Glimpser

Introduction

Glimpser is a straightforward yet powerful real-time monitoring application designed to capture, analyze, and summarize live data from various sources such as cameras, dashboards, and video streams. Utilizing advanced image processing techniques and AI models, Glimpser provides insightful summaries and alerts. It’s highly configurable, allowing users to tailor it to their specific monitoring needs through an easy-to-use interface.

For more documentation, see the documentation index. You can find a summary of the documentation structure in docs/README.md. Read a high-level Architecture Overview to understand how the pieces fit together. See OpenSSF Scorecard for details on the security badge.

Features

• Real-time Monitoring: Continuously captures data from multiple sources. Whether it's a traffic camera or a weather dashboard, Glimpser ensures you’re always up-to-date with the latest information.

• Image Processing: Employs advanced techniques to compare images and detect even subtle changes, making it ideal for monitoring evolving situations effectively.

• Motion Detection: Automatically detects motion in the captured images and videos, triggering alerts and actions as configured by the user.

• AI Integration: Integrates with models like ChatGPT to provide intelligent insights. It can summarize data, detect anomalies, and generate alerts based on predefined rules.

• Auto-captioning: Automatically generates concise and informative captions for images and videos, providing quick insights into the content.

• Auto-summarization: Summarizes data from multiple sources into a coherent and concise format, highlighting the most important information.

• RTSP Streaming: Exposes a basic RTSP endpoint (/test.rtsp) so external NVRs can ingest the MJPEG stream. Supported verbs are OPTIONS, DESCRIBE, SETUP, PLAY, PAUSE, GET_PARAMETER, and TEARDOWN.

• MJPG Test Stream: Provides /test.mjpg for clients that lack RTSP support.

• Customizable Configuration: Easily configure different data sources and processing rules through the user-friendly interface. Glimpser’s configuration is fully database-driven, ensuring flexibility and ease of use.

• Data Retention Policies: Automatically manages storage by cleaning up old data, ensuring the system remains efficient without requiring constant manual intervention.

• HTTP Callbacks: When a template specifies a callback URL, Glimpser sends a JSON webhook with caption or motion updates to that endpoint. See the HTTP Callback Guide for setup details and payload examples.

• SMS Alerts: Configure Twilio credentials to receive important notifications by text message.

• Camera Discovery: Open Settings and switch to the Discover tab to automatically scan the local network for ONVIF, RTSP, RTMP, HTTP/MJPEG, HLS, and SSDP devices. The table now displays each camera's MAC address plus manufacturer and model information when available. Glimpser checks common system OUI databases, an online lookup service, and the ONVIF device service to gather these details.

• Camera Fix Suggestions: Validate a template and discover alternative URLs with /suggest_fix/<template>. See Camera Fix Suggestions.

• Local Cameras: The Discover tab also lists any available /dev/video* devices for easy webcam integration.

• MCP Integration: Delegate actions to an external control plane. See the MCP Integration Guide.

• Web Interface: A user-friendly web interface allows for easy monitoring and configuration. Users can view live feeds, summaries, and configure settings without delving into the code.

Installation

Prerequisites

• Python 3.8 to 3.13

Steps

1. Install the Package

sh pip install glimpser

Or, if you want to install from source:

sh git clone https://github.com/KristopherKubicki/glimpser.git cd glimpser pip install .

