diff --git a/.gitignore b/.gitignore
new file mode 100644
index 0000000000000000000000000000000000000000..6609e2b52a8167d7633d53087d93804b72ffd227
--- /dev/null
+++ b/.gitignore
@@ -0,0 +1,7 @@
+*env/
+.env
+
+*.egg-*
+*.pyc
+*.txt
+config.txt
diff --git a/Dockerfile b/Dockerfile
index 5910a680d795db4a6b5d7b7c7e7fc7de7ceccca3..6349f47c227493952580cee4888e07e93c5376f4 100644
--- a/Dockerfile
+++ b/Dockerfile
@@ -1,23 +1,25 @@
-FROM jupyter/scipy-notebook:hub-1.5.0
+FROM jupyter/datascience-notebook:hub-3.1.1
-# Install from APT repository
USER root
-RUN apt-get update -y
-RUN apt-get install -y git libpq-dev gcc
-# Set working directory
-WORKDIR /usr/src/app
+COPY requirements.txt environment.yml /tmp/
+RUN conda env update -q -f /tmp/environment.yml && \
+ /opt/conda/bin/pip install -r /tmp/requirements.txt && \
+ conda clean -y --all && \
+ conda env export -n "root" && \
+ jupyter lab build
-# Install basics
-USER jovyan
-RUN pip3 install --upgrade pip
-COPY .env .env
+COPY dash_proxy /tmp/dash_proxy/
+RUN pip install /tmp/dash_proxy/
-# Install python packages
-COPY requirements.txt requirements.txt
-RUN pip install --no-cache-dir -r requirements.txt
-ENV IPYTHONDIR /usr/src/app/ipython/
-WORKDIR /usr/src/app/ipython/profile_default/startup/
-COPY python_startup/ ./
-WORKDIR /home/
+COPY llm_utils /llm_utils/
+RUN pip install /llm_utils/
+ENV CONFIG_PATH=/home/jovyan/config.txt
+COPY app /dash/app/
+RUN chown -R jovyan /dash/app/
+
+# install some NLTK and spaCy data
+RUN python -m nltk.downloader stopwords
+RUN python -m nltk.downloader wordnet
+RUN python -m spacy download en_core_web_sm
diff --git a/README.md b/README.md
index 38011d6a37fc8d8cec424c34a4e464afbffda1cd..ef4faa61b32056f2b8678950612ad916b456bf0e 100644
--- a/README.md
+++ b/README.md
@@ -1,2 +1,3 @@
-# bücker
+# A Jupyterlab for LLM
+In order to run Dash or use the client, AZURE_OPENAI_API_KEY, AZURE_OPENAI_ENDPOINT, OPENAI_API_VERSION need to be stored in a config.txt file in the home directory.
diff --git a/app/app.py b/app/app.py
new file mode 100644
index 0000000000000000000000000000000000000000..b43ae3ac69e2d41ac5bf45446abf5d36aa5d7996
--- /dev/null
+++ b/app/app.py
@@ -0,0 +1,65 @@
+import sys
+sys.path.append("/home/jovyan/")
+
+import argparse
+import logging
+
+from urllib.parse import urlparse, urljoin
+
+from dash import Dash
+
+from jupyter_server.serverapp import list_running_servers
+
+from layout import layout
+from callbacks import register_callbacks
+
+logging.basicConfig(level=logging.INFO)
+
+# weird trick to find base_url for the jupyterlab
+def find_jupyterlab_base_url():
+ servers = list_running_servers()
+ for server in servers:
+ if server["port"] == 8888:
+ return server['url']
+ return None
+
+
+# get the correct port from proxy
+parser = argparse.ArgumentParser()
+parser.add_argument("--port", type=int)
+args = parser.parse_args()
+port: int = args.port
+
+if not port:
+ raise ValueError(f"Port of proxy server for Dash not found in {args}.")
+else:
+ logging.debug(f"Dash app running on port {port}.")
+
+
+base_url = find_jupyterlab_base_url()
+if base_url is None:
+ raise ValueError("Base URL of Jupyterlab could not be detected.")
