🌐 Multi-Language Support

English | Русский

GeoView

A web application for reservoir simulation, optimization, and visualization, powered by an AI agent.

Lightweight. Modern. Open source.

Features

• ECLIPSE Format Support: Parsing of standard reservoir model files
• Advanced Simulation: Fast and reliable modelling using the JutulDarcy simulator
• NPV Optimization: Automated well schedule tuning to maximize discounted Net Present Value
• Multi-Dimensional Visualization: Rich interactive analytics for static and dynamic data (3D, 2D, and 1D)
• AI-Assisted Workflows: Autonomous AI agent to streamline engineering tasks and data analysis.

Requirements

	Needed for	Note
Python >=3.13	everything
git on PATH	everything	GeoCode and GeoRead are installed from GitHub, not from PyPI
Julia	reservoir simulation	only for the Simulate and Optimize buttons
uv	the AI chat	GeoAgent runs in its own virtual environment, see Enabling the chat

Installation as a package

Create a new virtual environment with python 3.13 and install the project dependencies:

pip install "git+https://github.com/geo-kit/geoview.git"

This pulls GeoCode and GeoRead along with it. For reservoir simulation, install Julia from https://julialang.org/downloads/. Then install the JutulDarcy dependencies:

julia --project="$(python -c "import geocode, pathlib; print(pathlib.Path(geocode.__file__).parent / 'bin')")" -e "using Pkg; Pkg.instantiate()"

After installation, run in the terminal:

geoview

This should open a new tab in your default browser to http://localhost:8080/ with the application's home page.

• Contextual Help: Click the help icon in the top-right corner for a brief description of the current page.
• Tooltips: Hover over any button or icon to see its description.

The AI chat needs one more step, since the agent is a separate process pip cannot place for you: see Enabling the chat.

Installation from source code

Clone both repositories into the same directory:

git clone https://github.com/geo-kit/geoview.git
git clone https://github.com/geo-kit/geocode.git

Install dependencies in GeoCode and GeoView using

pip install -r requirements.txt

GeoRead is listed by GeoCode, so it arrives with it. For reservoir simulation, install Julia from https://julialang.org/downloads/. Then install the JutulDarcy dependencies:

julia --project=GeoCode/geocode/bin -e "using Pkg; Pkg.instantiate()"

Then navigate to the directory GeoView and run in the terminal

python -m geoview.app

to start the application.

Enabling the chat (GeoAgent)

The chat panel talks to GeoAgent, which GeoView starts and stops for you. GeoAgent runs in its own virtual environment, so it needs its own checkout whichever way you installed GeoView.

1. Clone the agent. Next to GeoView/ if you installed from source — that is where GeoView looks by default:

your-workspace/
├── GeoView/
└── GeoAgent/

git clone https://github.com/geo-kit/GeoAgent.git

If you installed GeoView as a package, put it wherever you like and see step 5.

2. Install it with uv, from inside GeoAgent/:

uv venv && uv pip install -e .

This creates the .venv that GeoView launches the agent from.

3. Add your API key. Copy .env.example to .env and fill in the key for the provider you use, one per line:

OPENAI_API_KEY=sk-...

GeoView reads this file at startup and passes the value to the agent process only.

4. Start GeoView with --agent:

geoview --agent --agent-provider openai --agent-model gpt-5-mini

The chat button appears on the Home tab. Ask about the model currently open (grid, wells, phases, whether results exist), or ask the agent to find and open a model, run the simulation, or fill in the optimization form.

5. Package installs only — point GeoView at the checkout. A packaged GeoView lives in site-packages and has no sibling directory to look in, so give it the absolute path once:

$env:GEOVIEW_AGENT_DIR = "C:\path\to\GeoAgent"
geoview --agent --agent-provider openai --agent-model gpt-5-mini

export GEOVIEW_AGENT_DIR=/path/to/GeoAgent
geoview --agent --agent-provider openai --agent-model gpt-5-mini

If GeoView exits with GeoAgent langgraph executable was not found: ..., step 2 has not been run in that directory, or GEOVIEW_AGENT_DIR points somewhere else.

You do not need to set GEOVIEW_RESULT_DIR. GeoView passes it to the agent itself; it matters only when you run the agent standalone, which GeoAgent's README covers.

Start parameters

You can add a few optional parameters to the application start command:

• --server to prevent a new tab from opening in the browser
• --app to launch the application in a separate window rather than in the browser
• --port 1234 to change the default port 8080 to, e.g., 1234
• --agent to start AI-agent
• --ru to switch the chat to Russian
• -vr or --vtk_remote to enable vtk remote rendering instead of the default local rendering using vtk webassembly.

Choosing the chat model

When started with the --agent flag, GeoView prompts for an LLM provider and model. The available options are LM Studio, Ollama, and OpenAI.

For a non-interactive launch, provide both values explicitly:

python -m geoview.app --agent `
    --agent-provider ollama `
    --agent-model qwen2.5:7b

Use --agent-base-url for a non-default local or OpenAI-compatible endpoint:

python -m geoview.app --agent `
    --agent-provider lmstudio `
    --agent-base-url http://127.0.0.1:1234/v1 `
    --agent-model some-model-id

The matching environment variables are GEOVIEW_AGENT_PROVIDER, GEOVIEW_AGENT_MODEL, and GEOVIEW_AGENT_BASE_URL.

Cloud API keys are read from the environment (OPENAI_API_KEY). If a key is not already set there, GeoView loads it from GeoAgent/.env, which is why step 3 above is enough. GeoView does not start local model servers or pull models.

Rendering options

By default, rendering is performed locally in the user's browser using vtk webassembly functionality. To enable vtk remote rendering, use the -vr or --vtk_remote option when starting the application. Note that the functionality of the application is slightly different between local and remote rendering.

Open-source reservoir models

An example reservoir model with dynamics simulation can be found in the open_data directory in the GeoCode repository https://github.com/geo-kit/GeoCode, as well as links to a number of other open source models.

Script writing

The application allows you to write and execute python scripts for reservoir model transformations and calculations. The script should be based on the GeoCode framework https://github.com/geo-kit/GeoCode. Read the documentation and see examples in the GeoCode repository to prepare a script.

Next releases

The project is developing. We are preparing new releases with new features. Your suggestions and issues reports will help to make the application even better.

What's inside

We use

• trame to build the web application
• GeoRead to read and GeoCode to process reservoir models
• JutulDarcy for reservoir simulation
• GeoAgent for AI-agent assistance

Citing

We hope that this project will help you in your research and you will decide to cite it as

GeoView web application (2026). GitHub repository, https://github.com/geo-kit/GeoView.