Data Wellness Q&A Chatbot

This is a reverse chatbot that asks the users questions about data integration practices and then gives advice based on a body of knowledge. This version operates a bit like an agent which tries to gather enough information to be able to give advice. So it may ask an unspecified number of questions.

We have used a specially changed version of the [Chainlit][https://chainlit.io] library that you can install from the wheels folder.

The source code for the hacked chainlit version is from this repo:

https://github.com/onepointconsulting/chainlit-data-wellness-agent.git

Setup

We suggest to use Conda to manage the virtual environment and then install poetry.

conda activate base
conda remove -n data_integration_questionnaire_agent --all
conda create -n data_integration_questionnaire_agent python=3.11
conda activate data_integration_questionnaire_agent
pip install poetry

Installation

poetry install
poetry add --editable ./wheels/chainlit-0.7.8.10-py3-none-any.whl

Running

Prepare the SQLLite database

Before you run for the first time, please run this command from the command line to initiate the SQLLite database:

python ./data_questionnaire_agent/utils/tracker_db_init.py

Running the main application

chainlit run .\data_questionnaire_agent\ui\data_questionnaire_chainlit.py --port 8080

Configuration

This is the content of the .env file, which needs to be saved in the project root folder.

OPENAI_API_KEY=<open-api-key>
# This model does not seem to perform too well.
# OPENAI_MODEL=gpt-4-1106-preview
OPENAI_MODEL=gpt-4-0613
REQUEST_TIMEOUT=300

VERBOSE_LLM=true
LANGCHAIN_CACHE=false
CHATGPT_STREAMING=false

UI_TIMEOUT = 60

# Email related
MAIL_FROM_PERSON=<mail sender>
MAIL_USER=<some valid email id>
MAIL_PASSWORD=<mail password>
MAIL_FROM=<some valid email sender>
MAIL_SERVER=smtp.gmail.com:587

# General stuff
PROJECT_ROOT=/development/playground/langchain/data_questionnaire_agent
QUESTION_CACHE_FOLDER=/tmp/data_questionnaire_agent/cache

# PDF Related
WKHTMLTOPDF_BINARY=/Program Files/wkhtmltopdf/bin/wkhtmltopdf.exe
TEMPLATE_LOCATION=/development/playground/langchain/data_questionnaire_agent/templates
PDF_FOLDER=/tmp/data_questionnaire_agent/pdfs

# Whether to show the task list or not
TASKLIST=false

# UI
SHOW_CHAIN_OF_THOUGHT=true

# Embedding related
# The following property is where your knowledge base is located
RAW_TEXT_FOLDER=/development/playground/langchain/data_questionnaire_agent/docs/raw_text
EMBEDDINGS_PERSISTENCE_DIR=/development/playground/langchain/data_questionnaire_agent/embeddings
EMBEDDINGS_CHUNK_SIZE=2500
SEARCH_RESULTS_HOW_MANY=2

# Question generation related
QUESTIONS_PER_BATCH=1
# Minimum questions asked before giving advice
MINIMUM_QUESTIONNAIRE_SIZE=4

# Token limit for chatgpt 4. Important to extend the context as much as possible using the vector DB search
# This could be higher as the TPM increased on the 6th of November
TOKEN_LIMIT=6000

# Not in use
IMAGE_LLM_TEMPERATURE=0.9

# Show session cost
SHOW_SESSION_COST=false
OPENAI_RETRY_ATTEMPTS=3
OPENAI_WAIT_FIXED=30

# SQLLite DB related
# Optional
ONEPOINT_SQL_LITE_DB=/tmp/SQL_LITE_DB.db

# Related to the tracker DB download
TRACKER_DB_LOGS_PASSWORD=<pass>

Here is the content of the config.toml file in the .chainlit folder:

[project]
# Whether to enable telemetry (default: true). No personal data is collected.
enable_telemetry = true

# List of environment variables to be provided by each user to use the app.
user_env = []

# Duration (in seconds) during which the session is saved when the connection is lost
session_timeout = 3600

# Enable third parties caching (e.g LangChain cache)
cache = false

# Follow symlink for asset mount (see https://github.com/Chainlit/chainlit/issues/317)
# follow_symlink = false

[features]
# Show the prompt playground
prompt_playground = true

[UI]
# Name of the app and chatbot.
name = "Chatbot"

# Description of the app and chatbot. This is used for HTML tags.
# description = ""

# Large size content are by default collapsed for a cleaner ui
default_collapse_content = true

# The default value for the expand messages settings.
default_expand_messages = false

# Hide the chain of thought details from the user in the UI.
hide_cot = false

# Link to your github repo. This will add a github button in the UI's header.
github = "https://onepointltd.com"

# Specify a CSS file that can be used to customize the user interface.
custom_css = '/public/css/styles.css'

# Specify a JS file that can be used to customize the user interface.
custom_js = '/public/js/onepoint.js'

# The text
watermark_text = "Built by"

# Override default MUI light theme. (Check theme.ts)
[UI.theme.light]
    #background = "#FAFAFA"
    #paper = "#FFFFFF"

    [UI.theme.light.primary]
        #main = "#F80061"
        #dark = "#980039"
        #light = "#FFE7EB"

# Override default MUI dark theme. (Check theme.ts)
[UI.theme.dark]
    #background = "#FAFAFA"
    #paper = "#FFFFFF"

    [UI.theme.dark.primary]
        #main = "#F80061"
        #dark = "#980039"
        #light = "#FFE7EB"


[meta]
generated_by = "0.7.1"

Note on sqllite3 on Windows

If you have a hard time installing sqllite3 on Windows follow these instructions:

https://zeljic.com/blog/sqlite-lib-windows-10/

1. Download source from source

   For example: source https://www.sqlite.org/2023/sqlite-amalgamation-3430100.zip

2. Download binary from binary

   For example: binary https://www.sqlite.org/2023/sqlite-dll-win64-x64-3430100.zip

3. Extract both archives to the same directory

4. Open Developer Command Prompt for VS 2017 by typing Developer Command in Windows Search

5. Go to directory where you've extracted source code and binary files (via opened cmd)

6. Run lib /DEF:sqlite3.def /OUT:sqlite3.lib /MACHINE:x64

7. Copy all of the files including sqlite3.h into one of the include folders used by cl.exe like e.g. -IC:\Users\gilfe\miniconda3\envs\data_integration_questionnaire_agent\Include

Blog post

Note on sqllite3 on Linux

Make sure that gcc is installed.

On Ubuntu these are the commands:

sudo apt update
sudo apt install build-essential
# Check if gcc is installed
gcc --version

Listing the interactions

If you want to list the interactions with the system from the database you can run this command into a CSV file:

python ./data_questionnaire_agent/utils/tracker_db_lister.py > temp.csv