@tanstack/ai-gemini 0.32.1 → 0.33.1

package/src/adapters/text.ts

@@ -1,5 +1,10 @@
 import { FinishReason } from '@google/genai'
-import { EventType, normalizeSystemPrompts } from '@tanstack/ai'
+import {
+  EventType,
+  fileReferenceFor,
+  isFileSource,
+  normalizeSystemPrompts,
+} from '@tanstack/ai'
 import { toRunErrorRawEvent } from '@tanstack/ai/adapter-internals'
 import { BaseTextAdapter } from '@tanstack/ai/adapters'
 import { convertToolsToProviderFormat } from '../tools/tool-converter'
@@ -107,8 +112,11 @@ function contentPartToInteraction(part: ContentPart): InteractionContent {
     source.type === 'data'
       ? source.mimeType
       : (source.mimeType ?? DEFAULT_MEDIA_MIME_TYPES[part.type])
-  const base =
-    source.type === 'data'
+  // A Gemini Files API handle maps to the `uri` field, same as a public URL;
+  // `fileReferenceFor` throws when another provider issued it.
+  const base = isFileSource(source)
+    ? { uri: fileReferenceFor(source, 'gemini'), mime_type: mimeType }
+    : source.type === 'data'
       ? { data: source.value, mime_type: mimeType }
       : { uri: source.value, mime_type: mimeType }
 
@@ -216,6 +224,8 @@ export class GeminiTextAdapter<
 > {
   override readonly kind = 'text' as const
   readonly name = 'gemini' as const
+  // Consumes Gemini Files API references (geminiFiles()) as fileData.fileUri.
+  override readonly supportsFileSources = true
 
   private readonly client: GoogleGenAI
 
@@ -920,6 +930,13 @@ export class GeminiTextAdapter<
       case 'audio':
       case 'video':
       case 'document': {
+        // File references (Gemini Files API) and public URLs both pass
+        // through as `fileData`; Gemini fetches the URI server-side. A
+        // file source's handle is the file URI (throws when another provider
+        // issued it).
+        const fileUri = isFileSource(part.source)
+          ? fileReferenceFor(part.source, this.name)
+          : part.source.value
         const geminiPart: Part =
           part.source.type === 'data'
             ? {
@@ -930,7 +947,7 @@ export class GeminiTextAdapter<
               }
             : {
                 fileData: {
-                  fileUri: part.source.value,
+                  fileUri,
                   // For URL sources, use provided mimeType or fall back to
                   // reasonable defaults.
                   mimeType:
@@ -1045,6 +1062,9 @@ export class GeminiTextAdapter<
                 },
               })
             } else {
+              const fileUri = isFileSource(part.source)
+                ? fileReferenceFor(part.source, this.name)
+                : part.source.value
               const defaultMimeType = {
                 image: 'image/jpeg',
                 audio: 'audio/mp3',
@@ -1053,7 +1073,7 @@ export class GeminiTextAdapter<
               }[part.type]
               mediaParts.push({
                 fileData: {
-                  fileUri: part.source.value,
+                  fileUri,
                   mimeType: part.source.mimeType ?? defaultMimeType,
                 },
               })

package/src/adapters/video.ts

@@ -2,7 +2,12 @@ import {
   GenerateVideosOperation,
   VideoGenerationReferenceType,
 } from '@google/genai'
-import { resolveMediaPrompt } from '@tanstack/ai'
+import {
+  fileReferenceFor,
+  isFileSource,
+  resolveMediaPrompt,
+  unsupportedFileSourceError,
+} from '@tanstack/ai'
 import { BaseVideoAdapter, snapToDurationOption } from '@tanstack/ai/adapters'
 import { arrayBufferToBase64 } from '@tanstack/ai-utils'
 import { createGeminiClient, getGeminiApiKeyFromEnv } from '../utils'
@@ -93,6 +98,14 @@ async function imagePartToVeoImage(
       mimeType: part.source.mimeType || 'image/png',
     }
   }
+  if (isFileSource(part.source)) {
+    // Veo's predict API accepts only inline bytes or a gs:// reference — there's
+    // no way to reference a Files API handle here.
+    throw unsupportedFileSourceError(
+      'gemini',
+      'for Veo video generation, which needs inline image bytes or a gs:// reference — pass a data: URI or gs:// URL',
+    )
+  }
   const url = part.source.value
   if (url.startsWith('gs://')) {
     return {
@@ -143,15 +156,21 @@ async function imagePartToVeoImage(
 function mediaPartToInteractionsContent(
   part: ImagePart<MediaInputMetadata> | VideoPart<MediaInputMetadata>,
 ): InteractionContent {
+  // A Gemini Files API reference maps to the `uri` field, same as a public
+  // URL (mirrors the Interactions text adapter). `fileReferenceFor` throws
+  // when another provider issued the handle.
+  const sourceValue = isFileSource(part.source)
+    ? fileReferenceFor(part.source, 'gemini')
+    : part.source.value
   const mimeType = part.source.mimeType
   if (part.type === 'image') {
     return part.source.type === 'data'
-      ? { type: 'image', data: part.source.value, mime_type: mimeType }
-      : { type: 'image', uri: part.source.value, mime_type: mimeType }
+      ? { type: 'image', data: sourceValue, mime_type: mimeType }
+      : { type: 'image', uri: sourceValue, mime_type: mimeType }
   }
   return part.source.type === 'data'
-    ? { type: 'video', data: part.source.value, mime_type: mimeType }
-    : { type: 'video', uri: part.source.value, mime_type: mimeType }
+    ? { type: 'video', data: sourceValue, mime_type: mimeType }
+    : { type: 'video', uri: sourceValue, mime_type: mimeType }
 }
 
