fix(install\_software.sh): update package list to include tensorrt-libs
docs(install\_software.sh): install lf terminal file manager for better navigation during the installation process
build(install\_software.sh): update the script to check if /etc/docker/daemon.json exists before configuring Nvidia container toolkit and restarting docker service
ci(install\_software.sh): add a new section for initializing Docker Swarm and deploying ollama stack using docker stack deploy command.
remove the following lines from install\_software.sh:
```bash
ffmpeg -vsync 0 -hwaccel cuda -hwaccel_output_format cuda `# we need nvidia acceleration because of low fps without it` \
--r 30 -video_size 640x480 -i /dev/video0 `# this is camera` \
--f alsa -i default -async 1 `# this is audio. Why not pulse? Nvidia?` \
--c:v h264_nvenc `# this is video codec` \
video.mp4 # output file
```
Also, remove the filter\_complex line and the x11grab section.
This refactoring is done to simplify the script and make it more focused on installing software rather than recording the screen.
This commit updates the install\_software.sh script by adding the "npm run build" command before installing the opencommit package using "sudo npm install". This change is intended to ensure that the OpenCommit package is built locally before installation, which may improve the stability and performance of the installed software.
In this commit, I updated the install\_software.sh script to include the installation of nvidia-container-toolkit, docker-compose, and opencommit. The script now checks if these packages are already installed using the `command -v` command and installs them if they are not present. Additionally, it adds a new service 'ollama' in ollama-stack.yaml file for running ollama container with the latest image.
Also, I added a new script named record\_screen.sh which records the screen and camera using ffmpeg with nvidia acceleration. The script uses h264\_nvenc as video codec to take advantage of NVIDIA GPU for encoding. It also sets the video size to 640x480 and frames per second to 30.
Lastly, I added a new file named ollama-stack.yaml which defines the services and volumes for running ollama and open-webui containers using docker-compose. The ollama service listens on port 11434 and maps the volume to /root/.ollama. The open-webui service listens on port 8080 and maps the volume to /app/backend/data. It also sets the OLLAMA\_BASE\_URL environment variable for open-webui to point to ollama container's IP address and port number.
Digital Studium