+logging.debug(f"Base URL: {base_url}")
+
+proxy_base_path = urlparse(urljoin(base_url + "/", f"proxy/{port}/")).path
+logging.debug(f"Proxy base path: {proxy_base_path}")
+
+# define Dash app
+app = Dash(
+ name=__name__,
+ requests_pathname_prefix=proxy_base_path
+)
+
+# define layout
+app.layout = layout
+
+# register all callback functions
+register_callbacks(app=app)
+
+# Run Dash app in the notebook
+app.run(
+ jupyter_mode="jupyterlab",
+ port=port,
+ host="0.0.0.0",
+ debug=True
+)
diff --git a/app/callbacks.py b/app/callbacks.py
new file mode 100644
index 0000000000000000000000000000000000000000..9a61c5cb2d483203df434aae8482552bac8dfc0c
--- /dev/null
+++ b/app/callbacks.py
@@ -0,0 +1,65 @@
+import os
+from datetime import datetime
+
+from dash import (
+ html,
+ Dash
+)
+from dash.dependencies import (
+ Input,
+ Output,
+ State
+)
+
+from llm_utils.client import ChatGPT, get_openai_client
+
+
+def format_chat_messages(chat_history):
+ chat_messages = []
+ for message in chat_history:
+ chat_messages.append(html.Div([
+ html.P(f'{message["sender"]}: {message["message"]}'),
+ html.P(f'Sent at: {message["timestamp"]}')
+ ]))
+ return chat_messages
+
+
+def register_callbacks(app: Dash):
+ model="gpt4"
+ client = get_openai_client(
+ model=model,
+ config_path=os.environ.get("CONFIG_PATH")
+ )
+ chat_gpt = ChatGPT(
+ client=client,
+ model="gpt4"
+ )
+
+ @app.callback(
+ [Output('chat-container', 'children'),
+ Output('chat-history', 'data')],
+ [Input('send-button', 'n_clicks')],
+ [State('user-input', 'value'),
+ State('chat-history', 'data')]
+ )
+ def update_chat(n_clicks, input_value, chat_history):
+ if chat_history is None:
+ chat_history = []
+
+ if n_clicks > 0 and input_value:
+ chat_history.append({
+ 'sender': 'User',
+ 'message': input_value,
+ 'timestamp': datetime.now().strftime("%Y-%m-%d %H:%M:%S")
+ })
+
+ response = chat_gpt.chat_with_gpt(input_value)
+
+ # Add response to chat history
+ chat_history.append({
+ 'sender': 'Language Model',
+ 'message': response,
+ 'timestamp': datetime.now().strftime("%Y-%m-%d %H:%M:%S")
+ })
+
+ return format_chat_messages(chat_history), chat_history
diff --git a/app/layout.py b/app/layout.py
new file mode 100644
index 0000000000000000000000000000000000000000..4d82a24710224074e9123450133daf5eefd3aac2
--- /dev/null
+++ b/app/layout.py
@@ -0,0 +1,40 @@
+from dash import (
+ html,
+ dcc
+)
+
+layout = html.Div([
+ dcc.Store(
+ id='chat-history',
+ data=[]
+ ),
+ html.H1(
+ "Simple Chat App",
+ style={'text-align': 'center'}
+ ),
+ html.Div(
+ id='chat-container',
+ style={'overflowY': 'scroll', 'height': '70vh', 'padding': '10px'}
+ ),
+ html.Div([
+ dcc.Input(
+ id='user-input',
+ type='text',
+ placeholder='Type your message...',
+ debounce=True
+ ),
+ html.Button(
+ 'Send',
+ id='send-button',
+ n_clicks=0
+ )
+ ], style={
+ 'display': 'flex',
+ 'alignItems': 'center',
+ 'justifyContent': 'center',
+ 'position': 'fixed',
+ 'bottom': 0,
+ 'width': '100%',
+ 'padding': '10px'
+ })
+], style={'position': 'relative'})
diff --git a/app/my_app.py b/app/my_app.py
new file mode 100644
index 0000000000000000000000000000000000000000..db083be592f45c76bd6c05d3f2783614b84cfcb0
--- /dev/null
+++ b/app/my_app.py
@@ -0,0 +1,70 @@
+import sys
+sys.path.append("/home/jovyan/")
+
+import argparse
+import logging
+
+from urllib.parse import urlparse, urljoin
+
+from dash import Dash
+
+from jupyter_server.serverapp import list_running_servers
+
+try:
+ from my_layout import layout
+ from my_callbacks import register_callbacks
+except ModuleNotFoundError:
+ # do not let Dash start
+ exit()
+
+
+logging.basicConfig(level=logging.INFO)
+
+# weird trick to find base_url for the jupyterlab
+def find_jupyterlab_base_url():
+ servers = list_running_servers()
+ for server in servers:
+ if server["port"] == 8888:
+ return server['url']
+ return None
+
+
+# get the correct port from proxy
+parser = argparse.ArgumentParser()
+parser.add_argument("--port", type=int)
+args = parser.parse_args()
+port: int = args.port
+
+if not port:
+ raise ValueError(f"Port of proxy server for Dash not found in {args}.")
+else:
+ logging.debug(f"Dash app running on port {port}.")
+
+
+base_url = find_jupyterlab_base_url()
+if base_url is None:
+ raise ValueError("Base URL of Jupyterlab could not be detected.")
