feat(ai): Add prefer_in_cloud option for inference mode #9236
Conversation
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
This change introduces a new InferenceMode option, prefer_in_cloud. When this mode is selected, the SDK will attempt to use the cloud backend first. If the cloud call fails with a network-related error, it will fall back to the on-device model if available.
🦋 Changeset detectedLatest commit: 337d34c The changes in this PR will be included in the next version bump. This PR includes changesets to release 2 packages
Not sure what this means? Click here to learn what changesets are. Click here if you're a maintainer who wants to add another changeset to this PR |
Size Report 1Affected Products
Test Logs |
Size Analysis Report 1Affected Products
Test Logs |
This commit adds a new test suite to verify that the GenerativeModel's methods correctly dispatch requests to either the on-device or cloud backends based on the selected InferenceMode. It covers generateContent, generateContentStream, and countTokens.
dlarocque
reviewed
Sep 3, 2025
Co-authored-by: Daniel La Rocque <[email protected]>
…rebase-js-sdk into feat/prefer-in-cloud
Co-authored-by: Daniel La Rocque <[email protected]>
…rebase-js-sdk into feat/prefer-in-cloud
rachelsaunders
requested changes
Sep 8, 2025
docs-devsite/ai.md
Outdated
|
|
||
| <b>(EXPERIMENTAL)</b> Determines whether inference happens on-device or in-cloud. | ||
|
|
||
| <b>PREFER\_ON\_DEVICE:</b> Attempt to make inference calls on-device. If on-device inference is not available, it will fall back to cloud. <br/> <b>ONLY\_ON\_DEVICE:</b> Only attempt to make inference calls on-device. It will not fall back to cloud. If on-device inference is not available, inference methods will throw. <br/> <b>ONLY\_IN\_CLOUD:</b> Only attempt to make inference calls to the cloud. It will not fall back to on-device. <br/> <b>PREFER\_IN\_CLOUD:</b> Attempt to make inference calls to the cloud. If not available, it will fall back to on-device. |
There was a problem hiding this comment.
Suggested change
| <b>PREFER\_ON\_DEVICE:</b> Attempt to make inference calls on-device. If on-device inference is not available, it will fall back to cloud. <br/> <b>ONLY\_ON\_DEVICE:</b> Only attempt to make inference calls on-device. It will not fall back to cloud. If on-device inference is not available, inference methods will throw. <br/> <b>ONLY\_IN\_CLOUD:</b> Only attempt to make inference calls to the cloud. It will not fall back to on-device. <br/> <b>PREFER\_IN\_CLOUD:</b> Attempt to make inference calls to the cloud. If not available, it will fall back to on-device. | |
| <b>PREFER\_ON\_DEVICE:</b> Attempt to make inference calls using an on-device model. If on-device inference is not available, the SDK will fall back to using a cloud-hosted model. <br/> <b>ONLY\_ON\_DEVICE:</b> Only attempt to make inference calls using an on-device model. The SDK will not fall back to a cloud-hosted model. If on-device inference is not available, inference methods will throw. <br/> <b>ONLY\_IN\_CLOUD:</b> Only attempt to make inference calls using a cloud-hosted model. The SDK will not fall back to an on-device model. <br/> <b>PREFER\_IN\_CLOUD:</b> Attempt to make inference calls to a cloud-hosted model. If not available, the SDK will fall back to an on-device model. |
There was a problem hiding this comment.
Addressed in source code comment and re-generated.
