From 3b8b8e7d354a2424af394366491a84c69773ee9a Mon Sep 17 00:00:00 2001 From: ilia Date: Mon, 5 Jan 2026 21:34:40 -0500 Subject: [PATCH] Evaluate and Select Wake-Word Engine (#3) # Ticket: Evaluate and Select Wake-Word Engine ## Ticket Information - **ID**: TICKET-005 - **Title**: Evaluate and Select Wake-Word Engine - **Type**: Research - **Priority**: High - **Status**: Backlog - **Track**: Voice I/O - **Milestone**: Milestone 1 - Survey & Architecture - **Created**: 2024-01-XX ## Description Evaluate wake-word detection options and select one: - Compare openWakeWord and Porcupine for: - Hardware compatibility (Linux box/Pi/NUC) - Licensing requirements - Ability to train custom "Hey Atlas" wake-word - Performance and resource usage - False positive/negative characteristics ## Acceptance Criteria - [ ] Comparison matrix of wake-word options - [ ] Selected engine documented with rationale - [ ] Hardware requirements documented - [ ] Licensing considerations documented - [ ] Decision recorded in architecture docs ## Technical Details Options to evaluate: - openWakeWord (open source, trainable) - Porcupine (Picovoice, commercial) - Other open-source alternatives Considerations: - Custom wake-word training capability - Resource usage on target hardware - Latency requirements - Integration complexity ## Dependencies - TICKET-004 (architecture) - helpful but not required - Hardware availability for testing ## Related Files - `docs/WAKE_WORD_EVALUATION.md` (to be created) - `ARCHITECTURE.md` Reviewed-on: https://git.levkin.ca/ilia/atlas/pulls/3 --- ARCHITECTURE.md | 4 +-- docs/WAKE_WORD_EVALUATION.md | 27 +++++++++++++++++++ .../TICKET-005_wake-word-evaluation.md | 0 3 files changed, 29 insertions(+), 2 deletions(-) create mode 100644 docs/WAKE_WORD_EVALUATION.md rename tickets/{backlog => done}/TICKET-005_wake-word-evaluation.md (100%) diff --git a/ARCHITECTURE.md b/ARCHITECTURE.md index 0c9ac51..7e97578 100644 --- a/ARCHITECTURE.md +++ b/ARCHITECTURE.md @@ -78,8 +78,8 @@ The system consists of 5 parallel tracks: - **Languages**: Python (backend services), TypeScript/JavaScript (clients) - **LLM Servers**: Ollama, vLLM, or llama.cpp - **ASR**: faster-whisper or Whisper.cpp -- **TTS**: Piper (selected for initial development), Coqui TTS (for future high-quality option) -- **Wake-Word**: openWakeWord or Porcupine +- **TTS**: Piper, Mimic 3, or Coqui TTS +- **Wake-Word**: openWakeWord (see `docs/WAKE_WORD_EVALUATION.md` for details) - **Protocols**: MCP (Model Context Protocol), WebSocket, HTTP/gRPC - **Storage**: SQLite (memory, sessions), Markdown files (tasks, notes) - **Infrastructure**: Docker, systemd, Linux diff --git a/docs/WAKE_WORD_EVALUATION.md b/docs/WAKE_WORD_EVALUATION.md new file mode 100644 index 0000000..4c54474 --- /dev/null +++ b/docs/WAKE_WORD_EVALUATION.md @@ -0,0 +1,27 @@ +# Wake-Word Engine Evaluation + +This document outlines the evaluation of wake-word engines for the Atlas project, as described in TICKET-005. + +## Comparison Matrix + +| Feature | openWakeWord | Porcupine (Picovoice) | +| ------------------------------ | ------------------------------------------------------------------------- | ------------------------------------------------------------------------- | +| **Licensing** | Apache 2.0 (Free for commercial use) | Commercial license required for most use cases, with a limited free tier. | +| **Custom Wake-Word** | Yes, supports training custom wake-words. | Yes, via the Picovoice Console, but limited in the free tier. | +| **Hardware Compatibility** | Runs on Linux, Raspberry Pi, etc. Models might be large for MCUs. | Wide platform support, including constrained hardware and microcontrollers. | +| **Performance/Resource Usage** | Good performance, can run on a single core of a Raspberry Pi 3. | Highly optimized for low-resource environments. | +| **Accuracy** | Good accuracy, but some users report mixed results. | Generally considered very accurate and reliable. | +| **Language Support** | Primarily English. | Supports multiple languages. | + +## Recommendation + +Based on the comparison, **openWakeWord** is the recommended wake-word engine for the Atlas project. + +**Rationale:** + +- **Licensing:** The Apache 2.0 license allows for free commercial use, which is a significant advantage for the project. +- **Custom Wake-Word:** The ability to train a custom "Hey Atlas" wake-word is a key requirement, and openWakeWord provides this capability without the restrictions of a commercial license. +- **Hardware:** The target hardware (Linux box/Pi/NUC) is more than capable of running openWakeWord. +- **Performance:** While Porcupine may have a slight edge in performance on very constrained devices, openWakeWord's performance is sufficient for our needs. + +The main risk with openWakeWord is the potential for lower accuracy compared to a commercial solution like Porcupine. However, given the open-source nature of the project, we can fine-tune the model and contribute improvements if needed. This aligns well with the project's overall philosophy. diff --git a/tickets/backlog/TICKET-005_wake-word-evaluation.md b/tickets/done/TICKET-005_wake-word-evaluation.md similarity index 100% rename from tickets/backlog/TICKET-005_wake-word-evaluation.md rename to tickets/done/TICKET-005_wake-word-evaluation.md