Skip to the content
2 layers42 × 35.6 mmXIAO Plus Base2 V1.0, the printed circuit board in 3D

XIAO Plus Base2 V1.0

lachlanchen/IdeasGlass

IdeasGlass is an AI-first wearable + PWA that captures and streams your ideas in real time.

Size
42 × 35.6 mm
Copper layers
2
Thickness
1.6 mm
Parts
4
Nets
23
Pads
59
Tracks, vias
61, 43
Pours
1
Routed
Yes
KiCad
KiCad 8.0 (file 20240108)
Created
8 Nov 2025
Updated
15 Oct 2026

Overview

XIAO Plus Base2 V1.0 is an open-source printed circuit board by lachlanchen, from the lachlanchen/IdeasGlass repository on GitHub. It is a 2-layer board of 42 × 35.6 mm with 4 parts on 23 nets, routed with 61 tracks and 43 vias.

Its parts include 2 connectors and 1 IC. It belongs with the wearables boards and the ESP32 boards of this library.

README

README.md

English · العربية · Español · Français · 日本語 · 한국어 · Tiếng Việt · 中文 (简体) · 中文(繁體) · Deutsch · Русский

LazyingArt banner — apps and four buddies

IdeasGlass

A wearable AI glass that turns ideas into actions, income, and creative momentum.

Voice-first wearable AI pipeline: capture from ESP32 glasses, process in FastAPI, and monitor/control via a live PWA dashboard.

Python FastAPI ESP32 PWA Streaming Locale

LanePurpose
🎙️ Wearable captureESP32 glasses send audio, photos, and telemetry in near-real-time
🧠 Backend intelligenceFastAPI ingests streams, transcribes, segments, and persists metadata
🖥️ DashboardPWA shows live waveform, transcripts, and device/account status
IdeasGlass App UI IdeasGlass hardware
App UI (left) · Hardware (right)

Explore community experiments at onlyideas.art.

🚀 Overview

IdeasGlass is an AI-first wearable system for voice-first idea capture and execution. In this repository, the primary runtime path is:

  • backend/glass/ for FastAPI APIs, WebSocket ingest, Whisper-based transcription, and the installable PWA dashboard.
  • IdeaGlass/firmware/ideasglass_arduino/IdeasGlassClient/ for XIAO ESP32S3 firmware streaming telemetry/audio/photos.

If you are new to this repo, start there first.

📚 Table of Contents

At a glance

AreaPrimary locationWhat it does
Backend API + PWAbackend/glass/FastAPI endpoints, WebSocket ingest/fanout, transcription, dashboard
FirmwareIdeaGlass/firmware/ideasglass_arduino/IdeasGlassClient/ESP32 capture/streaming client
Bridge notesreferences/ideasglass_bridge.mdTLS/WAN reliability notes and deployment field tips
README translationsi18n/Multilingual docs synced from the canonical README

✨ Why IdeasGlass

IdeasGlass is an AI-first wearable built for people who live in streams of ideas. It captures, translates, organizes, and executes on creativity the moment inspiration strikes, whether you are narrating a concept in motion or hosting a live session.

🧩 Features

Product vision features

  • Creation-native hardware – lightweight glasses and wearable inputs, tuned for voice-first capture plus subtle gesture shortcuts.
  • Instant translation – real-time language detection/translation so you can ideate across teams or audiences without switching tools.
  • EchoMind co-pilot – tight pairing with chat.lazying.art for brainstorming, script drafting, and multilingual content coaching.
  • Channel autopilot – drafts outlines, long-form scripts, short-form hooks, and schedules uploads to YouTube or other feeds.
  • Highlights & reels – auto-selects moments, generates thumbnails, subtitles, and social-ready clips.
  • Income layer – connects to LazyingArt Coin for tipping, credit payouts, and conversion to on-chain assets.
  • Spending & focus – tracks operational spend, surfaces profitable formats, and distills your personal strengths into next projects.

Repository/runtime features

  • FastAPI backend with REST + WebSocket endpoints for ingest (/api/v1/audio, /ws/audio-ingest) and live stream fanout (/ws/stream).
  • Deterministic audio segmentation (default ~15 s with overlap) into backend/glass/audio_segments/.
  • Optional openai-whisper streaming transcripts with configurable latency thresholds.
  • Optional Postgres persistence (DATABASE_URL) for messages, photos, chunks, segments, transcripts.
  • PWA dashboard with live waveform, transcript updates, and install support on desktop/mobile.
  • Arduino firmware support for XIAO ESP32S3 Sense camera + mic flows.

