Merge main to colbert

.dockerignore

@@ -0,0 +1,5 @@
+experiments
+colbert.ipynb
+test.ipynb
+.venv
+__pycache__

.gitignore

@@ -1,6 +1,8 @@
 __pycache__
 .env
+.venv
 src/prepline_sec_filings/__pycache__
 experiments
 colbert-env
-sec-earnings-call
+sec-earnings-call
+sec-earnings-call-openai

Dockerfile

@@ -0,0 +1,17 @@
+FROM python:3.11-slim-bullseye
+
+COPY ./requirements.txt .
+
+# RUN apt-get update && apt-get install -y 
+RUN pip3 install --upgrade pip
+
+RUN pip3 --no-cache-dir install -r requirements.txt
+
+WORKDIR /src
+COPY . /src
+EXPOSE 8501
+
+HEALTHCHECK CMD curl --fail http://localhost:8501/_stcore/health
+
+ENTRYPOINT [ "streamlit", "run" ]
+CMD [ "Intro.py", "--server.fileWatcherType", "none", "--browser.gatherUsageStats", "false", "--server.address", "0.0.0.0"]

Dockerfile_Streamlit

@@ -0,0 +1,17 @@
+FROM python:3.11-slim-bullseye
+
+COPY ./requirements.txt .
+
+# RUN apt-get update && apt-get install -y 
+RUN pip3 install --upgrade pip
+
+RUN pip3 --no-cache-dir install -r requirements.txt
+
+WORKDIR /src
+COPY . /src
+EXPOSE 8501
+
+HEALTHCHECK CMD curl --fail http://localhost:8501/_stcore/health
+
+ENTRYPOINT [ "streamlit", "run" ]
+CMD [ "Intro.py", "--server.fileWatcherType", "none", "--browser.gatherUsageStats", "false", "--server.address", "0.0.0.0"]

Intro.py

@@ -1,28 +1,38 @@
 import streamlit as st
-from src.vectorDatabase import create_database
+from src.vectorDatabaseDocker import create_database
+from datetime import datetime
 
+curr_year = datetime.now().year
 ticker = st.text_input(label="Ticker")
 year = st.text_input(label="Year")
 
 if year != "":
     int_year = int(float(year))
-
-if ticker != "" and year != "":
+submit_button = st.button(label="Submit")
+if ticker != "" and year != "" and submit_button:
+    if curr_year == int_year:
+        curr_year_bool = True
+    else:
+        curr_year_bool = False
     (
         qdrant_client,
         encoder,
         speakers_list_1,
         speakers_list_2,
         speakers_list_3,
+        speakers_list_4,
         sec_form_names,
         earnings_call_quarter_vals,
     ) = create_database(ticker=ticker, year=int_year)
     st.write("Created the database")
 
+    st.session_state["ticker"] = ticker
+    st.session_state["year"] = str(year)
     st.session_state["qdrant_client"] = qdrant_client
     st.session_state["encoder"] = encoder
     st.session_state["speaker_list_1"] = speakers_list_1
     st.session_state["speaker_list_2"] = speakers_list_2
     st.session_state["speaker_list_3"] = speakers_list_3
+    st.session_state["speaker_list_4"] = speakers_list_4
     st.session_state["sec_form_names"] = sec_form_names
     st.session_state["earnings_call_quarter_vals"] = earnings_call_quarter_vals

README.md

@@ -0,0 +1,18 @@
+## INSTRUCTIONS TO RUN
+
+1. Start the docker container for qdrant by running (See more instructions [here](https://qdrant.tech/documentation/guides/installation/#docker-and-docker-compose))
+```
+docker run -p 6333:6333 \
+    -v $(pwd)/path/to/data:/qdrant/storage \
+    qdrant/qdrant
+```
+
+2. Run the requirements file for installing the packages
+
+3. Run the streamlit file for the demo
+
+```
+streamlit run Intro.py
+```
+
+