1. Run the Application

sh glimpser

You can pass command-line options to customize the runtime configuration. The most common flags are:

```sh

Start without the background scheduler

glimpser --no-scheduler

Skip scheduling crawler jobs

glimpser --no-crawlers

Disable the watchdog thread

glimpser --no-watchdog ```

Run glimpser --help to see all available options.

For a full description of every command-line flag, including the separate credentials utility, see docs/command_line.md.

You will be prompted to create a secret key to initialize the local sqlite database. Follow the rest of the guided setup and then direct your browser to http://127.0.0.1:8082 to finish the rest of the setup.

Docker Quick Start

If you prefer to run Glimpser in Docker, copy .env.example to .env and set at least SECRET_KEY and API_KEY. You may also adjust SESSION_TIMEOUT_MINUTES and related cookie settings. The timeout resets after each request, so active use keeps the session alive. Then build and start the container using Gunicorn:

sh docker-compose up --build

The web interface will be available at http://localhost:8082.

Common Setup Issues

If you cannot log in or see video feeds, double-check that your .env file matches the configuration values in the database. Missing SECRET_KEY or API credentials often cause startup failures. Refer to Troubleshooting for more solutions.

Developer Dependencies

To install Python packages required for development, run:

sh uv sync --group dev

Then install linters and JavaScript tools with make setup (or scripts/setup_env.sh).

Usage

Configuration

Glimpser uses a database-driven configuration to manage data sources and processing rules. Users can easily add, update, or remove configurations through the web interface. SMS alerts can be enabled by setting TWILIO_SID, TWILIO_TOKEN, and TWILIO_NUMBER in the configuration.

Capturing Screenshots

The preferred method for capturing screenshots is through the Glimpser web interface. Simply navigate to the capture section, select your desired source, and click the capture button. This ensures a seamless and user-friendly experience.

Adding a Camera Source

Open the Discover tab under Settings to scan your network for ONVIF, RTSP, or local devices and click Add next to any result. You can also open Add Source in the web interface to manually supply a camera URL and group. See Camera Discovery for more details.

Running Tests

To ensure everything works as expected, you can run the included unit tests:

sh python -m coverage run -m pytest

Troubleshooting Quick Tips

If the summarizer stops working, ensure the scheduler is running and check /jobs for a summary entry. Review logs/glimpser.log for errors. More solutions are listed in Troubleshooting.

Motion Detection

Glimpser automatically detects motion in the captured images and videos. When motion is detected, the system can trigger alerts, capture additional data, and generate relevant summaries and captions.

Auto-captioning

Using advanced AI models, Glimpser generates concise and informative captions for images and videos. This feature helps users quickly understand the content and context of the captured data.

Auto-summarization

Glimpser can summarize data from multiple sources into a coherent and concise format. The summaries highlight the most important information, making it easier for users to stay informed.

RTSP Streaming

Glimpser exposes a simple RTSP endpoint at /test.rtsp. When a client issues the standard RTSP verbs, the /rtsp_stream route serves MJPEG frames packetized with RTP headers. An equivalent MJPEG feed is available at /test.mjpg for quick testing or clients without RTSP support.

Typical sequence:

1. OPTIONS
2. DESCRIBE
3. SETUP
4. PLAY
5. (optional) PAUSE / PLAY
6. Periodic GET_PARAMETER to keep the session alive
7. TEARDOWN to close the session

This allows external NVR software to ingest the stream as a basic camera source.

Development

To set up the project for development:

1. Clone the repository:

sh git clone https://github.com/KristopherKubicki/glimpser.git cd glimpser

1. Create and activate a virtual environment:

sh python -m venv env source env/bin/activate # On Windows use `env\Scripts\activate`

1. Install Python and JavaScript dependencies:

sh uv sync --group dev npm install

1. Install Git hooks and additional tooling:

sh uv run pre-commit install make setup # optional helper to configure tools

1. Verify hooks and run linters:

sh uv run pre-commit run --all-files uv run flake8 uv run pytest npm test -- --coverage

See Developer Guide and Testing for more details.

Releases

Release packages are built automatically when a version tag is pushed. The workflow runs tests and creates the Debian package on an Ubuntu runner. The Windows executable is built on a Windows runner with build_windows.py, and a macOS archive is produced with build_macos.py. When all steps finish, the resulting Debian package and platform binaries are uploaded to the GitHub release for that tag. If the repository contains a PYPI_API_TOKEN secret, the workflow also builds Python distributions and publishes them to PyPI. These files can be downloaded from the Releases page or installed with pip.

If the Releases page shows an older version than the footer, ensure the version tag was pushed. You can run scripts/auto_tag_release.py to create and push the tag manually.

Contributing

Contributions are always welcome. If you have an idea to improve Glimpser, feel free to fork the repository and submit a pull request. Please read our Code of Conduct to understand the expectations for participants and how to report issues.

Steps to Contribute

1. Fork the repository.
2. Create a feature branch. sh git checkout -b feature-branch
3. Commit your changes. sh git commit -m "Description of changes"
4. Push to the branch. sh git push origin feature-branch
5. Open a pull request.

Branching Workflow

All changes start in a feature branch and merge into staging first. The main branch is always deployable and receives commits only from the CD pipeline. Direct pushes to main are blocked by a branch protection workflow. See Staged GitHub Flow for a detailed diagram.

Recommendations

For detailed suggestions on how to use Glimpser effectively, please check out our Recommendations guide.

License

This project is licensed under the MIT License. See the LICENSE.md file for details.

Acknowledgements

We are grateful to the contributors and the open-source community. Special thanks to OpenAI for their powerful models that enable Glimpser's advanced features.