Welcome to Lunar Tools, a comprehensive toolkit designed to facilitate the programming of interactive exhibitions. Our suite of simple, modular tools is crafted to offer a seamless and hopefully bug-free experience for both exhibitors and visitors.

Make sure you have python>=3.10.

pip install lunar-tools

Sound playback via SoundPlayer requires simpleaudio, which is not installed by default. Install it manually if you need in-process playback:

pip install simpleaudio

For PyTorch-accelerated features (GPU rendering, torch tensor support):

pip install lunar-tools[torch]

pip install git+https://github.com/lunarring/lunar_tools

On Ubuntu, you may have to install additional dependencies for sound playback/recording.

sudo apt-get install libasound2-dev libportaudio2

For running the midi controllers, you might have to create a symlink:

cd /usr/lib/x86_64-linux-gnu/
sudo ln -s alsa-lib/libasound_module_conf_pulse.so libasound_module_conf_pulse.so

Many features in lunar-tools require API keys from various services. Set these as environment variables in your shell configuration file (e.g. ~/.bash_profile or ~/.zshrc or ~/.bashrc):

# AI/LLM Services
export OPENAI_API_KEY="XXX"          # For GPT models, DALL-E, TTS, and STT
export GEMINI_API_KEY="XXX"          # For Google Gemini models
export DEEPSEEK_API_KEY="XXX"        # For DeepSeek models

# Image Generation
export REPLICATE_API_TOKEN="XXX"     # For Replicate (SDXL models)
export FAL_KEY="XXX"                 # For fal.ai (Flux models)
export GLIF_API_KEY="XXX"            # For Glif API

# Audio/Voice Services
export ELEVEN_API_KEY="XXX"          # For ElevenLabs TTS
export DEEPGRAM_API_KEY="XXX"        # For Deepgram speech-to-text

# Health Monitoring
export TELEGRAM_BOT_TOKEN="XXX"      # For Telegram health reporting
export TELEGRAM_CHAT_ID="XXX"        # Your Telegram chat ID

Only add the keys for services you plan to use. Each module will let you know if a required API key is missing.

Runnable input snippets live in examples/inputs. Launch them from the repo root to validate your hardware and copy/paste the relevant code into your own project.

examples/inputs/audio_recorder_example.py exposes lt.AudioRecorder through two CLI flags so you can verify your microphone pipeline without touching code.

python examples/inputs/audio_recorder_example.py --seconds 5 --output myvoice.mp3

examples/inputs/webcam_live_renderer.py pairs lt.WebCam with lt.Renderer and displays a live preview window for whichever camera ID (or auto-probed device) you pass in.

python examples/inputs/webcam_live_renderer.py --cam-id auto

examples/inputs/meta_input_inspector.py uses lt.MetaInput to detect a MIDI controller (or keyboard fallback) and continuously prints one slider + one button so you can confirm your mappings on the spot.

python examples/inputs/meta_input_inspector.py

examples/inputs/movie_reader_example.py wraps lt.MovieReader with a CLI so you can inspect frame shapes, counts, and FPS before embedding any mp4 into your pipeline.

python examples/inputs/movie_reader_example.py my_movie.mp4 --max-frames 10

Runnable output demos live in examples/outputs. Each script is a ready-to-run showcase that you can copy into your own pipeline or execute as-is.

examples/outputs/sound_playback_generated_sine.py demonstrates lt.SoundPlayer by first writing a generated 440 Hz sine to disk, then streaming a 660 Hz tone directly from memory via play_audiosegment.

python examples/outputs/sound_playback_generated_sine.py

examples/outputs/display_multi_backend_example.py spins up lt.Renderer and cycles through NumPy, Pillow, and Torch backends (whichever are installed) to render random RGBA frames in one looping window.

python examples/outputs/display_multi_backend_example.py

note you can speed-up opengl render calls by upto a factor of 3 by disabling VSYNC on your system On Ubuntu do: Run nvidia-settings 2. Screen 0 > OpenGl > Sync to VBLank -> off

examples/outputs/realtime_console_updates_example.py combines lt.FPSTracker, lt.LogPrint, and dynamic_print to stream live progress messages while measuring per-segment timings.

python examples/outputs/realtime_console_updates_example.py

examples/outputs/logprint_example.py showcases lt.LogPrint formatting, highlighting how to stream colored, timestamped console output.

python examples/outputs/logprint_example.py

examples/outputs/movie_saver_example.py creates a short mp4 using random RGB frames so you can validate codec support and file permissions.

python examples/outputs/movie_saver_example.py --output my_movie.mp4 --frames 10 --fps 24

lunar_tools.comms.get_local_ip inspects network interfaces to determine the best IP to share with peers. Run the example below to print the detected address or see a friendly warning if one cannot be determined (for example, on air-gapped machines).

python examples/comms/get_local_ip_example.py

Low-latency data channel built on WebRTC for streaming numpy arrays, JSON blobs, PNG previews, and log text.

