felix
by wspotter
Overview
Real-time conversational AI voice assistant with local processing, autonomous tool execution, and PWA support.
Installation
./run.shEnvironment Variables
- WHISPER_MODEL
- WHISPER_DEVICE
- WHISPER_GPU_DEVICE
- OLLAMA_URL
- OLLAMA_MODEL
- TTS_ENGINE
- TTS_VOICE
- SERVER_HOST
- SERVER_PORT
- ADMIN_TOKEN
- ENABLE_AUTH
- ENABLE_TOOL_TUTOR
- TOOL_CONFIDENCE_THRESHOLD
- DATA_DIR
- SESSION_SAVE_INTERVAL
- FELIX_ENABLE_TUTOR_EMBEDDINGS
- OPENAI_API_KEY
- OPENROUTER_API_KEY
- LMSTUDIO_URL
Security Notes
The `calculate` tool uses `eval()` with input sanitization, which is inherently risky and could be a vector for code injection if not perfectly secured. The `ADMIN_TOKEN` in `server/config.py` can be empty or weak, potentially allowing unauthorized administrative access. User and session data is stored in local JSON files, which is suitable for a local-first application but lacks enterprise-grade security features like encryption at rest or robust access control. Tool argument sanitization needs consistent vigilance across all 56+ tools. `subprocess` is used for external binaries (Piper, Whisper.cpp, ComfyUI), which is common but requires careful input validation to prevent command injection.
Similar Servers
uLoopMCP
Acts as a Model Context Protocol (MCP) server to bridge Unity Editor operations with AI coding tools like Cursor and Claude Code, enabling autonomous development loops for compiling, testing, logging, and scene automation.
codebadger
A containerized Model Context Protocol (MCP) server providing static code analysis using Joern's Code Property Graph (CPG) technology.
seamless-agent
Empowers AI agents within GitHub Copilot Chat (or similar LM clients) to seek interactive user confirmation and feedback via specialized tools (ask_user, plan_review, walkthrough_review) before executing actions, ensuring user control.
mcp
This MCP server assists developers by identifying inefficient or outdated npm packages in install commands and source files, providing migration documentation.