Local Open-source micro-agents that observe, log and react, all while keeping your data private and secure.

Observer App Link

• Support me and the project!

An open-source platform for running local AI agents that observe your screen while preserving privacy.

Creating your own Observer AI agent is simple, and consist of three things:

• SENSORS - input that your model will have
• MODELS - models run by ollama or by Ob-Server
• TOOLS - functions for your model to use

1. Navigate to the Agent Dashboard and click "Create New Agent"
2. Fill in the "Configuration" tab with basic details (name, description, model, loop interval)
3. Give your model a system prompt and Sensors! The current Sensors that exist are:
   • Screen OCR ($SCREEN_OCR) Captures screen content as text via OCR
   • Screenshot ($SCREEN_64) Captures screen as an image for multimodal models
   • Agent Memory ($MEMORY@agent_id) Accesses agents' stored information
   • Clipboard ($CLIPBOARD) It pastes the clipboard contents
   • Microphone* ($MICROPHONE) Captures the microphone and adds a transcription
   • Screen Audio* ($SCREEN_AUDIO) Captures the audio transcription of screen sharing a tab.
   • All audio* ($ALL_AUDIO) Mixes the microphone and screen audio and provides a complete transcription of both (used for meetings).

* Uses a whisper model with transformers.js (only supports whisper-tiny english for now)

4. Decide what tools do with your models response in the Code Tab:

• notify(title, options) – Send notifications
• getMemory(agentId)* – Retrieve stored memory (defaults to current agent)
• setMemory(agentId, content)* – Replace stored memory
• appendMemory(agentId, content)* – Add to existing memory
• startAgent(agentId)* – Starts an agent
• stopAgent(agentId)* – Stops an agent
• time() - Gets current time
• sendEmail(content, email) - Sends an email
• sendSms(content, phone_number) - Sends an SMS to a phone number, format as e.g. sendSms("hello",+181429367")
• sendWhatsapp(content, phone_number) - Sends a whatsapp message, IMPORTANT: temporarily to counter anti spam, Observer is sending only static messages disregarding "content" variable.
• startClip() - Starts a recording of any video media and saves it to the recording Tab.
• stopClip() - Stops an active recording
• markClip(label) - Adds a label to any active recording that will be displayed in the recording Tab.

The "Code" tab now offers a notebook-style coding experience where you can choose between JavaScript or Python execution:

JavaScript agents run in the browser sandbox, making them ideal for passive monitoring and notifications:

// Remove Think tags for deepseek model
const cleanedResponse = response.replace(/<think>[\s\S]*?<\/think>/g, '').trim();

// Preserve previous memory
const prevMemory = await getMemory();

// Get time
const time = time();

// Update memory with timestamp
appendMemory(`[${time}] ${cleanedResponse}`);

Note: any function marked with * takes an agentId argument.
If you omit agentId, it defaults to the agent that’s running the code.

Python agents run on a Jupyter server with system-level access, enabling them to interact directly with your computer:

#python <-- don't remove this!
print("Hello World!", response, agentId)

# Example: Analyze screen content and take action
if "SHUTOFF" in response:
    # System level commands can be executed here
    import os
    # os.system("command")  # Be careful with system commands!

The Python environment receives:

• response - The model's output
• agentId - The current agent's ID

To use Python agents:

1. Run a Jupyter server on your machine with c.ServerApp.allow_origin = '*'
2. Configure the connection in the Observer AI interface:
   • Host: The server address (e.g., 127.0.0.1)
   • Port: The server port (e.g., 8888)
   • Token: Your Jupyter server authentication token
3. Test the connection using the "Test Connection" button
4. Switch to the Python tab in the code editor to write Python-based agents

✨ Major Update: Simpler Setup & More Flexibility! The observer-ollama service no longer requires SSL by default. This means no more browser security warnings for a standard local setup! It now also supports any backend that uses a standard OpenAI-compatible (v1/chat/completions) endpoint, like Llama.cpp.

There are a few ways to get Observer up and running with local inference. I recommend using Docker for the simplest setup.

This method uses Docker Compose to run everything you need in containers: the Observer WebApp, the observer-ollama translator, and a local Ollama instance. This is the easiest way to get a 100% private, local-first setup.

Prerequisites:

• Docker installed.
• Docker Compose installed.

Instructions:

1. Clone the repository and start the services:

   git clone https://github.com/Roy3838/Observer.git
   cd Observer
   docker-compose up --build

2. Access the Local WebApp:

   • Open your browser to http://localhost:8080. This is your self-hosted version of the Observer app.

3. Connect to your Ollama service:

   • In the app's header/settings, set the Model Server Address to http://localhost:3838. This is the observer-ollama translator that runs in a container and communicates with Ollama for you.

4. Pull Ollama Models:

   • Navigate to the "Models" tab and click "Add Model". This opens a terminal to your Ollama instance.
   • Pull any model you need, for example:

     ollama run gemma3:4b # <- highly recommended model!

For NVIDIA GPUs: it's recommended to edit docker-compose.yml and explicitly add gpu runtime to the ollama docker container. Add these to the ollama section of docker-compose.yml:

    volumes:
      - ollama_data:/root/.ollama
    # ADD THIS SECTION
    runtime: nvidia
    deploy:
      resources:
        reservations:
          devices:
            - driver: nvidia
              count: all
              capabilities: [gpu]
    # UP TO HERE
    ports:
      - "11434:11434"

To Stop the Docker Setup:

docker-compose down

Running an inference server on another device on the network is fully supported, just place the address to the OpenAI-compatible server from within the self-hosted webpage.

If you want to access your self-hosted Observer from another device that can't self host its own webpage (phones or tablets), you need to make one small change to disable authentication. This is due to Auth0 strict origin policy.

1. Edit the docker-compose.yml file:

   # In docker-compose.yml
   services:
     observer_app:
       build:
         context: .
         dockerfile: Dockerfile
         args:
           # Set this value to true to disable authentication
           - VITE_DISABLE_AUTH=true

2. Restart Docker:

   docker-compose up --build --force-recreate

   You can now access Observer by navigating to http://<YOUR_PC_IP_ADDRESS>:8080 on your other device.

Observer can connect directly to any server that provides a v1/chat/completions endpoint.

Prerequisites:

• Node.js v18+ (which includes npm).
• An already running OpenAI-compatible model server.

1. Self-host the WebApp: with run script

   git clone https://github.com/Roy3838/Observer
   cd Observer
   chmod +x run.sh
   ./run.sh
   

2. Run your Llama.cpp server:

   # Example command
   ./server -m your-model.gguf -c 4096 --host 0.0.0.0 --port 8001

3. Connect Observer: In the Observer app (http://localhost:8080), set the Model Server Address to your Llama.cpp server's address (e.g., http://127.0.0.1:8001).

Same as docker, just change the disable auth variable on run.sh to true:

# To disable auth
export VITE_DISABLE_AUTH=true 

Save your agent, test it from the dashboard, and export the configuration to share with others!

We welcome contributions from the community! Here's how you can help:

1. Fork the repository
2. Create your feature branch (git checkout -b feature/amazing-feature)
3. Commit your changes (git commit -m 'feat: add amazing feature')
4. Push to the branch (git push origin feature/amazing-feature)
5. Open a Pull Request

This project is licensed under the MIT License - see the LICENSE file for details.

• Website
• GitHub Repository
• twitter

• GitHub: @Roy3838
• Project Link: https://observer-ai.com

Built with ❤️ by Roy Medina for the Observer AI Community Special thanks to the Ollama team for being an awesome backbone to this project!