Merge pull request #41 from hpicgs/deployment-changes

Finish Repository Deployment

.streamlit/config.toml

@@ -0,0 +1,5 @@
+[theme]
+base="light"
+primaryColor="#B1063A"
+backgroundColor="#F1F1F1"
+secondaryBackgroundColor="#D4DADE"

Dockerfile

@@ -0,0 +1,22 @@
+# Builder image
+FROM continuumio/miniconda3 as builder
+
+RUN apt-get update -y && apt-get install default-jre make wget curl unzip build-essential -y
+COPY . /app/unCover
+
+WORKDIR /app/unCover
+RUN conda update conda --yes
+RUN conda env create -f environment.yml
+RUN cp -n .env.example .env
+RUN make -C tem/topic-evolution-model/
+RUN conda run -n unCover ./corenlp --no-run && conda run -n unCover ./prepare_models
+
+# Final image
+FROM continuumio/miniconda3
+LABEL authors="lucasliebe"
+
+COPY --from=builder /app /app
+COPY --from=builder /opt/conda /opt/conda
+
+WORKDIR /app/unCover
+CMD conda run -n unCover streamlit run main.py

README.md

@@ -2,19 +2,29 @@
 
 Detailed information about unCover can be found in the following publication:
 
-> Liebe L, Baum J, Schutze T, Cech T, Scheibel W, and Dollner J (2023). UNCOVER:
-> Identifying AI Generated News Articles by Linguistic Analysis and
-> Visualization
+> Liebe, L.; Baum, J.; Schütze, T.; Cech, T.; Scheibel, W. and Döllner, J. (2023). 
+> UNCOVER: Identifying AI Generated News Articles by Linguistic Analysis and Visualization.  
+> In Proceedings of the 15th International Joint Conference on Knowledge Discovery, 
+> Knowledge Engineering and Knowledge Management - Volume 1: KDIR, ISBN 
+> 978-989-758-671-2, ISSN 2184-3228, pages 39-50.    
 
-![Teaser](https://drive.google.com/uc?export=download&id=1i49F16U7TiHCS8-17lBv8ofPsnvd-RE0)
+![Teaser](https://drive.google.com/uc?export=download&id=1DU9HwazIUGxoFdI5cJ-liW3Q_-a-QV6G)
 
 An interactive example deployment of unCover can be found at 
-[uncover.streamlit.app](https://uncover.streamlit.app).
+[uncover.lucasliebe.de](https://uncover.lucasliebe.de).
 Our datasets and pre-trained models can be found in 
 [Google Drive](https://drive.google.com/drive/folders/1fMZgGC2Bnp5K-ZoANXB_S0AI02akye_c?usp=drive_link).
 Please note that for copyright reasons we removed the plain text of the scraped 
 news articles and only left the metadata and the generated texts in the dataset files.
 
+## Prerequisites
+
+Before you can use the installation script as described below, please make sure
+you have the following packages installed and working on your machine:
+- Anaconda or Miniconda
+- Java (Runtime Environment is sufficient)
+- Make and g++
+
 ## Setup
 
