This is the pipeline which is spawned for a single client
Install the requirements.txt in your venv in the folder above using
pip install -r requirements.txt
Create an environment file with the following content in the root folder of the pipeline:
CLOUDFLARE_API_KEY=""
CLOUDFLARE_ACCOUNT_ID=""
IMAGE_CLASSIFICATION_MODEL="@cf/microsoft/resnet-50"
OBJECT_DETECTION_MODEL="@cf/facebook/detr-resnet-50"
IMAGE_TO_TEXT_MODEL="@cf/llava-hf/llava-1.5-7b-hf" # in beta
LARGE_LANGUAGE_MODEL="@cf/meta/llama-3-8b-instruct" # in beta
One of the two commandline options needs to specified to run the pipeline
--uuidspecify the UUID of the upstream--devuses local images, need to provide directory--fastenables fast mode in the backend, which removes the LLM summary and leads to only the image 2 text message being sent and voiced in the frontend
You can use the directories in development for a more consistent experience