🔄 Sample Workflow

  1. Capture – Speak or sketch a concept; IdeasGlass transcribes, translates, and tags the intent.
  2. Co-create – EchoMind refines the idea, drafts scripts, and suggests CTAs tailored for each platform.
  3. Publish – The channel agent auto-produces highlight videos, gallery images, and uploads them with metadata.
  4. Monetize – Credits route through LazyingArt Coin (coin.lazying.art) and payouts sync with your preferred wallets.
  5. Reflect – Spending, reach, and engagement dashboards surface what to double down on next.

🗂️ Project Structure

text
IdeasGlass/
├── README.md
├── i18n/                                  # README translations
├── backend/
│   ├── glass/                             # Primary FastAPI + PWA backend
│   │   ├── app.py
│   │   ├── serve.py
│   │   ├── requirements.txt
│   │   ├── static/
│   │   ├── tools/
│   │   └── audio_segments/
│   ├── tornado_app/                       # Secondary/parallel ingest backend path
│   ├── memo/
│   ├── memo_legacy/
│   └── ngrok_bridge/
├── IdeaGlass/firmware/ideasglass_arduino/
│   ├── IdeasGlassClient/
│   ├── config.h
│   ├── WifiTest/WifiTest.ino
│   ├── wifi_credentials.example.h
│   └── README.md
├── references/ideasglass_bridge.md        # Bridge + deployment notes
├── docs/                                  # Additional site/docs assets
├── development_plan/

🧰 Prerequisites

  • Python 3.10+
  • pip (or conda environment with compatible Python)
  • Optional: NVIDIA GPU + CUDA/cuDNN for faster Whisper inference
  • Optional: PostgreSQL for persistence
  • For firmware: Arduino IDE or arduino-cli, Seeed XIAO ESP32S3 Sense, PSRAM enabled
ComponentRequirementNotes
Backend runtimePython 3.10+, pipUse venv or conda (glass)
GPU acceleration (optional)NVIDIA + CUDA/cuDNNImproves Whisper latency
Persistence (optional)PostgreSQLEnabled via DATABASE_URL
Firmware toolchainArduino IDE / arduino-cliUse XIAO ESP32S3 profile with PSRAM

⚙️ Installation

Backend dependencies

bash
cd backend/glass
python -m venv .venv
source .venv/bin/activate
pip install -r requirements.txt

Firmware prerequisites

  • Copy IdeaGlass/firmware/ideasglass_arduino/wifi_credentials.example.h to wifi_credentials.h (recommended) and set SSID/password.
  • In Arduino IDE, use board ESP32 -> XIAO_ESP32S3 with PSRAM: OPI PSRAM.
  • Partition scheme: Default with spiffs (3MB APP/1.5MB SPIFFS) or Maximum APP when filesystem is not needed.

▶️ Usage

Run backend (uvicorn)

bash
IDEASGLASS_WHISPER_MODEL=base IDEASGLASS_WHISPER_DEVICE=cuda \
uvicorn backend.glass.app:app \
  --host 0.0.0.0 \
  --port 8765 \
  --proxy-headers \
  --forwarded-allow-ips="*" \
  --reload

Run backend (helper)

bash
python backend/glass/serve.py --whisper-model base --whisper-device cuda --reload

Open dashboard

  • http://localhost:8765/
  • http://localhost:8765/healthz
EndpointPurpose
/Main dashboard (PWA-capable UI)
/healthzBackend liveness check
/ws/audio-ingestDevice ingest WebSocket
/ws/streamLive stream fanout to dashboard clients

Login and bind your device

  1. Register or login from the dashboard Settings/Account area.
  2. Bind your device ID in the Bind device field.
  3. Only bound devices will stream to your account.