 To set up this project to run on your own machine, run the following command
@@ -33,11 +43,15 @@ activate the environment with
 conda activate unCover
 ```
 
-To take full advantage of all capabilities in this repository you should update 
-all the information in `.env`. OpenAI-credentials are
+To take full advantage of all capabilities in this repository you should fill out 
+all the information in `.env.example` and save it as `.env`. OpenAI-credentials are
 only required for generation, however, fine-tuning the confidence thresholds to the 
 used models will greatly benefit performance and is required to achieve good results.
 
+Alternatively if you are only interested as running the web interface you can use
+the provided docker container for a quick deployment. It can be built yourself using 
+the Dockerfile or pulled from Docker Hub: `docker pull lucasliebe/uncover:latest`.
+
 ## Usage
 
 There are multiple ways to use unCover, depending on your use case.

definitions.py

@@ -25,7 +25,7 @@
 GPT_KEY = os.getenv("GPT_KEY", "")
 OPENAI_ORGA = os.getenv("OPENAI_ORGA", "")
 
-CHAR_MACHINE_CONFIDENCE = float(c) if (c := os.getenv("CHAR_MACHINE_CONFIDENCE")) else st.secrets["CHAR_MACHINE_CONFIDENCE"]
-CHAR_HUMAN_CONFIDENCE = float(c) if (c := os.getenv("CHAR_HUMAN_CONFIDENCE")) else st.secrets["CHAR_HUMAN_CONFIDENCE"]
-SEM_MACHINE_CONFIDENCE = float(c) if (c := os.getenv("SEM_MACHINE_CONFIDENCE")) else st.secrets["SEM_MACHINE_CONFIDENCE"]
-SEM_HUMAN_CONFIDENCE = float(c) if (c := os.getenv("SEM_HUMAN_CONFIDENCE")) else st.secrets["SEM_HUMAN_CONFIDENCE"]
+CHAR_MACHINE_CONFIDENCE = float(os.getenv("CHAR_MACHINE_CONFIDENCE", ""))
+CHAR_HUMAN_CONFIDENCE = float(os.getenv("CHAR_HUMAN_CONFIDENCE", ""))
+SEM_MACHINE_CONFIDENCE = float(os.getenv("SEM_MACHINE_CONFIDENCE", ""))
+SEM_HUMAN_CONFIDENCE = float(os.getenv("SEM_HUMAN_CONFIDENCE", ""))

install.sh

@@ -4,16 +4,17 @@ print_bold() {
     printf "\\033[1m$1\\033[0m\n"
 }
 
-if ! which java 1>/dev/null 2>/dev/null; then
-    print_bold "Please ensure Java is installed and on your PATH."
-    exit 1
-fi
 
 if ! which conda 1>/dev/null 2>/dev/null; then
     print_bold "Please ensure Anaconda is installed and on your PATH."
     exit 1
 fi
 
+if ! which java 1>/dev/null 2>/dev/null; then
+    print_bold "Please ensure Java is installed and on your PATH."
+    exit 1
+fi
+
 if ! which make 1>/dev/null 2>/dev/null; then
     print_bold "Please ensure Make is installed and on your PATH."
     exit 1
@@ -23,18 +24,16 @@ print_bold "Cloning repository"
 git clone --recurse-submodules https://github.com/hpicgs/unCover.git
 cd unCover
 
-print_bold "Compiling TEM"
-make -C tem/topic-evolution-model/
-
 print_bold "Creating Anaconda environment"
 conda env create -f environment.yml
-conda activate unCover
+
+print_bold "Compiling TEM"
+make -C tem/topic-evolution-model/
 
 print_bold "Installing CoreNLP"
-./corenlp --no-run
+conda run -n unCover ./corenlp --no-run
 
 print_bold "Downloading Models"
-./prepare_models
-cp ./.env.example ./.env
+conda run -n unCover ./prepare_models
 
 print_bold "Done!"

main.py

@@ -124,9 +124,10 @@ def get_prediction(style_prediction, te_prediction):
     col1.title("Welcome at unCover")
     col2.image(Image.open("unCover.png"), width=100)
     st.write(
-        " \nHere you can analyze a news article on topics and writing style to get further insights on whether this text "
-        "might have been written by an AI. This system was developed at Hasso-Plattner-Institute. To start, please choose "
-        "the type of input and enter the url/text in the field below.")
+        " \nHere you can analyze a news article on topics and writing style to get further insights on whether this "
+        "text might have been written by an AI. This system was developed at Hasso-Plattner-Institute. For more "
+        "information and the associated paper visit https://github.com/hpicgs/unCover.")
+    st.write("To start, please choose the type of input and enter the url/text in the field below.")
     col3, col4 = st.columns(2)
     input_type = col3.selectbox("type of input", ('URL', 'Text'), label_visibility="collapsed")
     text = ""