Add Docker image

Dockerfile

@@ -0,0 +1,19 @@
+FROM pytorch/pytorch:latest
+
+RUN apt update && apt install -y && rm -rf /var/lib/apt/lists/*
+
+COPY requirements.txt .
+RUN python3 -m pip install -r requirements.txt
+
+COPY setup.py VERSION .
+COPY src ./src
+RUN python3 -m pip install .
+
+COPY "demo.py" .
+EXPOSE 7860
+COPY start.sh .
+RUN chmod +x start.sh
+
+RUN mkdir /data
+ENV TRANSFORMERS_CACHE=/data
+CMD ["./start.sh"]

README.md

@@ -24,7 +24,15 @@ python -m pip install -e .
 Set an environemnt variable called "HF_HUB_TOKEN" with your Hugging Face token
 or create a `.env` file with that env var.
 
-In one terminal, start the background worker:
+### Running the app
+
+Run the `start.sh` script. This may require a `chmod +x start.sh` if not already executable.
+
+```sh
+./start.sh
+```
+
+Instead, you can also run each part individually. For this, in one terminal, start the background worker:
 
 ```sh
 python src/gistillery/worker.py
@@ -46,6 +54,22 @@ python demo.py
 
 and navigate to the indicated URL (usually http://127.0.0.1:7860).
 
+## Docker
+
+To run everything with Docker, first build the image:
+
+```sh
+docker build -t gistillery:latest .
+```
+
+Next run the container:
+
+```sh
+docker run -p 7860:7860 -e GRADIO_SERVER_NAME=0.0.0.0 -v $HOME/.cache/huggingface/hub:/data gistillery:latest
+```
+
+Note that the Hugging Face cache folder is mounted as a docker volume to make use of potentially available local model cache instead of downloading the transformers models each time the container is started. To prevent that, remove the `-v ...` parameter. The database used for storing the results is ephemeral and will be deleted when the docker container is stopped.
+
 ## Checks
 
 ### Running tests
@@ -61,3 +85,15 @@ mypy src/
 black src/ && black tests/
 ruff src/ && ruff tests/
 ```
+
+## TODOs
+
+### Tools
+
+i. Reading pdf in general
+i. Reading arxiv
+i. Generating text from youtube videos using whisper
+
+### Deployment
+
+i. Make DB location configurable, mountable when running in docker (otherwise, it will be deleted each time the container is stopped).

src/gistillery/worker.py

@@ -1,4 +1,5 @@
 import time
+import sys
 from dataclasses import dataclass
 
 from gistillery.base import JobInput
@@ -123,4 +124,4 @@ if __name__ == "__main__":
         main()
     except KeyboardInterrupt:
         print("Shutting down...")
-        exit(0)
+        sys.exit()

start.sh

@@ -0,0 +1,21 @@
+#!/usr/bin/env bash
+
+echo "Starting background worker"
+python3 src/gistillery/worker.py &
+pid_worker=$!
+
+echo "Starting web server"
+uvicorn src.gistillery.webservice:app --port 8080 --host=0.0.0.0 &
+pid_webserver=$!
+
+# kill with ctrl-c
+trap onexit INT
+function onexit() {
+    echo "Killing background worker"
+    kill $pid_worker
+    echo "Killing web server"
+    kill $pid_webserver
+}
+
+echo "Starting gradio app on default port 7860"
+python3 demo.py