Co-authored-by: rachelsaunders <[email protected]>
…rebase-js-sdk into feat/prefer-in-cloud
rachelsaunders
approved these changes
Sep 8, 2025
dlarocque
approved these changes
Sep 9, 2025
Merged
mansisampat
pushed a commit
that referenced
this pull request
Sep 25, 2025
Co-authored-by: Daniel La Rocque <[email protected]> Co-authored-by: rachelsaunders <[email protected]>
MarkDuckworth
added a commit
that referenced
this pull request
Oct 10, 2025
* Add FDC Integration Tests (#8703) * Upgrade two more `download-artifact` v3 actions to v4 in Firestore workflow (#8706) There were a pair of `download-artifact` GitHub actions that were still set to `v3` instead of `v4`. The `v3` actions were unable to find the build archive in the test-changed-firestore workflow. This PR upgrades those to `v4` actions. * Version Packages (#8705) v11.2.0 Release v11.2.0 Co-authored-by: github-actions[bot] <github-actions[bot]@users.noreply.github.com> * chore(deps): replace dependency npm-run-all with npm-run-all2 5.0.0 (#8112) Co-authored-by: renovate[bot] <29139614+renovate[bot]@users.noreply.github.com> * Add App Check token to FirebaseServerApp (#8651) FirebaseServerApp now accepts an optional App Check token at initialization. The product SDKs will look for this token, and if it's present, the SDKs will use this value in lieu of calling getToken on App Check. This change affects the following SDKs: Auth, Cloud Functions, Data Connect, Firestore, Realtime Database, Vertex AI * Update rollup-plugin-terser dep (#8709) * chore(deps): update actions/github-script action to v7 (#8711) Co-authored-by: renovate[bot] <29139614+renovate[bot]@users.noreply.github.com> * Fireperf web vitals (#8644) * Add support for capturing web vitals metrics in Firebase performance for Web (Largest Contentful Paint, Interaction to Next Paint, Cumulative Layout Shift) * Modifies export to use sendBeacon instead of fetch API, and shifts the upload time to the first time the page is hidden or unloaded. * Update a lot of dependencies (#8716) * chore(config): migrate config renovate.json (#8723) Co-authored-by: renovate[bot] <29139614+renovate[bot]@users.noreply.github.com> * chore(deps): update actions/setup-node action to v4 (#8712) Co-authored-by: renovate[bot] <29139614+renovate[bot]@users.noreply.github.com> * chore(deps): update dependency firebase-tools to v13 [security] (#8607) Co-authored-by: renovate[bot] <29139614+renovate[bot]@users.noreply.github.com> * Have renovate ignore karma-webpack (#8725) * Workflow CI: Update CHROME_VALIDATED_VERSION for Auth tests (#8722) Update our CI's logged `CHROME_VALIDATED_VERSION` to the currently tested version. We haven't updated it in a while. Keeping this up to date should reduce support burden later. * chore(deps): update dependency @rollup/plugin-json to v6 (#8727) Co-authored-by: renovate[bot] <29139614+renovate[bot]@users.noreply.github.com> * Update dev dependency firebase-tools 13.6.0 -> 13.29.2 * Discard earliest heartbeat once there are 30 heartbeats (#8724) * Delete earliest heartbeats only once there are 30 heartbeats * Add changeset * Use max heartbeat const in tests * Fix test name in all caps * Fix test names * Formatting * Fix negative offset of reconnect delay (#8718) (#8719) RTDB Fix negative offset of reconnect delay. Fixes #8718. * Swap Performance's owner from jposuna to visumickey (#8737) * Give `check-version` job pull request write permissions (#8743) * Add pull_request write permissions * fix typo * Skip flaky auth recaptcha test (#8753) * Update dependencies in packages and repo-scripts (#8729) * Add ActionCodeSettings.linkDomain and deprecate ActionCodeSettings.dynamicLinkDomain (#8738) * Add linkDomain field to ActionCodeSettings (#8428) * Add linkDomain field to ActionCodeSettings * Update API reports * Update error message for ERROR_INVALID_HOSTING_LINK_DOMAIN to include that default hosting domains cannot be used. * Use constants for test values --------- Co-authored-by: NhienLam <[email protected]> * FDL Deprecation & Hosting link Integration Test (#8603) * Implement Integration test for passwordless email sign-in via firebase-hosting links. * Add license to new test file created' did not match any files * Remove unwanted _ * Add support for setting custom hosting link domain in test app (#8614) * Add afterEach method and apply formatting to the hosting links integration test (#8615) * Mark ActionCodeSettings.dynamicLinkDomain as deprecated * Add changeset * Update API reports * Address review comments * Fix changeset * Update refdocs and address review comments * Update deprecated