前廊AI
AI Skill Hub 推荐使用：前廊AI 是一款优质的AI工具。AI 综合评分 7.5 分，在同类工具中表现稳健。如果你正在寻找可靠的AI工具解决方案，这是一个值得深入了解的选择。
📚 深度解析
**为什么要使用开源工具而非商业 SaaS？**
对于个人开发者和有隐私需求的用户，本地部署的开源工具意味着数据不离本机，不受第三方服务商的数据政策约束。同时，开源工具通常没有使用次数限制和月度费用，一次安装即可长期使用，对于高频使用场景的总拥有成本（TCO）远低于订阅制商业工具。
**安装与环境准备**
前廊AI 依赖 Dart 运行环境。建议通过 pyenv（Python）或 nvm（Node.js）管理 Dart 版本，避免全局环境污染。对于新手用户，推荐先创建虚拟环境（python -m venv venv && source venv/bin/activate），再安装依赖，这样即使出现问题也可以随时删除虚拟环境重新开始，不影响系统稳定性。
**社区与维护**
GitHub Issue 和 Discussion 是获取帮助的最快渠道。在提问前建议先检查 Closed Issues（已关闭的问题），大多数常见问题都已有解答。遇到 Bug 时，提供 pip list 的输出、完整错误堆栈和最小可复现示例，能显著提高开发者响应速度。AI Skill Hub 将持续追踪 前廊AI 的版本更新，及时通知重要功能变化。
📋 工具概览
前廊AI 是一款基于 Dart 开发的开源工具，专注于 ai、ai-companion、character-ai 等核心功能。作为 GitHub 开源项目，它拥有活跃的社区支持和持续的版本迭代，代码完全透明可审计，支持本地部署以保护数据隐私。无论是个人使用还是集成到企业工作流，都能提供稳定可靠的解决方案。
📖 中文文档
前廊AI 是一款基于 Dart 开发的开源工具，专注于 ai、ai-companion、character-ai 等核心功能。作为 GitHub 开源项目，它拥有活跃的社区支持和持续的版本迭代，代码完全透明可审计，支持本地部署以保护数据隐私。无论是个人使用还是集成到企业工作流，都能提供稳定可靠的解决方案。
- 开源免费，支持本地部署，数据完全自主可控
- 活跃的 GitHub 开源社区，持续迭代更新
- 提供详细文档和使用示例，新手友好
- 支持自定义配置，灵活适配不同使用环境
- 可作为基础组件集成进现有技术栈或进行二次开发
- 本地部署运行，保护数据隐私，满足合规要求
- 自定义集成到现有系统，扩展技术栈能力
- 作为开源基础组件进行商业化二次开发
# 克隆仓库 git clone https://github.com/linux4life1/front-porch-AI cd front-porch-AI # 查看安装说明 cat README.md # 按 README 完成环境依赖安装后即可使用
- 访问 GitHub 仓库页面
- 按照 README 文档完成依赖安装
- 根据系统环境完成初始化配置
- 参考官方示例或文档开始使用
- 遇到问题可在 GitHub Issues 中查找解答
# 查看帮助 front-porch-ai --help # 基本运行 front-porch-ai [options] <input> # 详细使用说明请查阅文档 # https://github.com/linux4life1/front-porch-AI
# front-porch-ai 配置说明 # 查看配置选项 front-porch-ai --config-example > config.yml # 常见配置项 # output_dir: ./output # log_level: info # workers: 4 # 环境变量（覆盖配置文件） export FRONT_PORCH_AI_CONFIG="/path/to/config.yml"
Front Porch AI
A privacy-first AI companion for Windows, Linux, and macOS. Runs fully offline with local LLMs (KoboldCpp, etc.) by default, but also supports remote APIs like OpenRouter, Nano-GPT, and OpenAI with no lock-in when you want them.
💬 Join the Discord — questions, feedback, and hanging out welcome. Also on Matrix.
<p align="center"> <img src="docs/screenshots/home_new.png" width="800" alt="Front Porch AI — Character Library"> </p>
---
<p align="center"> <strong>Download v0.9.8</strong><br><br> <a href="https://github.com/linux4life1/front-porch-ai/releases/latest"><strong>Windows • macOS • Linux</strong></a> </p>
---
✨ Features
🆕 What's New in v0.9.8
v0.9.8 is a substantial release focused on making the app feel more alive and reliable. The headline feature is Character Expressions, but the release also delivers major maturation of the Realism Engine, a much more robust local TTS experience, .kcpps preset support, custom chat backgrounds, and dozens of quality-of-life and stability improvements.
