惯性聚合 高效追踪和阅读你感兴趣的博客、新闻、科技资讯
阅读原文 在惯性聚合中打开

推荐订阅源

L
LangChain Blog
S
SegmentFault 最新的问题
V
Visual Studio Blog
J
Java Code Geeks
宝玉的分享
宝玉的分享
美团技术团队
博客园 - Franky
酷 壳 – CoolShell
酷 壳 – CoolShell
H
Hackread – Cybersecurity News, Data Breaches, AI and More
有赞技术团队
有赞技术团队
量子位
Martin Fowler
Martin Fowler
MyScale Blog
MyScale Blog
Google DeepMind News
Google DeepMind News
Jina AI
Jina AI
博客园 - 叶小钗
月光博客
月光博客
P
Proofpoint News Feed
D
DataBreaches.Net
Blog — PlanetScale
Blog — PlanetScale
博客园_首页
腾讯CDC
Microsoft Azure Blog
Microsoft Azure Blog
Stack Overflow Blog
Stack Overflow Blog

Hacker News - Newest: "AI"

AI can't read an investor deck AI as an attorney? Student uses ChatGPT, Gemini to sue UW over alleged racial discrimination Hacking MCP Servers in AI Systems – The Rug Pull: Tool Changes After Approval GitHub - MeepCastana/KubeezCut: Free Web based video editor Can AI judge journalism? A Thiel-backed startup says yes, even if it risks chilling whistleblowers Coming soon: 10 Things That Matter in AI Right Now DARPA built an AI to fact-check enemy weapons claims What explains heterogeneity in AI adoption? When AI Meets Muscle: Context-Aware Electrical Stimulation Promises a New Way to Guide Human Movements - Department of Computer Science AI Changed How We Build. It Did Not Change What Matters. Linux rules on using AI-generated code - Copilot is OK, but humans must take 'full responsibility for the… Meta spins up AI version of Mark Zuckerberg to engage with employees Code Mode: Let Your AI Write Programs, Not Just Call Tools | TanStack Blog GitHub - Delavalom/graft: Go framework for building AI agents. Type-safe tools, multi-provider (OpenAI, Anthropic, Gemini, Bedrock), zero vendor SDKs. India's TCS tops estimates, says new AI models did not dent services demand Gen Z's fading AI hype Strong feeling: we are in a folded AI reality GitHub - machinarii/total-recall-catalog: A reference catalog of latest knowledge retrieval, memory & RAG systems GitHub - mensfeld/code-on-incus: Give each AI agent its own isolated machine with root, Docker, and systemd. Active defense detects and stops threats automatically.. Quantization, LoRA, and the 8% Problem: Benchmarking Local LLMs for Production AI Iran war: We spoke to the man making Lego-style AI videos that experts say are powerful propaganda Powell, Bessent discussed Anthropic's Mythos AI cyber threat with major U.S. banks GitHub - immartian/bellamem: Persistent belief-graph memory for AI agents. Retrieves decisive context by importance — not recency, not RAG, not /compact. recursive-mode: The Repo-Native Operating System for AI Engineering After the attack on Sam Altman's home, will AI CEO's go on the offensive? The biggest advance in AI since the LLM Opus 4.6 vs GPT 5.4 One Prompt Unity World Generation Test “AI polls” are fake polls Client Challenge Can AI be a 'child of God'? Inside Anthropic's meeting with Christian leaders
GitHub - joshuatic/voxel: Voxel is a local-first AI assis...
joshuatic · 2026-05-10 · via Hacker News - Newest: "AI"

Voxel logo

Voxel is a local-first AI assistant that runs on your own machine through a lightweight web dashboard.

It is designed to feel like a private desktop copilot:

  • answers questions with a local GGUF model,
  • uses web search when the internet is available,
  • supports push-to-talk voice input,
  • and can speak responses with local voices.

What Voxel does

  • Local-first answering with llama-cpp-python and GGUF models in models/
  • Web-assisted answers using DuckDuckGo search (ddgs) when online
  • Offline fallback when internet is unavailable (or user forces local mode)
  • Voice output via local Piper voices (/voice/speak)
  • Voice input via Faster-Whisper transcription (/voice/transcribe)
  • Search history + settings persisted in local SQLite (data/voxel.db)
  • Assistant personality control (get/set/reset)
  • Voice pack import from ZIP files with validation/safety checks
  • Cache visibility + clearing for generated/downloaded audio data

Tech stack (current)

  • Backend: FastAPI
  • Frontend: static HTML/CSS/JS dashboard in static/
  • Local LLM runtime: llama-cpp-python
  • Web search: ddgs
  • Speech-to-text: faster-whisper
  • Data storage: SQLite (data/voxel.db)
  • Key storage for cloud providers: system keyring backend

No cloud dependency is required for a core local operation.


Project layout

app/
  main.py             # FastAPI app + API routes
  command_router.py   # Online/offline routing logic
  local_ai.py         # GGUF model selection/loading/inference
  search.py           # DuckDuckGo search + result formatting
  transcription.py    # Faster-Whisper transcription pipeline
  tts.py              # Piper CLI speech synthesis
  voice.py            # Voice listing/selection/phrase serving
  voice_import.py     # Voice ZIP import + validation
  api_keys.py         # Provider key + provider state management
  personality.py      # Assistant personality storage
  storage.py          # SQLite schema + read/write operations
  cache_manager.py    # Cache size/status/clear operations
  network.py          # Simple internet reachability check
  config.py           # Paths and app-level constants
  logging_setup.py    # Startup log rotation + logger setup

static/
  index.html          # Dashboard shell
  app.js              # UI logic + API calls + voice controls
  styles.css          # UI styling

models/               # Put GGUF model files here
voices/               # Voice packs (Piper or phrase packs)
data/                 # SQLite DB + runtime cache data
logs/                 # latest.log + rotated compressed logs

Quick start (Windows / PowerShell)

  1. Create and activate a virtual environment:
python -m venv .venv
.\.venv\Scripts\Activate.ps1
  1. Install Python dependencies:
pip install -r requirements.txt
  1. Add at least one GGUF model to models/.

    • Voxel auto-selects the first .gguf file in that folder when set to auto.
  2. Ensure the required local tooling is available:

    • piper command available in PATH (for TTS)
    • FFmpeg/av installed if audio decoding issues appear during transcription
  3. Run the app:

.\run.ps1

By default, the server runs at http://127.0.0.1:8787.


Runtime behavior notes

  • If the internet is reachable, command routing can do search + summarize.
  • If the internet is unavailable (or the user uses local/offline prefix), Voxel uses local-only answering.
  • API provider keys (OpenAI/OpenRouter/Gemini) are optional and managed through keyring-backed settings endpoints.
  • Voice import enforces safe ZIP paths and allowed file types.

API highlights

  • Health/model: /health, /model/status, /model/list, /model/select
  • Search/commands: /search, /command
  • History/suggestions: /history/recent, /history/clear, /suggestions
  • Voice: /voice/list, /voice/select, /voice/phrase, /voice/speak, /voice/transcribe, /voice/import
  • Personality: /personality, /personality/reset
  • API keys/providers: /api-keys/status, /api-keys/set, /api-keys/clear, /api-keys/enabled, /api-keys/active
  • Cache/network: /cache/status, /cache/clear, /network/status

Privacy model

Voxel is built to run locally first. Data (history/settings/logs/cache) is stored on the local machine. Internet usage is primarily for web search and any explicitly enabled external provider usage.


Contributing

Please read CONTRIBUTING.md before opening a PR.


License

Voxel is licensed under the GNU General Public License v3.0 or later.

See LICENSE for details.