comment * Add FDL deprecation FAQ link * Fix linkDomain link in refdocs --------- Co-authored-by: NhienLam <[email protected]> Co-authored-by: mansisampat <[email protected]> * FIX: sort strings in UTF-8 encoded byte order (#8691) * Added more granular tracking for web frameworks * fix "TextEncoder undefined" issue in some platforms (#8765) * Version Packages (#8766) Co-authored-by: github-actions[bot] <github-actions[bot]@users.noreply.github.com> * Adds a "firebase" changeset to release PR as needed (#8770) * Use Node's `TextEncoder` in E2E JSDOM Environment (#8783) * Update Firebase index.d.ts type definition file reference (#8543) Signed-off-by: Emmanuel Ferdman <[email protected]> * Revert the UTF-8 encoding in string sorting (#8782) * Version Packages (#8786) Co-authored-by: github-actions[bot] <github-actions[bot]@users.noreply.github.com> * Emulator Idempotency: Firestore (#8780) Update `connectFirestoreEmulator` to support its invocation more than once. If the Firestore instance is already in use, and `connectFirestoreEmulator` is invoked with the same configuration, then the invocation will now succeed instead of assert. The implementation takes the Data Connect implementation as inspiration. Data Connect stores the parameters passed to `connectDataConnectEmulator` on the instance of Data Connect itself, so that they can be quickly checked to see if subsequent invocations match. This PR implements a similar storage and compare process with the optional `emulatorOptions` parameter (host and port are already stored). This PR unlocks support for SSR frameworks which render the page numerous times with the same instances of Firestore. Before this PR customers were required to guard against calling `connectFirestoreEmulator` in their SSR logic, which added to code complexity. Now the Firebase SDK does that guarding logic so that our users' apps don't have to. * Emulator Idempotency: Database (#8769) Update the `connectDatabaseEmulator` function to support its invocation more than once. If the Database instance is already in use, and `connectDatabaseEmulator` is invoked with the same configuration, then the invocation will now succeed instead of assert. This unlocks support for web frameworks which may render the page numerous times with the same instances of RTDB. Before this PR customers needed to add extra code to guard against calling `connectDatabaseEmulator` in their SSR logic. Now we do that guarding logic on their behalf which should simplify our customer's apps. Fixes #6824. * SSR related improvements for RC client SDK. (#8699) SSR related improvements for RC client SDK. * Exposes an initialFetchResponse arg that pre-hydrates the client SDK state * Adds an argument that allows setting an alternate template ID to fetch from * Splits storage impls into a simple in-memory version that can be run in SSR contexts * Adds some basic tests for the API functions, which were previously untested * Emulator Idempotency: Auth (#8750) Update the `connectAuthEmulator` function to support its invocation more than once. If the Auth instance is already in use, and `connectAuthEmulator` is invoked with the same configuration, then the invocation will now succeed instead of assert. This unlocks support for web frameworks which may render the page numerous times with the same instances of auth. Before this PR customers needed to add extra code to guard against calling `connectAuthEmulator` in their SSR logic. Now we do that guarding logic on their behalf which should simplify our customer's apps. Fixes #6824. * Upgrade dependencies (#8781) * first batch * @rollup/plugin-node-resolve 13 -> 16 * Publish VertexAI API's tagged with `@beta` (#8788) * Filter empty text parts when streaming (#8736) * Filter empty text parts when streaming * Add changeset * Remove unused filterEmptyTextParts() * Move logic into a function * final * dont throw if mock response file isn't found * update responses version to 6 * Only ignore empty text parts in the aggregated response. * review fixes * Throw in `aggregateResponses` if `newPart` has no properties * Change API Documenter (BETA) annotation to (Public Preview) (#8789) * Change (BETA) annotation to (Public Preview) * format * Minor bump * Create VertexAI Node CJS and ESM bundles (#8728) * Create Node CJS and ESM bundles * Add changeset * Run formatter * register version with node * Catch errors in debug token exchange logic (#8792) * Throw error if mock response file not found (#8739) * Fix vectorvalue docs (#8453) * Fix vectorvalue docs * Use @link vector * Fix vector link