Sender (hosts an embedded signaling server and streams mixed payloads):

python examples/comms/webrtc_data_sender.py --session demo

Receiver (auto-discovers the sender session via the cached signaling endpoint):

python examples/comms/webrtc_data_receiver.py --session demo

• --sender-ip defaults to the detected local address (via lunar_tools.comms.utils.get_local_ip).
• When the sender hosts the embedded signaling server it stores the endpoint details per session in ~/.lunar_tools/webrtc_sessions.json. Receivers can omit --sender-ip to reuse the most recent entry for the requested session, which keeps the bootstrap process simple.
• If you prefer using your own signaling server, start it separately (or pass --no-server in the sender example) and point both peers to the same http://<sender-ip>:<port> URL.

High-level OSC helper built on python-osc. The receiver example spawns the live grid visualizer, and the sender emits demo sine/triangle waves.

Receiver:

python examples/comms/osc_receiver.py --ip 0.0.0.0 --port 8003

Sender:

python examples/comms/osc_sender.py --ip 127.0.0.1 --port 8003 --channels /env1 /env2 /env3

One-to-one ZeroMQ stream that carries JSON blobs, compressed images, and raw PCM audio. Start the receiver first on the same machine (or pass --ip 0.0.0.0 if you want to accept remote peers), then launch the sender.

Receiver (binds locally):

python examples/comms/zmq_receiver.py --port 5556

Sender (connects to the receiver):

python examples/comms/zmq_sender.py --ip 127.0.0.1 --port 5556

ZMQPairEndpoint uses ZeroMQ's PAIR pattern, which is strictly one-to-one: exactly one sender and one receiver must be connected, and neither side can reconnect while the other is running. If you need fan-out/fan-in or resilient reconnection, prefer REQ/REP, PUB/SUB, or ROUTER/DEALER and stitch together the behavior you need on top of the raw zmq library.

Voice-focused demos live in examples/voice. Each script below can be run directly from the repo root and pairs with the API snippets that follow.

examples/voice/realtime_voice_example.py is an interactive CLI that lets you start/pause/mute a RealTimeVoice session, inject messages, and update instructions on the fly.

python examples/voice/realtime_voice_example.py

examples/voice/deepgram_realtime_transcribe_example.py uses lt.RealTimeTranscribe (Deepgram SDK) to stream microphone audio and print live transcripts. Set DEEPGRAM_API_KEY before running.

python examples/voice/deepgram_realtime_transcribe_example.py

examples/voice/openai_speech_to_text_example.py records a short microphone clip and prints the transcript, with an optional flag to save the text to disk.

python examples/voice/openai_speech_to_text_example.py --seconds 5 --output transcript.txt

examples/voice/openai_text_to_speech_example.py converts text to speech, saves it to an mp3, and can optionally stream the audio immediately with --play-inline.

python examples/voice/openai_text_to_speech_example.py --text "Testing 1 2 3" --voice nova --play-inline

examples/voice/elevenlabs_text_to_speech_example.py targets the ElevenLabs API with inline playback plus flags for stability, similarity, style, and speaker boost.

python examples/voice/elevenlabs_text_to_speech_example.py --text "Hi from ElevenLabs" --voice-id EXAVITQu4vr4xnSDxMaL --play-inline

examples/ai/dalle3_generate_example.py calls lt.Dalle3ImageGenerator, saves the resulting PNG, and prints the revised prompt.

python examples/ai/dalle3_generate_example.py --prompt "A red house with snow and a chimney"

examples/ai/sdxl_turbo_example.py uses the Replicate-powered lt.SDXL_TURBO helper and stores the PNG plus source URL for reference.

python examples/ai/sdxl_turbo_example.py --prompt "An astronaut riding a rainbow unicorn" --width 768 --height 512

examples/ai/nano_banana_edit_gradio.py launches a Gradio UI for interactive Flux/Nano Banana edits—drop in prompts, tweak sliders, and preview changes.

python examples/ai/nano_banana_edit_gradio.py

Obtain a bot here: https://docs.tracardi.com/qa/how_can_i_get_telegram_bot/ Next you will need to update your bashrc or bash_profile with the telegram bot env variables.

export TELEGRAM_BOT_TOKEN='XXX'
export TELEGRAM_CHAT_ID='XXX'

See examples/health/telegram_health_reporter_example.py for a runnable heartbeat + alert demo (requires the env vars above):

python examples/health/telegram_health_reporter_example.py --name "My Exhibit" --interval 2 --count 5

pip install pytest

make sure you are in base folder

python -m pytest lunar_tools/tests/