🎭 Character Expressions — Live Emotion Portraits
Your characters now look the way they feel. As the conversation evolves, their portrait automatically swaps to match their current emotional state.
- Dual classification engine: Toggle between a fast local ONNX path (distilbert-based, ~300 ms) and the deeper LLM path via the Realism Engine. Both run entirely on-device.
- One-click model download: Download the ONNX classifier directly from Settings → Expression Images with a beautiful glassmorphic progress overlay.
- SillyTavern compatible: Works with any standard expression pack (26 emotion categories supported).
- Flexible presentation: Sidebar mode for focused chats or fullscreen cinematic overlay.
- Graceful fallback: Falls back cleanly to neutral if an image is missing.
🧠 Realism Engine – Major Maturation
The Realism Engine received its most significant round of refinements to date:
- Bond and Trust ranges expanded to ±300 with updated tier naming to match the character creator.
- Arousal system expanded to ±100 with new tier-based labels.
- Improved spatial awareness logic and better behaviour when "passage of time" is disabled.
- Realism evaluations are now more reliable on thinking models (higher token limits, hardened JSON generation parameters).
- GBNF grammar disabled for KoboldCpp realism evals (dramatically improves completion rates on many models).
- Much more robust cancellation handling — interrupting a response now cleanly aborts in-flight realism evaluations.
- Better one-shot eval behaviour for remote APIs and improved tooltip explanations in the UI.
🗣️ Voice & Narration (Kokoro TTS)
Local voice output is now significantly more reliable and pleasant:
- Persistent Kokoro worker pool — enables fast, high-quality "read everything" (verbatim) narration without the previous stuttering or slow startup.
- "Only narrate quotes" mode is now much more dependable.
- Improved local bundling of both Kokoro and Piper engines.
- Better concurrency controls and pre-load behaviour.
⚙️ .kcpps Presets & Context Management
- Full support for loading KoboldCpp
.kcppslaunch presets. - When a preset is active, context size (and other generation parameters) are driven by the preset — the UI disables editing and shows a clear tooltip.
- All context size logic has been consolidated into
StorageServicefor consistency across the app.
✨ UI Polish & Quality of Life
- Custom chat backgrounds — upload and name your own images per chat.
- Google Fonts picker for chat text styling.
- Per-character chat bubble colours that persist correctly when exporting to PNG.
- Scenario field is now expandable in the character editor.
- UI Settings dialog is now properly scrollable.
- Window size and position are remembered across restarts.
- Many small improvements to tooltips, log copying, preset validation, and widget stability.
🐛 Stability & Fixes
- Numerous Realism Engine interruption and regeneration fixes.
- Lorebook improvements: constant entries now persist correctly, better deduplication and wildcard/word-boundary matching.
- macOS build quality: proper bundle naming for Metal, improved DMG packaging.
- Many Tooltip, preset, and widget tree crashes resolved.
- Various packaging and CI improvements for cleaner releases.
This release represents one of the largest cumulative improvements to day-to-day feel and reliability since the Realism Engine was first introduced.
---
Prerequisites
- Flutter SDK
- Rust toolchain (for the RAG embedding server)
- Git
- Windows, Linux, or macOS
Linux Extra Dependencies
Ubuntu/Debian
sudo apt-get install clang cmake ninja-build pkg-config libgtk-3-dev liblzma-dev libstdc++-12-dev libwpewebkit-1.0-devArch Linux
sudo pacman -S clang cmake ninja pkgconf gtk3 xz wpewebkitFedora
sudo dnf install clang cmake ninja-build pkgconf-pkg-config gtk3-devel xz-devel libstdc++-devel wpewebkit-develwpewebkit is required for the built-in browser (Chub.ai / aicharactercards.com import). Pre-built AppImages bundle it automatically.📥 Install
Beta Builds
Beta releases (e.g. the 0.9.8-Beta series) are available for early access to new features. They are completely isolated from your stable installation:
- Data directory:
~/Documents/FrontPorchAI-Beta/ - All preferences are namespaced with a
beta_prefix - Stable builds will never offer a beta update (and vice versa)
This isolation protects your main library while you test new features. Beta builds are recommended only for users comfortable with occasional rough edges.