with code tag * Re-enable auth redirect tests (#8803) * Re-enable redirect tests * Fix lint * Add Image Generation support using Imagen (#8683) * WIP * Introduce VertexAIModel base class, add documentation, and respond to other comments * Generate devsite docs * revert dataconnect change * formatting * WIP * Introduce VertexAIModel base class, add documentation, and respond to other comments * Generate devsite docs * revert dataconnect change * formatting * Fix ImagenRequestConfig comment * Improve tests * Formatting * Add ImagenGenerationResponse * Formatting * Update to match API changes * fixes * format * Update API reports * add changeset * update * format * Update API reports * remove public docs * Nest imageFormat in outputOptions * remove unecessary console.log * Log warning if compressionQuality outside of range * format * Imagen Documentation (#8776) * Revert "remove public docs" This reverts commit f095563. * update docs * update docs * format * minor fixes * format * mark all APIs beta instead of public * review fixes * format * Replace (BETA) with (Public Preview) in docs * fix links * another pass * format * review fixes * Cleanup * update changeset --------- Co-authored-by: dlarocque <[email protected]> * Add support for token-based usage metrics. (#8757) * Add support for token-based usage metrics. Token measurement is broken down by modaliy, with separate counters for image, audio, etc. Tests are in version 6.*, so this change also includes bumping update_responses.sh * Use right name for unspecified enum * Update API reports * Make promptTokensDetails an array * Update API reports * Add missing documentation * Add changeset entry * Fix formatting errors * Update API reports * Add missing refdoc * Update API reports * Use correct formatting for Modality documentation. * Update API reports * Update docs * Update packages/vertexai/src/types/enums.ts Co-authored-by: rachelsaunders <[email protected]> * Update md files * Revert changes to api-extractor.json --------- Co-authored-by: rachelsaunders <[email protected]> * Add Vertex and DataConnect E2E smoke tests (#8805) * Update dev script dependency @changesets/types to v6 (#8806) Co-authored-by: renovate[bot] <29139614+renovate[bot]@users.noreply.github.com> * Bump cross-spawn from 7.0.3 to 7.0.6 in /e2e (#8645) Bumps [cross-spawn](https://github.com/moxystudio/node-cross-spawn) from 7.0.3 to 7.0.6. - [Changelog](https://github.com/moxystudio/node-cross-spawn/blob/master/CHANGELOG.md) - [Commits](moxystudio/node-cross-spawn@v7.0.3...v7.0.6) --- updated-dependencies: - dependency-name: cross-spawn dependency-type: indirect ... Signed-off-by: dependabot[bot] <[email protected]> Co-authored-by: dependabot[bot] <49699333+dependabot[bot]@users.noreply.github.com> * Revert "Bump cross-spawn from 7.0.3 to 7.0.6 in /e2e (#8645)" (#8811) This reverts commit 117279e. * App Hosting JS SDK autoinit (#8483) Adds a `postinstall` step to `@firebase/util` which hardcodes autoinit defaults into `@firebase/util/dist/postinstall.(m)js` for both CJS and ESM variants [go/firestack-autoinit](http://goto.google.com/firestack-autoinit). Hardcoding is important since environment variable substitution doesn't dive into a developer's `node_modules` folder. This gives us a framework agnostic way to inject configuration and is a more robust solution than the _experimental_ autoinit methods provided by Web Frameworks [go/firebase-api-client-autoinit](https://goto.google.com/firebase-api-client-autoinit). Once this lands, we'll backport to Hosting and Functions and aim to deprecate the other autoinit methods. * Added non-null properties to database.ThenableReference (#8800) * Version Packages (#8794) Co-authored-by: github-actions[bot] <github-actions[bot]@users.noreply.github.com> * DataConnect: Update requests to point to v1 backend endpoints instead of v1beta (#8820) * update API requests to point to v1 instead of v1beta * update changelog * Remove auto-commit API report workflow (#8808) * Add missing `BlockReason` and `FinishReason` enum values (#8741) * Convert enum comments to JSDoc comments (#8742) * Add `systemInstruction`, `tools`, and `generationConfig` to `CountTokensRequest` (#8827) * Update yarn.lock in e2e (#8831) * Fix transport service retry logic condition (Fixes #8813) (#8814) Modify the retry mechanism to stop when remaining tries is less than or equal to zero, improving the robustness of the retry handling. Fixes #8813 * Give `check-changeset` job pull request write permissions (#8833) * Fix: sort strings in UTF-8 encoded byte order with lazy encoding (#8787) * fix golden files test (#8843) * fix golden files test * update responses version --------- Co-authored-by: David Motsonashvili <[email protected]> * Surface all App Check errors in public getToken() method (#8842) * Grant Create PR workflow pull-request write perms. (#8846) Our staging release workflow failed to create a Version Packages PR. Grant the PR creation workflow `pull-request` `write` permissions. * Grant Create PR content write permissions. (#8848) Our staging release workflow failed to merge branch. Grant the PR creation workflow `content` `write` permissions to fix this. * Version Packages (#8849) - v11.5.0 Release version 11.5.0. Co-authored-by: github-actions[bot] <41898282+github-actions[bot]@users.noreply.github.com> * Deprecate `GroundingAttribution` (#8832) * Stream full logs of test-all firestore job (#8851) * Set github actions to use commit hashes (#8852) * Implement Partial Errors for FDC (#8821) * Add `X-Firebase-AppId` header to VertexAI requests (#8809) * Update dependency webpack to v5.94.0 [SECURITY] (#8734) Co-authored-by: renovate[bot] <29139614+renovate[bot]@users.noreply.github.com> * Update dependency express to v4.20.0 [SECURITY] (#8759) Co-authored-by: renovate[bot] <29139614+renovate[bot]@users.noreply.github.com> * karma.base.js: replace "BROWSER" with "BROWSERS" in log messages, to reflect the actual name of the environment variable (see #8491) (#8875) * Auth cookie persistence (#8839) Adding `Persistence.COOKIE` a new persistence method backed by cookies. The `browserCookiePersistence` implementation is designed to be used in conjunction with middleware that ensures both your front and backend authentication state remains synchronized. * firestore: fix `undefined` document snapshot data after "clear site data" (#8871) * Use base browsers variable in Firestore karma config (#8876) * [chore] Update some dependencies (#8868) * Version Packages (#8878) Co-authored-by: github-actions[bot] <41898282+github-actions[bot]@users.noreply.github.com> * Fix 'window is not defined' error when calling clearIndexedDbPersistence * Fix Data Connect Types (#8898) * test(vertexai): update mock responses to v8 (#8904) * Fix: display WebChannel error message (#8907) * test(vertexai): add `backendName` param to mock response getters (#8906) `backendName` can be either `googleAI` or `vertexAI`. This can be passed to `getMockResponse` or `getMockResponseStreaming` to specify whether the mock response lookup should be done in the set of vertexAI or googleAI mock files. Modified the `convert-mocks.ts` script to read mock responses from the 'developerapi' directory, and add an export to the generated file for the new lookup object with those mock responses. * fix: remove `null` value inclusion from `!=` and `not-in` filter results (#8915) * Add assertion IDs that will be included in production log statements for fail and hardAsserts (#8313) * docs(vertexai): remove HTML `<code>` tags (#8900) We wrap links that reference public APIs with `<code>` tags so that they can be rendered as code in devsite. Unfortunately, VSCode's hover feature can not render these links at all. Since there are no other alternatives for creating code-formatted links, we should remove the problematic HTML `<code>` tags. * test(vertexai): update developerapi mock response dir to googleai (#8921) * Fix languageCode parameter in action_code_url (#8912) * Fix languageCode parameter in action_code_url * Add changeset * test(vertexai): update mock responses to v10 (#8959) * Version Packages (#8956) Co-authored-by: github-actions[bot] <41898282+github-actions[bot]@users.noreply.github.com> * Improve the integration test coverage for online vs offline comparisons. (#8975) * Improve the integration test coverage for online vs offline comparisons. * prettier. * Add missing cases. * fix(vertexai): add missing quote to chat role error message (#8979) * Auto Enable SSL for Firebase Studio (#8980) * Add support for running the emulators in Cloud Workstation (#8968) * Add Cookie Support For Firebase Studio (#8986) * Fix Auth Port for Firebase Studio (#8998) * Change automaticDataCollectionEnabled to default to true. (#8999) * Version Packages (#8997) Co-authored-by: github-actions[bot] <41898282+github-actions[bot]@users.noreply.github.com> * Roll back #8999 (#9007) * Revert "Change automaticDataCollectionEnabled to default to true. (#8999)" This reverts commit 3789b5a. * Revert #8999 - backend is not ready * Version Packages (#9008) Co-authored-by: github-actions[bot] <41898282+github-actions[bot]@users.noreply.github.com> * fix(vertexai): pass `GenerativeModel`'s `BaseParams` to `ChatSession` (#8972) startChat wasn't passing GenerativeModel's BaseParams to ChatSession. So, if a model had a generationConfig, it would never be passed to ChatSession. * feat(vertexai): Gemini multimodal output (#8922) Adds new ResponseModality enum that allows users to specify which modalities should be included in a response. Since we provide a text() accessor, a similar inlineDataParts() accessor was added to return all InlineDataPart[] in the first candidate. * test: Clean up vscode launch file and add vertexai debug config (#9000) * Fix encoding of CSI keys in Safari (#8993) * Restore #8999 (#9010) * Revert "Roll back #8999 (#9007)" This reverts commit 51e7b48. * Changeset * feat(ai): Migrate to `AI` and add GoogleAI support (#8931) * Add Emulator Overlay (#8977) * Rename `@firebase/vertexai` package to `@firebase/ai` (#9011) * Add Emulator Overlay (#8977) (#9031) Co-authored-by: Maneesh Tewani <[email protected]> * add temporary release branch (#9032) * Allow custom branch for release (#9033) * add temporary release branch * restore Vertex token for temp publish (#9036) * Version Packages (manual PR) (#9034) * Fixed scroll behavior (#9043) * Version Packages (manual PR) (#9044) * Remove redundant changeset * Update e2e tests to reflect AI package and new API (#9017) * Version Packages (#9016) Co-authored-by: github-actions[bot] <41898282+github-actions[bot]@users.noreply.github.com> * Revert "Fixed scroll behavior (#8977)" (#9053) * Revert "Fixed scroll behavior (#9043)" This reverts commit 9bcd1ea. * Revert "Add Emulator Overlay (#8977) (#9031)" This reverts commit 8593fa0. * Undid revert of util * Create dirty-crews-cross.md * Removed unnecessary markdown file * Updated to only show banner when calling connect*Emulator (#9055) * Version Packages (#9054) Co-authored-by: github-actions[bot] <41898282+github-actions[bot]@users.noreply.github.com> * Close webchannel instances on terminate (#9041) Explicitly close any unclosed WebChannel instances on Firestore#terminate() * Reorder reporters to prevent log message duplication in CI (#9040) * docs(ai): Fix 'occured' typo in error codes (#9021) * feat(ai): Add `HarmBlockThreshold.OFF` (#9050) * feat(ai): Schema `title`, `maximum`, `minimum`, and `propertyOrdering` (#9047) * feat(ai): Add support for `minItems` and `maxItems` to `Schema` (#9026) * Build out e2e subdirectories (#9065) * Added test project to fdc (#8924) * Version Packages (#9073) Co-authored-by: github-actions[bot] <41898282+github-actions[bot]@users.noreply.github.com> Release version 11.9.0. * test(ai): add integration tests (#8853) * Update issue template to rename VertexAI > AI (#9074) * ci: Fix Firefox error and streamline cross-browser tests (#9015) * Fixed issue where requestSts wasn't including the Studio cookie in it (#9075) * Fix issue where we ping an endpoint that hasn't been implemented (#9059) * Updated firebase auth changeset (#9088) * build(all): Version Packages v11.9.1 (#9089) Co-authored-by: github-actions[bot] <41898282+github-actions[bot]@users.noreply.github.com> v11.9.1 release. * test(ai): Upgrade mock responses to v14 (#9091) * Feat/google3 typings support (#9085) * fix(ai): Add deprecation tag to `totalBillableCharacters` (#9090) * test(ai): Add `gemini-2.5-flash` to integration tests (#9110) Added gemini-2.5-flash to our integration tests now that it's publicly available. The token counts differ slightly between 2.0-flash and 2.5-flash so I introduced conditionals when checking token counts. * firestore: minor refactor of listener registration of "versionchange" indexedb events (#9087) * Fixed issue where storage wasn't passing cookies when using firebase Studio (#9111) * fix: Add "react-native" entry point to @firebase/app (#9112) * Feat(Firestore) JSON serialization of types to improve SSR support. (#8926) Support the ability to resume `onSnapshot` listeners in the CSR phase based on serialized `DataSnapshot`s and `QuerySnapshot`s built in the SSR phase. Allow Firestore result types to be serialized with `toJSON` and then deserialized with `fromJSON` methods on the objects. `DocumentSnapshot` and `QuerySnapshot` deserialization methods will be standalone, tree-shakable functions `dataSnapshotFromJSON` and `querySnapshotFromJSON`. * chore(Firestore) add changeset for PR #8926 (#9115) Add the missing changset for PR #8926. * Update Fireperf logging to use sendBeacon only if the payload is under the 64KB limit (#9120) * Update Fireperf logging to use sendBeacon only if the payload is under the 64KB limit for most