搜尋套件、儲存庫與 APP 分類

套件詳情v1.8.6

flutter_gemma_litertlm

LiteRT-LM (.litertlm) 是用於 flutter_gemma 的裝置端推論引擎,透過 dart:ffi(5 個原生平台)+ web 實作。可選擇性使用 InferenceEngineProvider。同時提供 LiteRT C API 嵌入式後端(LiteRtEmbeddi [...]

AndroidiOSLinuxmacOSWebWindows
近 30 天下載
3萬
喜歡數
0
Pub 分數
130

安裝套件

flutter pub add flutter_gemma_litertlm

套件介紹

根據套件公開資料產生的選型評估,僅供參考。

一個輕量級的本機推理引擎,適用於FlutterGemma,使用Dart:ffi,支援五個原生平台與Web。透過可選的提供者整合,利用LiteRT C API嵌入後端實現高效能的本機大型語言模型推理。專為低延遲、注重隱私的生成式AI應用設計,可直接在裝置上執行,無需依賴雲端服務。此套件簡化了在Android、iOS、Linux、macOS、Windows及Web環境中的部署。

適用情境

  • 設備端語言模型推論
  • 保護隱私的聊天應用
  • 離線AI驅動的應用程式
  • 用於大語言模型的邊緣運算
  • 輕量級 LLM 移動與桌面端部署

主要特色

  • 原生平台支援(5+ 平台)
  • 透過 Dart:ffi 實現 Web 相容性
  • 選擇性啟用 InferenceEngineProvider
  • LiteRT C API 嵌入式後端
  • 最小的執行時開銷
  • 跨平台一致性

適合對象

  • 開發離線大型語言模型應用的開發者
  • 需要資料隱私的應用程式
  • 行動與桌面AI原型
  • 資源有限的邊緣AI專案
  • 使用 Flutter 進行生成式 AI 的團隊

使用建議

  • 需要原生程式碼編譯
  • 僅限與 LiteRT 相容的模型
  • Web 支援取決於 WASM 的可用性
  • 效能因裝置硬體而異
  • 無內建的模型管理或快取

原始 README

以下為關聯原始碼儲存庫的 README 原文快照,最新內容請造訪原始碼儲存庫。

Flutter Edge AI

pub package License: MIT

On-device AI for Flutter on Android, iOS, Web, macOS, Windows and Linux. Run any model a registered engine runs, or the model the operating system already ships, with text, vision, audio, function calling, embeddings, RAG and speech. No server, no cloud.

Website · Docs · Models · Codelabs · pub.dev

The core package is small: it installs models and hands them to engines. Every runtime, store and feature is a separate opt-in package, so an app ships only what it uses. Start with the core README or Getting started.

Packages

Package What it does Docs
flutter_edge_ai Core: model installation, the engine registry, sessions and chat. Required by every package except diagnostics and genkit_hybrid. pub.dev · site
Engines
flutter_edge_ai_litertlm LiteRT-LM (.litertlm) through dart:ffi on the five native platforms plus web; CPU, GPU and NPU; the LiteRT embedding backend. pub.dev · site
flutter_edge_ai_mediapipe MediaPipe (.task) on Android, iOS and web. Google's LLM Inference API is maintenance-only; new apps should use .litertlm. pub.dev · site
flutter_edge_ai_llama llama.cpp (.gguf) through llamadart on Android, iOS, macOS, Windows, Linux and web; native LoRA. pub.dev · site
flutter_edge_ai_onnx ONNX Runtime: GenAI inference and embeddings natively, Transformers.js and onnxruntime-web on web. pub.dev · site
flutter_edge_ai_builtin_ai The OS model, nothing to download: Gemini Nano (Android), Apple Foundation Models (iOS, macOS), Windows AI Foundry, the browser Prompt API. pub.dev · site
Embeddings and RAG
flutter_edge_ai_embeddings Embedding tokenizers (Gemma SentencePiece, BERT WordPiece) for .tflite and .onnx embedders. pub.dev · site
flutter_edge_ai_rag On-device RAG: RagIndex, embedding profiles, filters and the vector-store contract. pub.dev · site
flutter_edge_ai_sqlite SQLite (sqlite-vec) vector store on every platform, web included. pub.dev · site
flutter_edge_ai_qdrant Qdrant Edge vector store on native platforms. pub.dev · site
Speech, agents, diagnostics
flutter_edge_ai_speech Speech-to-text, text-to-speech and the VoiceSession voice loop. pub.dev · site
flutter_edge_ai_agent Agent skills (SKILL.md) the model picks and runs: text, JS, native intents and MCP tools. pub.dev · site
flutter_edge_ai_diagnostics Process memory diagnostics on Android, iOS, macOS, Linux and Windows. pub.dev · site
Genkit
genkit_flutter_edge_ai Genkit Dart plugin: on-device models and embedders as Genkit models. pub.dev · site
genkit_hybrid Hybrid routing for Genkit: on-device and cloud models behind one routing policy. Depends only on genkit. pub.dev · site

Which package you need for which task, and how they fit together: Packages and Capabilities.

Maintainers

Maintainer Looks after
Sasha Denisov (@DenisovAV) Lead maintainer: the core API, the engines, the native builds and every release
Jhin Lee (@leehack) flutter_edge_ai_llama
Francesco Vezzani (@kekko7072) flutter_edge_ai_builtin_ai
Samarthsinh Vala (@samarthsinh2660) flutter_edge_ai_diagnostics
Daisuke Majima (@john-rocky) The Hugging Face integration: installing a model straight from its repository

Everyone else who has contributed, and more about each maintainer: flutteredge.ai/contributors.

Repository

More: Migration guide · Desktop support · Contributing. The family shipped as flutter_gemma* up to core 1.11.4; those names stay published as they are.

License

MIT — see LICENSE.

所屬分類

依賴

依賴清單 7 項

原始碼儲存庫

使用此套件的專案

暫無相關資料。

所在榜單