---
🛠️ Build from Source
Build & Run
git clone https://github.com/linux4life1/front-porch-ai.git
cd front-porch-ai
flutter pub get
flutter runmacOS release build (includes RAG embedding server):
./scripts/build-macos.shLinux / Windows release build:
cargo build --release --manifest-path tools/embed_server/Cargo.toml
flutter build linux # or windowstools/embed_server/target/release/embed_server next to the built executable under embed_server/embed_server.
---
<details> <summary><strong>📦 Old Release Notes</strong></summary>
⚙️ Configuration
- Backend — go to Settings → Download Backend to fetch KoboldCpp, or point it at an existing binary.
- Model — go to Manage Models → HuggingFace Search, find a GGUF model (recommended:
Q4_K_MorQ5_K_M), download. - Optimize — hit Auto-Configure to let the app pick the best GPU layer split and thread count for your hardware.
---
⚙️ KoboldCpp Integration
- Automated download and update of the KoboldCpp backend
- Hardware detection — Vulkan on PC, Metal on Apple Silicon, Intel ARC support, Nvidia Blackwell (RTX 50-series) support
- Model Hub: search and download GGUF models directly from HuggingFace
- Start/Stop KoboldCpp from inside the Character Creator
- Advanced Launch Options — collapsible panel exposing Flash Attention, Context Shift, mlock, GPU ID selector, and prefill batch size with sane auto-selected defaults
---
Linux — Package Manager
Debian / Ubuntu / Mint / Pop!_OS
curl -fsSL https://apt.dreamersai.art/install.sh | bash
sudo apt install front-porch-aicurl -fsSL https://apt.dreamersai.art/front-porch-ai.gpg | sudo gpg --dearmor -o /etc/apt/keyrings/front-porch-ai.gpg
echo "deb [signed-by=/etc/apt/keyrings/front-porch-ai.gpg] https://apt.dreamersai.art stable main" | sudo tee /etc/apt/sources.list.d/front-porch-ai.list
sudo apt update && sudo apt install front-porch-aiFedora / RHEL / openSUSE
sudo dnf config-manager --add-repo https://rpm.dreamersai.art/front-porch-ai.repo
sudo dnf install front-porch-aiArch Linux (AUR)
yay -S front-porch-ai-bin # Stable (recommended)
yay -S front-porch-ai-beta-bin # Beta / Early accessFuture updates arrive through your normal system updates (apt upgrade, dnf upgrade, yay -Syu).
Beta builds are available for early access to new features. They install to a completely separate directory (~/Documents/FrontPorchAI-Beta/) and usebeta_preference keys so they never touch your stable data. See the Beta Builds section below for details.
🆚 How Does Front Porch AI Compare?
If you're evaluating local AI tools, here's an honest breakdown. Every project on this list is doing something right — the goal isn't to trash competitors, it's to help you pick the right tool for you.
| Feature | **Front Porch AI** | SillyTavern | Jan.ai | Backyard AI |
|---|---|---|---|---|
| **Native desktop app** | ✅ Flutter (Win/Mac/Linux) | ❌ Web-based (local server) | ✅ Electron | ✅ (abandoned) |
| **Fully offline — no cloud required** | ✅ | ✅ | ✅ | ✅ |
| **Remote LLM Endpoints** | ✅ Native multi-provider support (OpenRouter, Nano-GPT, custom, etc.) with deep integration | ✅ Strong native support for custom OpenAI-compatible endpoints | ⚠️ Limited | ❌ (service discontinued) |
| **Built-in TTS (50+ voices)** | ✅ Kokoro + Piper + ElevenLabs + OpenAI | ⚙️ Extension required | ❌ | ❌ |
| **Speech-to-text (push-to-talk)** | ✅ Whisper, built-in | ⚙️ Extension required | ❌ | ❌ |
| **Local image generation** | ✅ A1111, Forge, Draw Things | ⚙️ Extension required | ❌ | ❌ |
| **Realism Engine** | ✅ Time, trust, emotion, chaos, objectives | ❌ | ❌ | ❌ |
| **Character Expressions** | ✅ ONNX + LLM, live avatar swap | ⚙️ Extension required | ❌ | ❌ |
| **RAG memory (local)** | ✅ ONNX embeddings, no cloud | ⚙️ Extension required | ❌ | ❌ |
| **Novel / story generator** | ✅ Porch Stories pipeline | ❌ | ❌ | ❌ |
| **Cloud sync** | ✅ Google Drive / WebDAV | ❌ | ❌ | ❌ |
| **Character card compatibility** | ✅ V2 spec + Backyard .byaf import | ✅ V2 spec | ❌ | .byaf only |
| **Group chat** | ✅ | ✅ | ❌ | ❌ |
| **Extension / plugin ecosystem** | ❌ | ⭐ Very large | Moderate | ❌ |
| **Open source license** | ✅ AGPL-3.0 | ✅ AGPL-3.0 | ✅ MIT | ❌ |
| **Best for** | Polished AI companion + storytelling | Power users / heavy customization | Simple local chat | — |
SillyTavern's extension ecosystem is genuinely impressive and unmatched for customization depth. If you want maximum flexibility and don't mind configuration work, it's excellent. Front Porch AI prioritises everything working out of the box for users who want to chat, not configure.