colbert.ipynb

@@ -1382,6 +1382,33 @@
     "sd = build_index(\"AAPL\",2023)"
    ]
   },
+  {
+   "cell_type": "code",
+   "execution_count": 3,
+   "metadata": {},
+   "outputs": [
+    {
+     "name": "stdout",
+     "output_type": "stream",
+     "text": [
+      "y\n"
+     ]
+    }
+   ],
+   "source": [
+    "import os\n",
+    "from src.config import *\n",
+    "from qdrant_client import QdrantClient\n",
+    "\n",
+    "QdrantClient(\n",
+    "    \n",
+    ")\n",
+    "ticker = \"AAPL\"\n",
+    "year = 2023\n",
+    "\n",
+    "if os.path.exists(f\"{DATABASE_FOLDER}/{ticker}-{year}-db\"):\n"
+   ]
+  },
   {
    "cell_type": "code",
    "execution_count": null,
@@ -1406,7 +1433,7 @@
    "name": "python",
    "nbconvert_exporter": "python",
    "pygments_lexer": "ipython3",
-   "version": "3.8.18"
+   "version": "3.8.0"
   }
  },
  "nbformat": 4,

modal.py

@@ -0,0 +1,117 @@
+import pathlib
+
+import modal
+
+with open("requirements.txt") as f:
+    r = f.read()
+
+packages_list = r.split("\n")
+
+image = (
+    modal.Image.debian_slim()
+    .apt_install("git")
+    .pip_install(
+        "aiohttp==3.8.5",
+        "fastapi==0.109.2",
+        "langchain==0.1.5",
+        "nltk==3.8.1",
+        "numpy==1.25.0",
+        "openai==0.27.8",
+        "pandas==2.0.3",
+        "# pydantic==1.10.12",
+        "python-dotenv==1.0.1",
+        "qdrant_client==1.7.3",
+        "ratelimit==2.2.1",
+        "Requests==2.31.0",
+        "scikit_learn==1.3.0",
+        "sentence_transformers==2.3.1",
+        "starlette==0.36.3",
+        "streamlit==1.24.1",
+        "tenacity==8.2.2",
+        "torch==2.0.1",
+        "tqdm==4.65.0",
+        "typing_extensions==4.9.0",
+        "unstructured==0.8.1",
+        "uvicorn",
+        "streamlit-feedback==0.1.3",
+    )
+    # Use fork until https://github.com/valohai/asgiproxy/pull/11 is merged.
+    .pip_install("git+https://github.com/modal-labs/asgiproxy.git")
+)
+
+stub = modal.Stub(name="Financial-Docs-LLM", image=image)
+
+streamlit_script_local_path = pathlib.Path(__file__).parent / "Intro.py"
+streamlit_script_remote_path = pathlib.Path("/root/Intro.py")
+
+if not streamlit_script_local_path.exists():
+    raise RuntimeError(
+        "Intro.py not found! Place the script with your streamlit app in the same directory."
+    )
+
+streamlit_script_mount = modal.Mount.from_local_file(
+    streamlit_script_local_path,
+    streamlit_script_remote_path,
+)
+
+HOST = "127.0.0.1"
+PORT = "8000"
+
+
+def spawn_server():
+    import socket
+    import subprocess
+
+    process = subprocess.Popen(
+        [
+            "streamlit",
+            "run",
+            str(streamlit_script_remote_path),
+            "--browser.serverAddress",
+            HOST,
+            "--server.port",
+            PORT,
+            "--browser.serverPort",
+            PORT,
+            "--server.enableCORS",
+            "false",
+        ]
+    )
+
+    # Poll until webserver accepts connections before running inputs.
+    while True:
+        try:
+            socket.create_connection((HOST, int(PORT)), timeout=1).close()
+            print("Webserver ready!")
+            return process
+        except (socket.timeout, ConnectionRefusedError):
+            # Check if launcher webserving process has exited.
+            # If so, a connection can never be made.
+            retcode = process.poll()
+            if retcode is not None:
+                raise RuntimeError(f"launcher exited unexpectedly with code {retcode}")
+
+
+@stub.function(
+    # Allows 100 concurrent requests per container.
+    allow_concurrent_inputs=100,
+    mounts=[streamlit_script_mount],
+)
+@modal.asgi_app()
+def run():
+    from asgiproxy.config import BaseURLProxyConfigMixin, ProxyConfig
+    from asgiproxy.context import ProxyContext
+    from asgiproxy.simple_proxy import make_simple_proxy_app
+
+    spawn_server()
+
+    config = type(
+        "Config",
+        (BaseURLProxyConfigMixin, ProxyConfig),
+        {
+            "upstream_base_url": f"http://{HOST}:{PORT}",
+            "rewrite_host_header": f"{HOST}:{PORT}",
+        },
+    )()
+    proxy_context = ProxyContext(config)
+    return make_simple_proxy_app(proxy_context)