Generate a device ID + QR image:

bash
python backend/glass/tools/generate_device_id.py --out logs/device-id.png

Bind through API (cookie session required):

bash
curl -X POST http://localhost:8765/api/v1/devices/bind \
  -H 'Content-Type: application/json' \
  -d '{"device_id":"<your-device-id>"}' \
  -b cookies.txt -c cookies.txt

Verify current account and bound devices:

bash
curl -s http://localhost:8765/api/v1/auth/me -b cookies.txt -c cookies.txt | jq

Optional migration (rename historical data to a new device ID):

bash
curl -X POST http://localhost:8765/api/v1/devices/rename \
  -H 'Content-Type: application/json' \
  -d '{"from_id":"old-id","to_id":"<your-device-id>"}' \
  -b cookies.txt -c cookies.txt

Firmware build/upload (Arduino CLI)

bash
FQBN='esp32:esp32:XIAO_ESP32S3:PartitionScheme=default_8MB,PSRAM=opi'
SKETCH='IdeaGlass/firmware/ideasglass_arduino/IdeasGlassClient'
PORT='/dev/ttyACM0'

bin/arduino-cli compile --fqbn "$FQBN" "$SKETCH"
bin/arduino-cli upload -p "$PORT" --fqbn "$FQBN" "$SKETCH"

If port is busy: fuser -k /dev/ttyACM0. If permission denied: sudo usermod -aG dialout $USER then re-login (or temporary sudo chmod a+rw /dev/ttyACM0).

Firmware power UX (XIAO ESP32S3)

  • Hold the button ~0.8 s at power-up to boot.
  • Hold ~2.5 s while running to enter deep sleep.
  • Short press while running still triggers capture.

🛠️ Configuration

Core environment variables

  • DATABASE_URL: optional Postgres DSN for persistent storage.
  • IDEASGLASS_WHISPER_MODEL: base (default), small, medium, large-v3, large-v3-turbo.
  • IDEASGLASS_WHISPER_DEVICE: cuda or cpu.
  • IDEASGLASS_WHISPER_FP16: 1 for GPU mixed precision, 0 for CPU.
  • IDEASGLASS_TRANSCRIBE: 1 (default) to enable transcription, 0 to disable.
  • IDEASGLASS_TRANSCRIPT_INTERVAL_MS: rolling transcript interval.
  • IDEASGLASS_TRANSCRIPT_THRESHOLDS_MS: comma-separated thresholds (default 3000,6000,15000).
VariableDefault / optionsEffect
DATABASE_URLunset by defaultEnables Postgres persistence for account/device data
IDEASGLASS_WHISPER_MODELbase (small, medium, large-v3, large-v3-turbo)Controls accuracy vs latency
IDEASGLASS_WHISPER_DEVICEcuda or cpuInference backend
IDEASGLASS_WHISPER_FP161 GPU, 0 CPU-safeMixed precision control
IDEASGLASS_TRANSCRIBE1Toggle transcription pipeline
IDEASGLASS_TRANSCRIPT_INTERVAL_MSruntime configuredRolling transcript push interval
IDEASGLASS_TRANSCRIPT_THRESHOLDS_MS3000,6000,15000Progressive transcript emission thresholds

Safe DATABASE_URL examples:

  • export DATABASE_URL="postgresql://<db_user>@localhost/ideasglass_db" (peer/local auth)
  • export DATABASE_URL="postgresql://<db_user>:<db_password>@localhost/ideasglass_db" (password auth)

Audio gain and segmentation knobs

  • IDEASGLASS_GAIN_TARGET (default 0.032)
  • IDEASGLASS_GAIN_MAX (default 1.8)
  • IDEASGLASS_GAIN_MIN_RMS (default 0.008)
  • IDEASGLASS_SPEECH_RMS (default 0.03)
  • IDEASGLASS_SPEECH_MARGIN (default 0.005)
  • IDEASGLASS_SEGMENT_TARGET_MS (default 15000)
  • IDEASGLASS_SEGMENT_OVERLAP_MS (default 2000)
  • IDEASGLASS_SEGMENT_GAIN_TARGET (defaults to chunk gain target)
Audio knobDefaultPurpose
IDEASGLASS_GAIN_TARGET0.032Target RMS normalization
IDEASGLASS_GAIN_MAX1.8Upper clamp for gain amplification
IDEASGLASS_GAIN_MIN_RMS0.008Floor to avoid amplifying near-silence
IDEASGLASS_SPEECH_RMS0.03Speech activity RMS baseline
IDEASGLASS_SPEECH_MARGIN0.005Margin around speech threshold
IDEASGLASS_SEGMENT_TARGET_MS15000Segment length target
IDEASGLASS_SEGMENT_OVERLAP_MS2000Segment overlap for continuity
IDEASGLASS_SEGMENT_GAIN_TARGETinherits chunk gainSegment-level normalization target

Model prefetch (optional)

bash
python backend/glass/tools/prefetch_whisper_models.py \
  --models tiny,base,small,medium,large-v3 \
  --device cuda \
  --fp16 1

🧪 Examples

Generate and bind a device ID

bash
python backend/glass/tools/generate_device_id.py --out logs/device-id.png

Then set kDeviceId in:

  • IdeaGlass/firmware/ideasglass_arduino/IdeasGlassClient/IdeasGlassClient.ino

Dashboard flow:

  1. Register/login in Settings.
  2. Bind the device in the Account panel.
  3. Only bound devices stream to your account.

