@ai-sdk/google 4.0.0-beta.7 → 4.0.0-beta.82

package/src/interactions/google-interactions-language-model-options.ts

@@ -0,0 +1,262 @@
+import {
+  lazySchema,
+  zodSchema,
+  type InferSchema,
+} from '@ai-sdk/provider-utils';
+import { z } from 'zod/v4';
+
+/**
+ * Type-only union of Gemini model IDs that the Interactions API accepts via
+ * `model:`. Mirrors `Model` from `googleapis/js-genai`
+ * `src/interactions/resources/interactions.ts`.
+ *
+ * Kept as a separate type from `GoogleModelId` even though most IDs overlap;
+ * the two surfaces (`:generateContent` vs `/interactions`) are independent and
+ * may diverge over time.
+ */
+export type GoogleInteractionsModelId =
+  | 'gemini-2.5-computer-use-preview-10-2025'
+  | 'gemini-2.5-flash'
+  | 'gemini-2.5-flash-image'
+  | 'gemini-2.5-flash-lite'
+  | 'gemini-2.5-flash-lite-preview-09-2025'
+  | 'gemini-2.5-flash-native-audio-preview-12-2025'
+  | 'gemini-2.5-flash-preview-09-2025'
+  | 'gemini-2.5-flash-preview-tts'
+  | 'gemini-2.5-pro'
+  | 'gemini-2.5-pro-preview-tts'
+  | 'gemini-3-flash-preview'
+  | 'gemini-3-pro-image-preview'
+  | 'gemini-3-pro-preview'
+  | 'gemini-3.1-pro-preview'
+  | 'gemini-3.1-flash-image-preview'
+  | 'gemini-3.1-flash-lite-preview'
+  | 'gemini-3.1-flash-tts-preview'
+  | 'gemini-3.5-flash'
+  | 'lyria-3-clip-preview'
+  | 'lyria-3-pro-preview'
+  | (string & {});
+
+/**
+ * Provider-options schema for `google.interactions(...)` calls. Read from the
+ * shared `providerOptions.google.*` namespace (per PRD); per-call options that
+ * the AI SDK doesn't natively expose live here.
+ *
+ * All fields are `.nullish()` per the existing google provider convention.
+ */
+export const googleInteractionsLanguageModelOptions = lazySchema(() =>
+  zodSchema(
+    z.object({
+      previousInteractionId: z.string().nullish(),
+      store: z.boolean().nullish(),
+
+      agent: z.string().nullish(),
+      agentConfig: z
+        .union([
+          z
+            .object({
+              type: z.literal('dynamic'),
+            })
+            .loose(),
+          z.object({
+            type: z.literal('deep-research'),
+            thinkingSummaries: z.enum(['auto', 'none']).nullish(),
+            visualization: z.enum(['off', 'auto']).nullish(),
+            collaborativePlanning: z.boolean().nullish(),
+          }),
+        ])
+        .nullish(),
+
+      thinkingLevel: z.enum(['minimal', 'low', 'medium', 'high']).nullish(),
+      thinkingSummaries: z.enum(['auto', 'none']).nullish(),
+
+      /**
+       * Output-format entries that map directly to the API's `response_format`
+       * array. Use this to request image, audio, or non-JSON text outputs
+       * with full control over `mime_type`, `aspect_ratio`, and `image_size`.
+       *
+       * Entries are sent in order. The AI SDK call-level `responseFormat: {
+       * type: 'json', schema }` still drives JSON-mode and adds a matching
+       * text entry automatically; entries listed here are appended.
+       */
+      responseFormat: z
+        .array(
+          z.union([
+            z
+              .object({
+                type: z.literal('text'),
+                mimeType: z.string().nullish(),
+                schema: z.unknown().nullish(),
+              })
+              .loose(),
+            z
+              .object({
+                type: z.literal('image'),
+                mimeType: z.string().nullish(),
+                aspectRatio: z
+                  .enum([
+                    '1:1',
+                    '2:3',
+                    '3:2',
+                    '3:4',
+                    '4:3',
+                    '4:5',
+                    '5:4',
+                    '9:16',
+                    '16:9',
+                    '21:9',
+                    '1:8',
+                    '8:1',
+                    '1:4',
+                    '4:1',
+                  ])
+                  .nullish(),
+                imageSize: z.enum(['1K', '2K', '4K', '512']).nullish(),
+              })
+              .loose(),
+            z
+              .object({
+                type: z.literal('audio'),
+                mimeType: z.string().nullish(),
+              })
+              .loose(),
+          ]),
+        )
+        .nullish(),
+
+      /**
+       * @deprecated Use `responseFormat` with a `{ type: 'image', ... }`
+       * entry instead. Retained for backwards compatibility; the SDK
+       * translates it into a matching `response_format` image entry and
+       * emits a warning when set.
