Nosia is a platform that allows you to run an AI model on your own data. It is designed to be easy to install and use.

You can follow this README or go to the Nosia Guides.

• Quickstart
• API
• Upgrade
• Start
• Stop
• Troubleshooting

It will install Docker, Ollama, and Nosia on a macOS, Debian or Ubuntu machine.

curl -fsSL https://raw.githubusercontent.com/nosia-ai/nosia-install/main/nosia-install.sh | sh

You should see the following output:

✅ Setting up environment
✅ Setting up Docker
✅ Setting up Ollama
✅ Starting Ollama
✅ Starting Nosia

You can now access Nosia at https://nosia.localhost

By default, Nosia sets up ollama locally.

To use a remote Ollama instance, set the OLLAMA_BASE_URL environment variable during configuration.

Example:

Replace $OLLAMA_HOST_IP with the FQDN or IP address of your Ollama host and run:

curl -fsSL https://raw.githubusercontent.com/nosia-ai/nosia-install/main/nosia-install.sh \
  | OLLAMA_BASE_URL=http://$OLLAMA_HOST_IP:11434 sh

By default, Nosia uses:

1. Completion model: qwen2.5
2. Embeddings model: nomic-embed-text
3. Checking model: bespoke-minicheck

You can use any completion model available on Ollama by setting the LLM_MODEL environment variable during the installation.

Example:

To use the mistral model, run:

curl -fsSL https://raw.githubusercontent.com/nosia-ai/nosia-install/main/nosia-install.sh \
  | LLM_MODEL=mistral sh

At this time, the nomic-embed-text embeddings model is required for Nosia to work.

On macOS, install Homebrew:

/bin/bash -c "$(curl -fsSL https://raw.githubusercontent.com/Homebrew/install/HEAD/install.sh)"

Then install Ollama with Homebrew:

Replace $OLLAMA_HOST_IP with the IP address of the Ollama host machine and run the following command:

brew install ollama
ollama pull qwen2.5
ollama pull bespoke-minicheck
ollama pull nomic-embed-text
OLLAMA_BASE_URL=$OLLAMA_HOST_IP:11434 OLLAMA_MAX_LOADED_MODELS=3 ollama serve

On the Debian/Ubuntu VM:

Replace $OLLAMA_HOST_IP with the IP address of the host machine and run the following command:

curl -fsSL https://raw.githubusercontent.com/nosia-ai/nosia-install/main/nosia-install.sh \
  | OLLAMA_BASE_URL=http://$OLLAMA_HOST_IP:11434 sh

You should see the following output:

✅ Setting up environment
✅ Setting up Docker
✅ Setting up Ollama
✅ Starting Ollama
✅ Starting Nosia

From the VM, you can access Nosia at https://nosia.localhost

If you want to access Nosia from the host machine, you may need to forward the port from the VM to the host machine.

Replace $USER with the username of the VM, $VM_IP with the IP address of the VM, and $LOCAL_PORT with the port you want to use on the host machine, 8443 for example, and run the following command:

ssh $USER@$VM_IP -L $LOCAL_PORT:localhost:443

After running the command, you can access Nosia at https://nosia.localhost:$LOCAL_PORT.

1. Go as a logged in user to https://nosia.localhost/api_tokens
2. Generate and copy your token
3. Use your favorite OpenAI chat completion API client by configuring API base to https://nosia.localhost/v1 and API key with your token.

Follow the guide

You can upgrade the services with the following command:

./script/upgrade

You can start the services with the following command:

./script/start

You can stop the services with the following command:

./script/stop

If you encounter any issue:

• during the installation, you can check the logs at ./log/production.log
• during the use waiting for an AI response, you can check the jobs at http://<IP>:3000/jobs
• with Nosia, you can check the logs with docker compose -f ./docker-compose.yml logs -f
• with the Ollama server, you can check the logs at ~/.ollama/logs/server.log

If you need further assistance, please open an issue!