Project Overview: We are building a local-only, voice-first assistant on a Raspberry Pi 5. The system should be fully offline, modular, and capable of handling voice commands via a wake word. Core functionalities include local speech recognition (STT), text-to-speech (TTS), lightweight memory, and a FastAPI-based skill layer. No cloud or web interface required at this stage. This is not a research or experimental project. We want a working, reproducible v1 setup with clear documentation, suitable for future modular expansion. Hardware (provided by client): -Raspberry Pi 5 (8GB RAM) -ReSpeaker 4-Mic Array (compatible with Pi) -Creative Pebble V3 Speakers (3.5mm input) -32GB microSD card -(Optional for future: 1TB NVMe SSD via UGREEN USB-C enclosure) -(Optional: GPIO button for wake trigger — off by default) Software Stack Requirements: -OS: Raspberry Pi OS Lite (or best alternative) -Wake Word: openWakeWord (or similar lightweight local option) -Audio Processing: WebRTC VAD, AGC, NS, AEC (as needed) -STT: Vosk (fully offline) -TTS: Piper (fully offline) -API Layer: FastAPI in Python -Memory: SQLite (persistent conversation context, able to wipe) -Logging: CLI + file logs, no GUI required Functionality Requirements: -Boots on startup, loads assistant automatically -Wake-word detection or GPIO button activation (toggleable) -Round-trip voice interaction: wake word > STT > command > TTS -Modular skill design (easy to add/remove/modify commands) -Lightweight memory to store context/history -Clear hooks for future upgrades (e.g. LLM, web interface) Deliverables: -Fully working install script or bootable SD image -All source code and config files -Setup + usage documentation -Modular skill structure (Python-based) -Support for local expansion (no cloud dependencies) -An IP transfer agreement will be required upon awarding Milestones: We are open to two or three milestones. Suggested breakdown: -Milestone 1 ($400): OS + audio + Vosk + Piper + wake word working as a full round-trip voice pipeline -Milestone 2 ($400–600): FastAPI layer, memory setup (SQLite), modular skills What We’re Looking For: -Experience with Raspberry Pi hardware and audio setup -Proven ability to implement offline STT/TTS solutions -Python + FastAPI expertise -Clean, modular coding approach (with docs) -Fluent communication + ability to work async To apply, please include: -Examples of similar projects you've completed (offline preferred) -Confirmation you can work within the $800–$1200 AUD range -Your proposed milestone structure and estimated timeframe We're ready to begin immediately and can award the project quickly if aligned. Looking forward to working with someone who can bring this to life.