browsers. - For the flush, attempt to use sendBeacon with a low number of events incase sendBeacon is also used by other libraries. * Add changeset and fix format * Add additional comments * Put max flush size behind remote config flag * Version Packages (#9126) Co-authored-by: github-actions[bot] <41898282+github-actions[bot]@users.noreply.github.com> * Increase unit test coverage for facebook.ts to 100% (#9129) * Increase unit test coverage for facebook.ts to 100% * Increase unit test coverage for github.ts to 100% (#9130) * Increase unit test coverage for github.ts to 100% * test: Use inline source maps for webpack test output (#9109) * Increase unit test coverage for saml.ts to 100% (#9131) * Increase unit test coverage for saml.ts to 100% * firestore: misc.ts: further improved performance of UTF-8 string comparison logic * fix(ai): typings for functionDeclaration.parameters. (#8948) * feat(ai): add support for grounding with google search (#9068) * feat!(ai): Remove `vertexai` import alias (#9081) * docs(firestore): add an example to the detailed comment in compareUtf8Strings() (#9154) * feat(ai): Add support for Thinking Budget (#9156) * feat!(ai): remove `VertexAI` APIs (#9063) * fix!(ai): convert TS enums exports in Firebase AI into const variables (#9137) * fix(firestore): revert fix for `undefined` document snapshot data after "clear site data" (#9162) This reverts #8871 because it had unexpected negative customer impacts (#9056) and caused CI failures. * build!: update build targets to ES2020 (#9101) * (feat) initializeServerApp support for App Hosting auto init (#9151) Implement Auto Init for `initializeServerApp`. Auto init was previously implemented for `initializeApp` in #8483. This PR adds the same functionality to the `initializeServerApp` API surface. Fixes #8863 * feat!(ai): Remove `GroundingAttribution` (#9079) * feat!(ai): add support for `anyOf` schemas (#9042) * chore!: update engines.node to minimum of 20 (#9128) * fix(ci): Update e2e tests to import from ai instead of vertexai (#9164) * chore: Add missing changeset for 9101 (#9167) * Revert "firestore: minor refactor of listener registration of "versionchange" indexedb events (#9087) (#9168) * Revert "firestore: minor refactor of listener registration of "versionchange" indexedb events (#9087)" This reverts commit f73e08b. * Create hungry-icons-dream.md * docs: fix doc typos found by g3 presubmit (#9166) * Version Packages (#9165) Co-authored-by: github-actions[bot] <41898282+github-actions[bot]@users.noreply.github.com> * Fixes a bug where the out-of-the-box core web vital metrics would throw an error when the target element name exceeds the max length of a custom attribute value. (#9178) * fix(app-check): Prevent redundant exchangeToken calls in debug mode (#9187) * Update BrowsersList DB (#9188) * Firebase AI Hybrid Inference Implementation (#9029) Co-authored-by: Daniel La Rocque <[email protected]> Co-authored-by: Erik Eldridge <[email protected]> Co-authored-by: Christina Holland <[email protected]> * Version Packages (#9195) Co-authored-by: github-actions[bot] <41898282+github-actions[bot]@users.noreply.github.com> * Backfill fixes for hybrid docs (#9197) * Fix Firebase Functions Emulator usage on Firebase Studio (#9204) * Used valid path for pingserver * Added test * Create three-balloons-collect.md * Fixed fromUrl * Removed only * Fixed typings * fix(firestore): fix empty message reject inside transaction body (#9177) * Add limitedUseToken option to AI SDK (#9201) * [Firebase AI] Add thought summary and signature support (#9192) * fix(ai): Only include unexpected `prediction` element in error message (#9222) * fix(ai): Only include unexpected prediction element in error message. If the `responseJson` from the Predict request has an image, the entire image base64 will be included in the error message. This can make the error message >400kB (!!), which could cause performance issues in apps. I ran into this when testing #9216, since the `safetyAttributes` would be included in a response and trigger this error, but the `predictions` array in `responseJson` would have images. I noticed that when the sample app rendered the `error.message`, it began running into serious performance issues. * fix(ai): Add `'includeSafetyAttributes'` to Predict request payloads (#9216) * fix(ai): keep window-specific code out of Node bundle (#9225) * feat(ai): add support for the Live API (#9224) * Version Packages (#9230) Co-authored-by: github-actions[bot] <41898282+github-actions[bot]@users.noreply.github.com> * remove accidental factory export (#9232) * Version Packages (#9233) Co-authored-by: github-actions[bot] <41898282+github-actions[bot]@users.noreply.github.com> * fix(ai): Fix test script broken by #9232 (#9235) * Update CHANGELOG.md (#9237) * feat(ai): Add prefer_in_cloud option for inference mode (#9236) Co-authored-by: Daniel La Rocque <[email protected]> Co-authored-by: rachelsaunders <[email protected]> * Remote Config Realtime for Web (#9221) * feat(ai): Add code execution (#9249) * chore(firestore): Update closure-net dependency to the latest release (#9242) * Update closure-net dependency to the latest release (6f48f578d3e80fe7a85e530a5d95b9351433d135) * Change documentation tags for hybrid inference from EXPERIMENTAL to public preview (#9253) * docs: Fix docgen script to clean out package temp folders first (#9250) * Version Packages (#9258) Co-authored-by: github-actions[bot] <41898282+github-actions[bot]@users.noreply.github.com> * Expose setUserProperties from internal analytics instance (#9263) * feat(ai): add support for URL context (#9254) * docs(ai): tag code execution with "beta" tag (public preview) (#9274) * fix(ai): Handle empty parts when streaming (#9262) * feat(ai): Add method to send function responses in a live session (#9272) * Update ai model name used for e2e tests (#9282) * docs(ai): promote gemini developer api and imagen to GA (#9255) * Version Packages (#9300) release v12.4.0 For JS SDK release v12.4.0 Co-authored-by: github-actions[bot] <41898282+github-actions[bot]@users.noreply.github.com> * remove emulator files that were readded * Update util.api.md --------- Signed-off-by: Emmanuel Ferdman <[email protected]> Signed-off-by: dependabot[bot] <[email protected]> Co-authored-by: Maneesh Tewani <[email protected]> Co-authored-by: DellaBitta <[email protected]> Co-authored-by: Google Open Source Bot <[email protected]> Co-authored-by: github-actions[bot] <github-actions[bot]@users.noreply.github.com> Co-authored-by: DellaBitta <[email protected]> Co-authored-by: renovate[bot] <29139614+renovate[bot]@users.noreply.github.com> Co-authored-by: Christina Holland <[email protected]> Co-authored-by: Bryan Atkinson <[email protected]> Co-authored-by: Daniel La Rocque <[email protected]> Co-authored-by: Alexander <[email protected]> Co-authored-by: jposuna <[email protected]> Co-authored-by: Nhien (Ricky) Lam <[email protected]> Co-authored-by: NhienLam <[email protected]> Co-authored-by: mansisampat <[email protected]> Co-authored-by: Mila <[email protected]> Co-authored-by: Stephen Rosa <[email protected]> Co-authored-by: Emmanuel Ferdman <[email protected]> Co-authored-by: Kevin Elko <[email protected]> Co-authored-by: dlarocque <[email protected]> Co-authored-by: Rodrigo Lazo <[email protected]> Co-authored-by: rachelsaunders <[email protected]> Co-authored-by: dependabot[bot] <49699333+dependabot[bot]@users.noreply.github.com> Co-authored-by: James Daniels <[email protected]> Co-authored-by: Rich Hodgkins <[email protected]> Co-authored-by: Leo <[email protected]> Co-authored-by: David Motsonashvili <[email protected]> Co-authored-by: David Motsonashvili <[email protected]> Co-authored-by: github-actions[bot] <41898282+github-actions[bot]@users.noreply.github.com> Co-authored-by: Denver Coneybeare <[email protected]> Co-authored-by: Pavan Shankar <[email protected]> Co-authored-by: Ehsan <[email protected]> Co-authored-by: eranc-google <[email protected]> Co-authored-by: Raymond Lam <[email protected]> Co-authored-by: Raymond Lam <[email protected]> Co-authored-by: Nicole <[email protected]> Co-authored-by: gsiddh <[email protected]> Co-authored-by: Erik Eldridge <[email protected]> Co-authored-by: cherylEnkidu <[email protected]> Co-authored-by: Andrew Heard <[email protected]> Co-authored-by: Daymon <[email protected]> Co-authored-by: Tushar Khandelwal <[email protected]>
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
Add prefer_in_cloud option - opposite of prefer_on_device.
Note: Used gemini-cli
Had it add tests to generative-model.test.ts instead of on generateContent/generateContentStream/countTokens for all 4
InferenceModecases because I think the more of the pipeline it tests, the better it is - a lot of errors come from omitting or incorrectly passing params through multiple functions.Had it extract the logic for on-device vs cloud into a helper function in helpers.ts even though it is only used in generate-content.ts for now, because we'll need to re-use the same logic for countTokens when we implement that (Chrome API should be ready now).
Docs review: Only need to review
docs-devsite/ai.md.