 /**
@@ -255,6 +274,9 @@ export class GeminiVideoAdapter<
   GeminiVideoModelDurationByName
 > {
   readonly name = 'gemini' as const
+  // The Interactions path consumes Gemini Files API references as content
+  // `uri`s; the Veo path still rejects them (raw bytes / gs:// only).
+  override readonly supportsFileSources = true
 
   protected client: GoogleGenAI
   private readonly allowUrlFetch: boolean

package/src/experimental/text-interactions/adapter.ts

@@ -1,4 +1,4 @@
-import { EventType } from '@tanstack/ai'
+import { EventType, fileReferenceFor, isFileSource } from '@tanstack/ai'
 import { BaseTextAdapter } from '@tanstack/ai/adapters'
 import { parse as parsePartialJSON } from 'partial-json'
 import {
@@ -200,6 +200,8 @@ export class GeminiTextInteractionsAdapter<
 > {
   override readonly kind = 'text' as const
   override readonly name = 'gemini-text-interactions' as const
+  // Consumes Gemini Files API references (geminiFiles()) as content `uri`s.
+  override readonly supportsFileSources = true
 
   private readonly client: GoogleGenAI
   // Tracks the most recent server-assigned interaction id per threadId
@@ -911,6 +913,13 @@ function contentPartToBlock(part: ContentPart): ContentBlock {
   if (part.type === 'text') {
     return { type: 'text', text: part.content }
   }
+  // A Gemini Files API handle maps to the `uri` field (isData stays false),
+  // same as a public URL. `fileReferenceFor` checks the issuer against
+  // 'gemini', the name geminiFiles() stamps on its handles, not this
+  // adapter's own name.
+  const sourceValue = isFileSource(part.source)
+    ? fileReferenceFor(part.source, 'gemini')
+    : part.source.value
   const isData = part.source.type === 'data'
   switch (part.type) {
     case 'image': {
@@ -920,8 +929,8 @@ function contentPartToBlock(part: ContentPart): ContentBlock {
         'image',
       )
       return isData
-        ? { type: 'image', data: part.source.value, mime_type }
-        : { type: 'image', uri: part.source.value, mime_type }
+        ? { type: 'image', data: sourceValue, mime_type }
+        : { type: 'image', uri: sourceValue, mime_type }
     }
     case 'audio': {
       const mime_type = validateMime(
@@ -930,8 +939,8 @@ function contentPartToBlock(part: ContentPart): ContentBlock {
         'audio',
       )
       return isData
-        ? { type: 'audio', data: part.source.value, mime_type }
-        : { type: 'audio', uri: part.source.value, mime_type }
+        ? { type: 'audio', data: sourceValue, mime_type }
+        : { type: 'audio', uri: sourceValue, mime_type }
     }
     case 'video': {
       const mime_type = validateMime(
@@ -940,8 +949,8 @@ function contentPartToBlock(part: ContentPart): ContentBlock {
         'video',
       )
       return isData
-        ? { type: 'video', data: part.source.value, mime_type }
-        : { type: 'video', uri: part.source.value, mime_type }
+        ? { type: 'video', data: sourceValue, mime_type }
+        : { type: 'video', uri: sourceValue, mime_type }
     }
     case 'document': {
       const mime_type = validateMime(
@@ -950,8 +959,8 @@ function contentPartToBlock(part: ContentPart): ContentBlock {
         'document',
       )
       return isData
-        ? { type: 'document', data: part.source.value, mime_type }
-        : { type: 'document', uri: part.source.value, mime_type }
+        ? { type: 'document', data: sourceValue, mime_type }
+        : { type: 'document', uri: sourceValue, mime_type }
     }
   }
 }

package/src/index.ts

@@ -19,6 +19,14 @@ export {
   type GeminiSummarizeModel,
 } from './adapters/summarize'
 
+// Files adapter - upload media to the Gemini Files API and reference by file URI
+export {
+  GeminiFilesAdapter,
+  createGeminiFiles,
+  geminiFiles,
+  type GeminiFilesConfig,
+} from './adapters/files'
+
 // Image adapter
 export {
   GeminiImageAdapter,