+logging.debug(f"Base URL: {base_url}")
+
+proxy_base_path = urlparse(urljoin(base_url + "/", f"proxy/{port}/")).path
+logging.debug(f"Proxy base path: {proxy_base_path}")
+
+# define Dash app
+app = Dash(
+ name=__name__,
+ requests_pathname_prefix=proxy_base_path
+)
+
+# define layout
+app.layout = layout
+
+# register all callback functions
+register_callbacks(app=app)
+
+# Run Dash app in the notebook
+app.run(
+ jupyter_mode="jupyterlab",
+ port=port,
+ host="0.0.0.0",
+ debug=True
+)
diff --git a/dash_proxy/dash_proxy.py b/dash_proxy/dash_proxy.py
new file mode 100644
index 0000000000000000000000000000000000000000..309e977e30e4ae08b01c6f3c3734e673b94fb444
--- /dev/null
+++ b/dash_proxy/dash_proxy.py
@@ -0,0 +1,16 @@
+def setup_dash_proxy():
+ command = [
+ 'python',
+ '/dash/app/app.py',
+ '--port',
+ '{port}'
+ ]
+
+ return {
+ "command": command,
+ "new_browser_tab": False,
+ "launcher_entry": {
+ "enabled": True,
+ 'title': 'Dash'
+ }
+ }
diff --git a/dash_proxy/my_app_proxy.py b/dash_proxy/my_app_proxy.py
new file mode 100644
index 0000000000000000000000000000000000000000..e91a8649001749f630ee85b4e6706f99049fb634
--- /dev/null
+++ b/dash_proxy/my_app_proxy.py
@@ -0,0 +1,16 @@
+def setup_my_app_proxy():
+ command = [
+ 'python',
+ '/dash/app/my_app.py',
+ '--port',
+ '{port}'
+ ]
+
+ return {
+ "command": command,
+ "new_browser_tab": False,
+ "launcher_entry": {
+ "enabled": True,
+ 'title': 'MyApp'
+ }
+ }
diff --git a/dash_proxy/setup.py b/dash_proxy/setup.py
new file mode 100644
index 0000000000000000000000000000000000000000..8e1ebf5761c3de7f783ce204b25474e10f2b0d8d
--- /dev/null
+++ b/dash_proxy/setup.py
@@ -0,0 +1,17 @@
+import setuptools
+
+setuptools.setup(
+ author="Julian Rasch",
+ author_email="julian.rasch@fh-muenster.de",
+ description="A small module to run Dash inside a dockerized Jupyterlab.",
+ name="jupyter-dash-proxy",
+ py_modules=["dash_proxy", "my_app_proxy"],
+ entry_points={
+ "jupyter_serverproxy_servers": [
+ # name = packagename:function_name
+ "Dash = dash_proxy:setup_dash_proxy",
+ "MyApp = my_app_proxy:setup_my_app_proxy"
+ ]
+ },
+ install_requires=["jupyter-server-proxy==4.0.0"],
+)
diff --git a/environment.yml b/environment.yml
new file mode 100644
index 0000000000000000000000000000000000000000..eef32e175e1f18d058cee69d1e5baf91d7bdca45
--- /dev/null
+++ b/environment.yml
@@ -0,0 +1,7 @@
+name: "base"
+channels:
+ - defaults
+# dependencies:
+# - add packages here
+# - one per line
+prefix: "/opt/conda"
diff --git a/llm_utils/setup.py b/llm_utils/setup.py
new file mode 100644
index 0000000000000000000000000000000000000000..737729b7716520f1c36d7351fea156faee49258f
--- /dev/null
+++ b/llm_utils/setup.py
@@ -0,0 +1,14 @@
+import setuptools
+
+setuptools.setup(
+ author="Julian Rasch",
+ author_email="julian.rasch@fh-muenster.de",
+ description="Helper modules to work with LLMs.",
+ name="llm_utils",
+ package_dir={"": "src"},
+ packages=setuptools.find_packages(where="src"),
+ install_requires=[
+ "openai",
+ "python-dotenv"
+ ]
+)
diff --git a/llm_utils/src/__init__.py b/llm_utils/src/__init__.py
new file mode 100644
index 0000000000000000000000000000000000000000..e69de29bb2d1d6434b8b29ae775ad8c2e48c5391
diff --git a/llm_utils/src/llm_utils/__init__.py b/llm_utils/src/llm_utils/__init__.py
new file mode 100644
index 0000000000000000000000000000000000000000..e69de29bb2d1d6434b8b29ae775ad8c2e48c5391
diff --git a/llm_utils/src/llm_utils/client.py b/llm_utils/src/llm_utils/client.py
new file mode 100644
index 0000000000000000000000000000000000000000..1bca863281a30f6cca7798da46b09adbb125d0a7
--- /dev/null
+++ b/llm_utils/src/llm_utils/client.py
@@ -0,0 +1,81 @@
+import os
+from openai import AzureOpenAI
+from dotenv import load_dotenv
+
+from enum import Enum
+
+
+class OpenAIModels(Enum):
+ GPT_3 = "gpt3"
+ GPT_4 = "gpt4"
+ EMBED = "embed"
+
+ @classmethod
+ def get_all_values(cls):
+ return [member.value for member in cls]
+
+
+def get_openai_client(
+ model: str,
+ config_path: str
+ ) -> AzureOpenAI:
+ if not model in OpenAIModels.get_all_values():
+ raise ValueError(f"<model> needs to be one of {OpenAIModels.get_all_values()}.")
