Back to Home
wspotter icon

felix

by wspotter

Overview

A fully local, real-time AI voice assistant with barge-in support, speech-to-text, text-to-speech, and autonomous tool execution including image generation and task management, designed for high-performance AMD GPUs.

Installation

Run Command
./launch.sh

Environment Variables

  • WHISPER_MODEL
  • WHISPER_DEVICE
  • WHISPER_COMPUTE_TYPE
  • WHISPER_GPU_DEVICE
  • LLM_BACKEND
  • OLLAMA_URL
  • LMSTUDIO_URL
  • OPENAI_URL
  • OPENAI_API_KEY
  • OLLAMA_MODEL
  • OLLAMA_TEMPERATURE
  • OLLAMA_MAX_TOKENS
  • TTS_ENGINE
  • TTS_VOICE
  • SERVER_HOST
  • SERVER_PORT
  • AUDIO_SAMPLE_RATE
  • AUDIO_CHANNELS
  • AUDIO_CHUNK_MS
  • BARGE_IN_ENABLED
  • BARGE_IN_THRESHOLD
  • BARGE_IN_MIN_SPEECH_MS
  • LOG_LEVEL
  • OTEL_ENABLED
  • OTEL_ENDPOINT
  • VIP_API_URL
  • VIP_API_KEY
  • FACEBOOK_ACCESS_TOKEN
  • INSTAGRAM_ACCESS_TOKEN
  • FACEBOOK_PAGE_ID
  • INSTAGRAM_ACCOUNT_ID

Security Notes

The server uses `eval()` in `server/tools/builtin/web_tools.py` for mathematical calculations. While attempts are made to sanitize the input by checking for allowed characters, `eval()` is inherently dangerous and can lead to arbitrary code execution if not perfectly secured. Additionally, the system heavily relies on `subprocess.run` to execute external binaries like `whisper-cli`, `piper`, and `mpc`. This introduces potential vulnerabilities if the binaries are compromised or if command injections are possible through untrusted input. The `mcpart` component also interfaces with external social media APIs (Facebook, Instagram) and a 'VIP Pipeline' API, increasing the external attack surface.

Similar Servers

Stats

Interest Score35
Security Score3
Cost ClassLow
Avg Tokens1200
Stars1
Forks0
Last Update2025-12-05

Tags

Voice AssistantAIReal-timeLocalToolingConversational AISpeech-to-TextText-to-SpeechLarge Language ModelImage GenerationMusic ControlTask ManagementGPU Accelerated