REST ingest examples

bash
curl -X POST http://localhost:8765/api/v1/messages \
  -H 'Content-Type: application/json' \
  -d '{"device_id":"dev-001","message":"hello from curl"}'
bash
curl -X POST http://localhost:8765/api/v1/messages \
  -H 'Content-Type: application/json' \
  -d '{
    "device_id":"dev-001",
    "message":"photo demo",
    "photo_base64":"'"$(base64 -w0 sample.jpg)'"',
    "photo_mime":"image/jpeg"
  }'
bash
rec --bits 16 --channels 1 --rate 16000 -c 1 -b 16 -e signed-integer temp.raw trim 0 0.25
curl -X POST http://localhost:8765/api/v1/audio \
  -H 'Content-Type: application/json' \
  -d '{
    "device_id":"dev-001",
    "sample_rate":16000,
    "bits_per_sample":16,
    "duration_ms":250,
    "rms":0.05,
    "audio_base64":"'"$(base64 -w0 temp.raw)'"'
  }'
bash
curl http://localhost:8765/api/v1/audio/segments | jq '.[0]'
curl -o latest.wav http://localhost:8765/api/v1/audio/segments/<segment-id>

🧭 Development Notes

Focus area

This repo contains multiple backend tracks. Current contributor guidance and runtime focus is backend/glass/ unless otherwise requested.

Static/syntax check

bash
python -m compileall backend/glass/app.py

Developer docs

Note: In the current repository snapshot, some historical links above appear to have moved (for example, bridge notes now exist at references/ideasglass_bridge.md). The original links are preserved as canonical README content.

Quick device binding (preserved workflow)

  • Generate ID (in conda glass): python backend/glass/tools/generate_device_id.py
  • Set it in firmware: IdeaGlass/firmware/ideasglass_arduino/IdeasGlassClient/IdeasGlassClient.ino (kDeviceId)
  • Run backend and open http://localhost:8765, register/login, then bind the device ID in the Account panel

🆘 Troubleshooting

  • Port already in use: run backend on another port and update client settings.
  • Serial port busy: fuser -k /dev/ttyACM0.
  • Linux serial permission denied: sudo usermod -aG dialout $USER and re-login.
  • Postgres unavailable: backend can run without DB for partial functionality; verify DATABASE_URL and restart.
  • Whisper performance issues: use smaller models (base/small) or disable transcription via IDEASGLASS_TRANSCRIBE=0.
  • TLS/time sync instability on ESP32: verify Wi-Fi, NTP availability (UDP/123), and cert/host settings; see references/ideasglass_bridge.md for detailed field notes.
  • No live waveform updates: check backend logs and browser console for [IdeasGlass][wave] traces and confirm /ws/stream connectivity.

🌐 Ecosystem Links

BrandPurposeLink
🧠 EchoMindMultilingual AI companion for learning and creationchat.lazying.art
🌱 OnlyIdeasResearch-to-product community for bold conceptsonlyideas.art
💸 LazyEarnAutomations to turn small wins into incomeearn.lazying.art
📚 LazyLearnPhysics & chemistry tracks and notebookslearn.lazying.art
🤖 IdeasRobotAgent that turns ideas into drafts, tasks, and postsrobot.lazying.art
👓 IdeasGlassCapture, translate, and auto-produce highlight reelsglass.lazying.art
🪙 LazyingArt CoinRewards and payouts bridging contributions and on-chain valuecoin.lazying.art
🧪 IDEASNotebook of research notes and essaysideas.onlyideas.art
🎨 LazyingArtStudio behind OnlyIdeas, EchoMind, LazyEdit, and IdeasGlasslazying.art

🙏 Acknowledgements

We stand on the shoulders of great open projects — thank you to:

🛣️ Roadmap

  • Harden and document the end-to-end audio streaming path across WAN/TLS environments.
  • Continue improving transcript quality/latency tradeoffs (model/device/threshold presets).
  • Expand device management and account-scoped multi-device workflows in the dashboard.
  • Align or consolidate legacy/parallel backend tracks (tornado_app, memo, memo_legacy, ngrok_bridge) with the primary backend/glass path.
  • Maintain and refresh multilingual README variants under i18n/.