+
+ load_dotenv(
+ dotenv_path=config_path,
+ override=True
+ )
+
+ AZURE_OPENAI_API_KEY = os.environ.get("AZURE_OPENAI_API_KEY")
+ AZURE_OPENAI_ENDPOINT = os.environ.get("AZURE_OPENAI_ENDPOINT")
+ OPENAI_API_VERSION = os.environ.get("OPENAI_API_VERSION")
+
+ if any(p is None for p in (AZURE_OPENAI_API_KEY, AZURE_OPENAI_API_KEY, OPENAI_API_VERSION)):
+ raise ValueError(
+ f"""None of the following parameters can be None:
+ AZURE_OPENAI_API_KEY: {AZURE_OPENAI_API_KEY},
+ AZURE_OPENAI_API_KEY: {AZURE_OPENAI_API_KEY},
+ OPENAI_API_VERSION: {OPENAI_API_VERSION}
+ """
+ )
+
+ client = AzureOpenAI(
+ api_key=AZURE_OPENAI_API_KEY,
+ azure_endpoint=AZURE_OPENAI_ENDPOINT,
+ api_version=OPENAI_API_VERSION,
+ azure_deployment=model
+ )
+ return client
+
+
+class ChatGPT:
+ def __init__(self, client: AzureOpenAI, model: str):
+ self.model = model
+ self.client = client
+ self.messages = []
+
+ def chat_with_gpt(self, user_input: str):
+ self.messages.append({
+ "role": "user",
+ "content": user_input
+ })
+ response = self._generate_response(self.messages)
+ return response
+
+ def _generate_response(self, messages):
+ response = self.client.chat.completions.create(
+ model=self.model,
+ messages=messages,
+ temperature=0.2,
+ max_tokens=150,
+ top_p=1.0
+ )
+ response_message = response.choices[0].message
+ self.messages.append({
+ "role": response_message.role,
+ "content": response_message.content
+ })
+
+ return response_message.content
+
diff --git a/my_callbacks.py b/my_callbacks.py
new file mode 100644
index 0000000000000000000000000000000000000000..6640c9dab08cbf2a5826a0511b78b6e03a6e215c
--- /dev/null
+++ b/my_callbacks.py
@@ -0,0 +1,21 @@
+from dash.dependencies import (
+ Input,
+ Output
+)
+from dash import html
+
+
+def register_callbacks(app):
+ @app.callback(
+ Output('output-container-button', 'children'),
+ [Input('submit-btn', 'n_clicks')],
+ [Input('input-text', 'value')]
+ )
+ def update_output(n_clicks, input_value):
+ if n_clicks > 0:
+ return html.Div([
+ html.Label("You entered:"),
+ html.P(input_value)
+ ])
+ else:
+ return ''
diff --git a/my_layout.py b/my_layout.py
new file mode 100644
index 0000000000000000000000000000000000000000..e0cf7c4a9a74950db48449451e5e594fa8df9140
--- /dev/null
+++ b/my_layout.py
@@ -0,0 +1,13 @@
+from dash import html
+from dash import dcc
+
+
+layout = html.Div([
+ html.H1("Yeay, my app!"),
+ html.Div([
+ html.Label("Enter your text:"),
+ dcc.Input(id='input-text', type='text', value=''),
+ html.Button('Submit', id='submit-btn', n_clicks=0),
+ ]),
+ html.Div(id='output-container-button')
+])
diff --git a/requirements.txt b/requirements.txt
index c1d6a954ef9cf503b455d0e7529ff383ca13fdd6..20b821cca3adfc3d9e28610094f92494c3330ea0 100644
--- a/requirements.txt
+++ b/requirements.txt
@@ -1,16 +1,18 @@
-jupyterlab-git==0.34.0
-jupyterlab-gitlab==3.0.0
-psycopg2-binary
-psycopg2
-python-dotenv
-sqlalchemy
-urllib3
+jupyter-server-proxy==4.0.0
+jupyterlab-git==0.42.0
+jupyter_server>=2.0
+
+flake8
+
dash
+dash-bootstrap-components
+plotly
+
openai
-pandas
-numpy
-tenacity
+rapidfuzz
+nltk
+spacy
scikit-learn
-flask
-dash
-dash-bootstrap-components
+
+python-dotenv
+