+       */
+      imageConfig: z
+        .object({
+          aspectRatio: z
+            .enum([
+              '1:1',
+              '2:3',
+              '3:2',
+              '3:4',
+              '4:3',
+              '4:5',
+              '5:4',
+              '9:16',
+              '16:9',
+              '21:9',
+              '1:8',
+              '8:1',
+              '1:4',
+              '4:1',
+            ])
+            .nullish(),
+          imageSize: z.enum(['1K', '2K', '4K', '512']).nullish(),
+        })
+        .nullish(),
+      mediaResolution: z
+        .enum(['low', 'medium', 'high', 'ultra_high'])
+        .nullish(),
+
+      responseModalities: z
+        .array(z.enum(['text', 'image', 'audio', 'video', 'document']))
+        .nullish(),
+      serviceTier: z.enum(['flex', 'standard', 'priority']).nullish(),
+
+      /**
+       * Alternative to AI SDK `system` message. If both are set, the AI SDK
+       * `system` message wins and a warning is emitted.
+       */
+      systemInstruction: z.string().nullish(),
+
+      /**
+       * Per-block signature for round-tripping `thought.signature` and
+       * `function_call.signature` blocks. Set by the SDK on output reasoning /
+       * tool-call parts; passed back unchanged on input parts so the API
+       * accepts the prior turn.
+       */
+      signature: z.string().nullish(),
+
+      /**
+       * Set by the SDK on output assistant messages. The converter uses it to
+       * decide which messages to drop when compacting under
+       * `previousInteractionId`.
+       */
+      interactionId: z.string().nullish(),
+
+      /**
+       * Maximum time, in milliseconds, to poll a background interaction (agent
+       * call) before giving up. Defaults to 30 minutes. Long-running agents
+       * such as deep research can take tens of minutes — increase if needed.
+       */
+      pollingTimeoutMs: z.number().int().positive().nullish(),
+
+      /**
+       * Run the interaction in the background. Required for agents whose
+       * server-side workflow cannot complete within a single request/response.
+       * When `true`, the POST returns with a non-terminal status and the SDK
+       * polls `GET /interactions/{id}` until the work completes. Some agents
+       * reject `true`; see the agent's documentation for which mode it
+       * requires.
+       */
+      background: z.boolean().nullish(),
+
+      /**
+       * Environment configuration for the agent sandbox. Only applies to agent
+       * calls (`google.interactions({ agent })`); ignored on model-id calls.
+       *
+       *   - `"remote"`: provision a fresh sandbox for this call.
+       *   - any other string: an existing `environment_id` to reuse.
+       *   - object: provision a fresh sandbox and optionally preload `sources`
+       *     and/or constrain outbound traffic via `network`.
+       */
+      environment: z
+        .union([
+          z.string(),
+          z.object({
+            type: z.literal('remote'),
+            sources: z
+              .array(
+                z.union([
+                  z.object({
+                    type: z.literal('gcs'),
+                    source: z.string(),
+                    target: z.string().nullish(),
+                  }),
+                  z.object({
+                    type: z.literal('repository'),
+                    source: z.string(),
+                    target: z.string().nullish(),
+                  }),
+                  z.object({
+                    type: z.literal('inline'),
+                    content: z.string(),
+                    target: z.string(),
+                  }),
+                ]),
+              )
+              .nullish(),
+            network: z
+              .union([
+                z.literal('disabled'),
+                z.object({
+                  allowlist: z.array(
+                    z.object({
+                      domain: z.string(),
+                      transform: z
+                        .array(z.record(z.string(), z.string()))
+                        .nullish(),
+                    }),
+                  ),
+                }),
+              ])
+              .nullish(),
+          }),
+        ])
+        .nullish(),
+    }),
+  ),
+);
+
+export type GoogleLanguageModelInteractionsOptions = InferSchema<
+  typeof googleInteractionsLanguageModelOptions
+>;