feat: unify AI providers with shared abstractions in expo-example #179
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
- Consolidated Llama and MLC providers under Apple screen with provider selector - Removed standalone Llama and MLC screens - Created shared configuration for models (LLAMA_MODELS, SPEECH_LLAMA_MODELS) - Implemented reusable components: - ChatUI: Provider-agnostic chat interface - ProviderSetup: Language model download and initialization - SpeechProviderSetup: Speech model with vocoder setup - ProviderSelector: Provider switching UI component - Upgraded to AI SDK v6 with LanguageModelV3 and SpeechModelV3 - Removed unnecessary dynamic imports (Llama available on all platforms) - Added Llama speech support with OuteTTS model - Fixed code duplication and inconsistencies - Platform-specific implementations for iOS and Android
|
The latest updates on your projects. Learn more about Vercel for GitHub.
|
grabbou
changed the base branch from
main
to
fix/streaming-first-char-missing
grabbou
changed the title
- Replace separate repo/filename fields with single modelId format - Replace nested vocoder object with vocoderId field - Remove redundant mmproj and vocoder.size fields - Update SpeechProviderSetup to use new interface with getFilename helper
Contributor
Author
Not the sexiest UI out there, but at least we can do quite a lot w/o hardcoding providers anymore! |
Contributor
There was a problem hiding this comment.
Pull request overview
This PR refactors the expo-example app to consolidate multiple AI providers (Apple, Llama, and formerly MLC) under a unified architecture with shared abstractions and reusable components. The changes include:
Changes:
- Created a provider abstraction layer with SetupAdapter interface for unified model management (download, initialization, deletion)
- Upgraded dependencies:
llama.rnfrom 0.10.0-rc.0 to 0.10.1 andaipackage from 5.0.56 to 6.0.0 - Removed MLC provider and consolidated screens into provider-agnostic implementations
- Added llama.rn plugin to Expo configuration (previously missing)
- Exported additional storage utilities (ModelInfo, getDownloadedModels, removeModel) from llama package
Reviewed changes
Copilot reviewed 24 out of 25 changed files in this pull request and generated 11 comments.
Show a summary per file
| File | Description |
|---|---|
| packages/llama/src/index.ts | Exports additional storage utilities for model management |
| packages/llama/src/ai-sdk.ts | Updates speech model generation with guide tokens and improved audio handling |
| packages/llama/package.json | Upgrades llama.rn to 0.10.1 |
| apps/expo-example/src/config/providers.ts | Defines provider abstraction layer with SetupAdapter interface |
| apps/expo-example/src/components/adapters/*.ts | Implements adapters for Apple and Llama providers |
| apps/expo-example/src/components/ProviderSelector.tsx | New component for provider selection |
| apps/expo-example/src/components/ProviderSetup.tsx | New component for model download/setup |
| apps/expo-example/src/components/ChatUI.tsx | Refactored to accept any language model and tools |
| apps/expo-example/src/screens/apple/ChatScreen/index.tsx | New unified chat screen supporting multiple providers |
| apps/expo-example/src/screens/apple/PlaygroundScreen/index.tsx | Converted from iOS-specific to cross-platform with provider support |
| apps/expo-example/src/screens/apple/SpeechScreen/index.tsx | Converted from iOS-specific to cross-platform with provider support |
| apps/expo-example/src/screens/apple/TranscribeScreen/index.tsx | Converted from iOS-specific to cross-platform |
| apps/expo-example/src/App.tsx | Updated navigation to use unified screens and removed MLC/Llama tabs |
| apps/expo-example/polyfills.ts | Added DOMException polyfill |
| apps/expo-example/package.json | Removed MLC dependency, upgraded ai and llama.rn versions |
| apps/expo-example/app.json | Changed plugin from @react-native-ai/mlc to llama.rn |
| bun.lock | Updated lockfile with new dependency versions |
Comments suppressed due to low confidence (2)
apps/expo-example/src/components/ChatUI.tsx:67
- The removed line 'setMessages(updatedMessages)' means the user's message is no longer added to the UI immediately after sending. The message will only appear after the assistant's placeholder message is added on line 61-67. This creates a delay in showing the user's own message in the chat interface. Consider adding back 'setMessages(updatedMessages)' or updating line 61 to show the user message immediately.
apps/expo-example/src/components/ChatUI.tsx:75 - The tools prop defaults to an empty object but is then checked with 'tools ?? undefined'. When tools is an empty object (the default), it will be passed to streamText as an empty object rather than undefined. This could cause unexpected behavior. Consider changing the default to 'undefined' in the props interface and removing the '?? undefined' check, or checking if the object is empty before passing it.
💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.
apps/expo-example/src/components/adapters/llamaSpeechSetupAdapter.ts
Outdated
Show resolved
Hide resolved
Co-authored-by: Copilot <175728472+Copilot@users.noreply.github.com>
…ter.ts Co-authored-by: Copilot <175728472+Copilot@users.noreply.github.com>
Co-authored-by: Copilot <175728472+Copilot@users.noreply.github.com>
grabbou
added a commit
that referenced
this pull request
Jan 24, 2026
* refactor(expo-example): unify AI providers with shared abstractions - Consolidated Llama and MLC providers under Apple screen with provider selector - Removed standalone Llama and MLC screens - Created shared configuration for models (LLAMA_MODELS, SPEECH_LLAMA_MODELS) - Implemented reusable components: - ChatUI: Provider-agnostic chat interface - ProviderSetup: Language model download and initialization - SpeechProviderSetup: Speech model with vocoder setup - ProviderSelector: Provider switching UI component - Upgraded to AI SDK v6 with LanguageModelV3 and SpeechModelV3 - Removed unnecessary dynamic imports (Llama available on all platforms) - Added Llama speech support with OuteTTS model - Fixed code duplication and inconsistencies - Platform-specific implementations for iOS and Android * refactor: simplify SpeechModelOption to match ModelOption shape - Replace separate repo/filename fields with single modelId format - Replace nested vocoder object with vocoderId field - Remove redundant mmproj and vocoder.size fields - Update SpeechProviderSetup to use new interface with getFilename helper * chore: some other tweaks to get proper version * feat: updates to the playground app * chore: complete updates to the UI * tweaks * Update apps/expo-example/src/screens/apple/SpeechScreen/index.tsx Co-authored-by: Copilot <175728472+Copilot@users.noreply.github.com> * Update apps/expo-example/src/components/adapters/llamaSpeechSetupAdapter.ts Co-authored-by: Copilot <175728472+Copilot@users.noreply.github.com> * Update apps/expo-example/src/components/ProviderSetup.tsx Co-authored-by: Copilot <175728472+Copilot@users.noreply.github.com> * chore: get rid --------- Co-authored-by: Copilot <175728472+Copilot@users.noreply.github.com>
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.
This PR refactors the
expo-exampleapp to consolidate multiple AI providers (Apple, Llama, MLC) under a unified architecture with shared abstractions and reusable components. It also upgrades Llama to latest and updates Expo configuration (previously, it was missing a plugin).The goal is to expose all llama features from within same UI and have a runaway to test other on-device providers as well. The UI needs a lot of love for sure, but let's focus on functionality at this point.
UX is definitely not good, and this is meant for internal testing mostly. Will improve in follow-up PRs.
Testing Checklist