🤝 Contribution

Contributions are welcome. For repository-specific workflow guidance, follow AGENTS.md.

Recommended local validation before opening a PR:

bash
python -m compileall backend/glass/app.py

When submitting changes:

  • Keep commit subjects short and action-driven (present tense).
  • Mention relevant env vars (for example IDEASGLASS_TRANSCRIPT_THRESHOLDS_MS) in PR notes when behavior depends on them.
  • Include testing evidence (backend logs, dashboard behavior, firmware output).
  • Never commit secrets (DATABASE_URL, API tokens, credentials files).

❤️ Support

GitHub Sponsors LazyingArt Support PayPal Support Stripe Support

📄 License

No top-level LICENSE file was detected in this repository snapshot. Until an explicit license file is added, treat usage and redistribution as requiring maintainer approval.

Parts

4
QtyValueFootprintReferences
2Header 1x13 2.54mmConnector:H11-2.54-35J1, J2
1LOGOLogo:LOGO-SEEED-NEW-2_1X19.1MMG***
1XIAO-ESP32-S3-Plus-SMDSeeed_Studio_XIAO_Series:XIAO-ESP32-S3-Plus-SMDU1

Similar boards

  • 2 layers18.5 × 26 mm24g2usb rp2350zero, the printed circuit board in 3D

    24g2usb rp2350zero

    joypad-ai/joypad-os

    Universal controller firmware for building adapters, dongles, and custom controllers.

    4 parts1.7kApache-2.0
  • 2 layers116 × 33 mm966595-1 and 963356-1, the printed circuit board in 3D

    966595-1 and 963356-1

    rusefi/rusefi

    rusefi - GPL internal combustion engine control unit

    4 parts1.2k
  • 2 layers19 × 18 mmDaughter board, the printed circuit board in 3D

    Daughter board

    hackclub/OnBoard

    💡 Join 1,000 teenagers and make your first PCB with a $100 grant!

    4 parts915
  • 2 layers24.8 × 24.8 mmNrf24 breakout, the printed circuit board in 3D

    Nrf24 breakout

    tardate/LittleArduinoProjects

    a collection of "Little Electronic & Arduino Projects", most involving electronics or an Arduino in one way or another!

    4 parts803MIT
  • 2 layers26 × 20 mmCap chain, the printed circuit board in 3D

    Cap chain

    drandyhaas/KiCadRoutingTools

    A set of Python/Rust tools to aid routing in KiCad

    4 parts744MIT
  • 2 layers34.7 × 20 mmXG Station Power MSI, the printed circuit board in 3D

    XG Station Power MSI

    osy/XG_Mobile_Station

    Open source eGPU dock for ROG Ally and ROG Flow

    4 parts712CC-BY-4.0
  • 2 layers20.3 × 24.5 mmJack power input, the printed circuit board in 3D

    Jack power input

    Luos-io/luos_engine

    Open-source and real-time orchestrator for cyber-physical-systems, to easily design, test and deploy embedded applications and digital twins.

    4 parts546MIT

Questions

How big is the XIAO Plus Base2 V1.0 PCB?

It measures 42 × 35.6 mm and has 2 copper layers, 1.6 mm thick.

What parts does the XIAO Plus Base2 V1.0 use?

4 parts. The full list of its parts is on this page, with links to each part's symbol and footprint where PCBKit's library has them.

Where can I download the XIAO Plus Base2 V1.0 KiCad files?

From the lachlanchen/IdeasGlass repository on GitHub, at the commit this page was made from: the board is seeed_studio_xiao_esp32s3_dev/hardware/base_boards/plus_base_without_pad/XIAO Plus Base2 V1.0/XIAO Plus Base2 V1.0.kicad_pcb and its schematic seeed_studio_xiao_esp32s3_dev/hardware/base_boards/plus_base_with_pad/XIAO Plus Base V1.0/XIAO Plus Base V1.0.kicad_sch. PCBKit opens the board in the browser, with nothing to install.

Can I use the XIAO Plus Base2 V1.0 design in my own project?

Its repository states no licence, so its author keeps all rights: ask lachlanchen before reusing the design.