket` available in both Browsers and Node >= 22.
+ *
+ * @internal
+ */
+export declare class WebSocketHandlerImpl implements WebSocketHandler {
+    private ws?;
+    constructor();
+    connect(url: string): Promise<void>;
+    send(data: string | ArrayBuffer): void;
+    listen(): AsyncGenerator<unknown>;
+    close(code?: number, reason?: string): Promise<void>;
+}