pages/EarningsChat.py

@@ -1,6 +1,6 @@
 from src.chat_earnings_call import get_openai_answer_earnings_call
 from src.queryDatabase import query_database_earnings_call
-
+from streamlit_feedback import streamlit_feedback
 import streamlit as st
 from dotenv import load_dotenv
 import openai
@@ -17,10 +17,15 @@
 speaker_list_1 = st.session_state["speaker_list_1"]
 speakers_list_2 = st.session_state["speaker_list_2"]
 speakers_list_3 = st.session_state["speaker_list_3"]
+speakers_list_4 = st.session_state["speaker_list_4"]
 earnings_call_quarter_vals = st.session_state["earnings_call_quarter_vals"]
 quarter = st.selectbox("Quarter Name", tuple(earnings_call_quarter_vals))
 
 st.session_state["quarter"] = quarter
+ticker = st.session_state["ticker"]
+year = st.session_state["year"]
+
+st.title(f"{ticker}-{year}")
 
 
 def generate_response(input_text):
@@ -31,6 +36,8 @@ def generate_response(input_text):
         speakers_list = speakers_list_2
     elif quarter == "Q3":
         speakers_list = speakers_list_3
+    elif quarter == "Q4":
+        speakers_list = speakers_list_4
 
     relevant_text = query_database_earnings_call(
         input_text, quarter, qdrant_client, encoder, speakers_list
@@ -64,5 +71,10 @@ def generate_response(input_text):
             st.write(docs)
             expander = st.expander("See relevant sources")
             expander.write(relevant_text)
+            feedback = streamlit_feedback(
+                feedback_type="thumbs",
+                optional_text_label="Please describe the feedback in detail",
+            )
+            # print(feedback)
     message = {"role": "assistant", "content": docs}
     st.session_state.messages.append(message)

pages/SECChat.py

@@ -1,6 +1,6 @@
 from src.chat_sec import get_openai_answer_sec
 from src.queryDatabase import query_database_sec
-
+from streamlit_feedback import streamlit_feedback
 import streamlit as st
 from dotenv import load_dotenv
 import openai
@@ -16,6 +16,10 @@
 sec_form_names = st.session_state["sec_form_names"]
 form_name = st.selectbox("Form Name", tuple(sec_form_names))
 st.session_state["form_name"] = form_name
+ticker = st.session_state["ticker"]
+year = st.session_state["year"]
+
+st.title(f"{ticker}-{year}")
 
 
 def generate_response(input_text):
@@ -52,5 +56,9 @@ def generate_response(input_text):
             st.write(docs)
             expander = st.expander("See relevant sources")
             expander.write(relevant_text)
+            feedback = streamlit_feedback(
+                feedback_type="thumbs",
+                optional_text_label="Please describe the feedback in detail",
+            )
     message = {"role": "assistant", "content": docs}
     st.session_state.messages.append(message)

requirements.txt

@@ -1,23 +1,23 @@
-aiohttp==3.8.4
 aiohttp==3.8.5
 fastapi==0.109.2
 langchain==0.1.5
 nltk==3.8.1
 numpy==1.25.0
-openai==0.27.8
+openai==1.26.0
 pandas==2.0.3
-pydantic==1.10.12
-pydantic_settings==2.1.0
+# pydantic==1.10.12
 python-dotenv==1.0.1
 qdrant_client==1.7.3
 ratelimit==2.2.1
 Requests==2.31.0
 scikit_learn==1.3.0
 sentence_transformers==2.3.1
-starlette==0.37.0
+starlette==0.36.3
 streamlit==1.24.1
 tenacity==8.2.2
 torch==2.0.1
 tqdm==4.65.0
 typing_extensions==4.9.0
 unstructured==0.8.1
+uvicorn
+streamlit-feedback==0.1.3