---
GPL 3开源，替代Backyard AI
- 构建多智能体协作系统的 Agent 开发者
- 构建企业知识库 / RAG 检索应用的团队
- 做语音类 AI 产品的开发者
- 本地部署优先选 GGUF 量化模型，节省显存并保持响应速度
- 分块大小建议 256-512 tokens，向量库优选 pgvector 或 Qdrant
- Agent 任务先做 dry-run 验证工具调用链，再开启自主执行
- API key 直接提交到 git 仓库（请用 .env 并加入 .gitignore）
- embedding 模型与查询模型不一致导致检索失效
- 显存不足直接 OOM — 优先降低 context 或换更小的量化模型
- 本地部署：CPU 8GB 起，GPU 推荐 16GB+ 显存
- 云端托管：可放在 Vercel / Railway / Fly.io 等 PaaS 平台
⚡ 核心功能
- 开源免费，支持本地部署，数据完全自主可控
- 活跃的 GitHub 开源社区，持续迭代更新
- 提供详细文档和使用示例，新手友好
- 支持自定义配置，灵活适配不同使用环境
- 可作为基础组件集成进现有技术栈或进行二次开发
- 构建多智能体协作系统的 Agent 开发者
- 构建企业知识库 / RAG 检索应用的团队
- 做语音类 AI 产品的开发者
- 本地部署优先选 GGUF 量化模型，节省显存并保持响应速度
- 分块大小建议 256-512 tokens，向量库优选 pgvector 或 Qdrant
- Agent 任务先做 dry-run 验证工具调用链，再开启自主执行
- API key 直接提交到 git 仓库（请用 .env 并加入 .gitignore）
- embedding 模型与查询模型不一致导致检索失效
- 显存不足直接 OOM — 优先降低 context 或换更小的量化模型
👥 适合人群
🎯 使用场景
- 本地部署运行，保护数据隐私，满足合规要求
- 自定义集成到现有系统，扩展技术栈能力
- 作为开源基础组件进行商业化二次开发
⚖️ 优点与不足
- +完全开源免费，无授权费用
- +本地部署，数据完全自主可控
- +开发者社区支持，遇问题可查可问
- −安装和初始配置可能需要一定技术基础
- −功能完整性通常不如成熟商业产品
- −技术支持主要依赖开源社区，响应速度不稳定
该工具使用 AGPL-3.0 协议，商用场景请仔细阅读协议条款，必要时咨询法律意见。
AI Skill Hub 为第三方内容聚合平台，本页面信息基于公开数据整理，不对工具功能和质量作任何法律背书。
建议在沙箱或测试环境中充分验证后，再部署至生产环境，并做好必要的安全评估。
⚠️ AGPL 3.0 — 最严格的 Copyleft，网络服务端使用也需开源，SaaS 使用受限。
🔗 相关工具推荐
❓ 常见问题 FAQ
总体来看，前廊AI 是一款质量良好的AI工具，在同类工具中具备一定竞争力。AI Skill Hub 将持续追踪其更新动态，建议收藏备用，结合自身场景选择合适时机引入使用。
| 原始名称 | front-porch-AI |
| 原始描述 | 开源AI工具：GPL 3 replacement of defunct Backyard AI character creation/chat platform。⭐39 · Dart |
| Topics | aiai-companioncharacter-aichatbotdesktop-appdart |
| GitHub | https://github.com/linux4life1/front-porch-AI |
| License | AGPL-3.0 |
| 语言 | Dart |
收录时间：2026-06-02 · 更新时间：2026-06-02 · License：AGPL-3.0 · AI Skill Hub 不对第三方内容的准确性作法律背书。