added paper, modified read me, cleaned the analysis file

README.md

@@ -1,33 +1,49 @@
 # JinnyDB
 
-JinnyDB is a Natural Language to Databases (NLIDB) application powered by LangChain and GPT 3.5 Turbo that allows users to prompt their database using natural language to retrieve data from their databases
+JinnyDB is a Natural Language to Databases (NLIDB) application powered by LangChain and GPT 3.5 Turbo model that allows users to prompt their database using natural language to retrieve data from their databases
+
+This experiment is based on my research paper which you can read [here](https://github.com/ChelcieDeAlmeida/JinnyDB/tree/main/research_paper). 
 
 ## Table of Contents
 
 - [Installation](#installation)
 - [Usage](#usage)
-- [License](#license)
 - [TO DO](#todo)
 
 ## Installation
 
+Once you download this repo, run the following command from the project directory.
 
+```pip install -r requirements.txt```
 
-## Usage
+I used .env files for my environment set up but you can do as you please.
 
-This program is available and open to all users
+You will need an account with OpenAI if you plan on using the GPT-3.5-Turbo-16K model but any suitable model should work fine. 
 
-## License
+## Usage
 
-Currently does not have any license
+This experiment used two databases through postgres:
+- AdventureWorks [Download/Installation Here](https://github.com/lorint/AdventureWorks-for-Postgres)
+- Bookings  [Download/Installation Here] (https://postgrespro.com/docs/postgrespro/10/demodb-bookings)
 
-## Contributing
+Note that I modified these databases to have it work for my purpose, you may choose to do the same or work on your own databases
+but should you decide that you'd like to do work on this exact project then you may find the queries I used in the db folder of this repo.
 
-No Active Contribution
+Some of the changes I made to the data:
+- AdventureWorks
+    - Migrated all the tables into a single schema called 'Prod' 
+    - Dropped the views that were built with the db
+- Bookings
+    - I converted the coordinates column in the airports_data table from point to text (my use case at the moment doesn't call for it)
+    - separated the coordinates column to longitude and latitude
+    - This isn't advisable if you intend to use them, Point is a better data type for coordinates (PostGIS is even better if you're interested in spatial data)
 
 
+Below are a list of tasks, nonexhaustive that I intend on completing but feel free to participate if you'd like
 
 ## TO DO:
+- [ ] Build Backend
+- [ ] Build frontend on React
 - [ ] Add template logic for dialect/user input/database object definitions
 - [ ] Integrate intermediate steps
 - [ ] Add logic to re-run when """OutputParserException: Could not parse LLM output: `I now know the final answer.`"""

lang_gpt_analysis.ipynb

@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "code",
-   "execution_count": 91,
+   "execution_count": 1,
    "metadata": {},
    "outputs": [],
    "source": [
@@ -19,7 +19,7 @@
   },
   {
    "cell_type": "code",
-   "execution_count": 92,
+   "execution_count": 2,
    "metadata": {},
    "outputs": [],
    "source": [
@@ -588,6 +588,13 @@
     "agent_executor.run(\"How many distinct flights departed from Moscow?\")"
    ]
   },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "metadata": {},
+   "outputs": [],
+   "source": []
+  },
   {
    "cell_type": "code",
    "execution_count": 51,
@@ -1573,7 +1580,7 @@
     }
    ],
    "source": [
-    "#TC12 : FAIL (checked the airports view which was a good idea and less complicated than checking airport_data table\n",
+    "#TC12 : FAIL (checked the airports view which was a good idea and less complicated than checking airport_data table which is what I did\n",
     "# but couldn't perform a nested query with aggregate function)\n",
     "agent_executor.run(\"Which cities have more than 1 airport? Return the airport code, airport name, and city in your response.\")"
    ]

requirements.txt

@@ -2,4 +2,6 @@ langchain==0.0.242
 langchain-experimental==0.0.8
 langsmith==0.0.14
 psycopg2==2.9.6
-python-dotenv==1.0.0
+python-dotenv==1.0.0
+pandas==1.5.3
+SQLAlchemy==2.0.19