Open in app Get unlimited access ChatGPT has become an integral tool that most people use daily

access ChatGPT has become an integral tool that most people use daily to automate various
tasks. If you have used ChatGPT for any length of time, you would have realized that
it can provide wrong answers and has limited to zero context on some niche topics.
New: Navigate Medium from
the top of the page, and This brings up the question of how we can tap into chatGPT to bridge the gap and
focus more on reading
Published as
in Better Programming allow ChatGPT to have more custom data.
you scroll.

Okay, got it A wealth of knowledge is distributed in various platforms we interact with daily, i.e.,
This is your last free member-only story this month. Upgrade for unlimited access. via confluence wiki pages at work, slack groups, company knowledge base, Reddit,
Stack Overflow, books, newsletters, and google documents shared by colleagues.
Keeping up with all these information sources is a full-time job in itself.
Timothy Mugayi Follow

Mar 17 · 11 min read · · Listen

Wouldn't it be nice if you could selectively choose your data sources and feed that
information into ChatGPT conversation with your data with ease?
Save

1. Feeding Data via Prompt Engineering

How To Build Your Own Custom ChatGPT With Before we jump into how we can extend ChatGPT, let's look at how we could extend

Custom Knowledge Base ChatGPT manually and what the issues are. The conventional approach to extending
ChatGPT is via prompt engineering.
Step-by-step guide on how to feed your ChatGPT bot with custom
data sources This is quite simple to do since ChatGPT is context aware. First, we need to interact
with ChatGPT by appending the original document content before the actual
questions.

I will ask you questions based on the following content:

- Start of Content-
Your very long text to give ChatGPT context
- End of Content-

The issue with this approach is the model has a limited context; it can only accept
approximately 4,097 tokens for GPT-3. You will soon run into a wall with this
approach as it's also quite a manual, tedious process to always have to paste in the
content.

Imagine having hundreds of PDF documents you wanted to inject into ChatGPT. You
will soon run into paywall issues. You might be thinking GPT-4 is the successor to
2.2K 41
GPT-3. It was just launched on March 14, 2023, and it can process 25,000 words —
about eight times as many as GPT-3 process images — and handle much more
Photo by Christian Wiediger on Unsplash
nuanced instructions than GPT-3.5. This still has the same fundamental problem of
data input limitation. How do we go about bypassing some of these limitations? We
can leverage a Python library called LlamaIndex.

2. Extending ChatGPT With LlamaIndex (GPT Index)

LlamaIndex, also known as the GPT Index, is a project that provides a central
interface to connect your LLMs with external data. Yeah, you read that correctly.
With LlamaIndex, we can build something that looks like the illustration below:

Custom data sources feeding into ChatGPT

LlamaIndex connects your existing data sources and types with available data
connectors, for example (APIs, PDFs, docs, SQL, etc.) It enables you to employ LLMs A few Word documents uploaded to your Google Docs. LlamaIndex supports
by offering indexes over your structured and unstructured data. These indices many different data sources. For this tutorial, we will demonstrate Google Docs.
facilitate in-context learning by removing typical boilerplate and pain points:
preserving context in an accessible manner for quick insertion. How It Works
1. Create a document data index with LlamaIndex.
Dealing with prompt restrictions — a 4,096 token limit for the GPT-3 Davinci and an
8,000 token limit for GPT-4 — when the context is too large becomes much more 2. Use natural language to search the index.
accessible and tackles the text-splitting issue by giving users a way to interact with
3. The pertinent pieces will be retrieved by LlamaIndex and passed to the GPT
the index. LlamaInde also abstracts the process of extracting relevant parts from the
prompt. LlamaIndex will transform your original document data into a query-
documents and feeding them to the prompt.
friendly vectorized index. It will utilize this index to find the most pertinent
sections based on how closely the query and data match. The information will
How To Add Custom Data Source
then be loaded into the prompt, which will be sent to GPT so that GPT has the
In this section, we'll use GPT "text-davinci-003" and LlamaIndex to create a Q&A
background necessary to respond to your question.
chatbot based on pre-existing documents.
4. After that, you may ask ChatGPT, given the feed in context.
Prerequisites
Create a new folder for your Python project, which you can call mychatbot,
Before we start, make sure you have access to the following:
preferably using a virtual environment or conda environment.
Python ≥ 3.7 installed on your machine
We'll need to install the dependency libraries first. Here’s how:
An OpenAI API key, which can be found on the OpenAI website. You can use
your Gmail account to single sign-on.
pip install openai
pip install llama-index
pip install google-auth-oauthlib

Next, we'll import the libraries in Python and set up your OpenAI API key in a new
main.py file.

# Import necessary packages

import os
import pickle

from google.auth.transport.requests import Request

 from google_auth_oauthlib.flow import InstalledAppFlow Project." Follow the prompts to give your project a name and select the
from llama_index import GPTSimpleVectorIndex, download_loader
organization you want to associate it with.

3. Once your project is created, please select it from the dropdown menu in the top
os.environ['OPENAI_API_KEY'] = 'SET-YOUR-OPEN-AI-API-KEY' navigation bar.

4. Go to the "APIs & Services" section from the left-hand menu and click on the "+
In the above snippet, we are explicitly setting the environment variable for clarity, ENABLE APIS AND SERVICES" button at the top of the page.
as the LlamaIndex package implicitly requires access to OpenAI. In a typical
production environment, you can put your keys in environment variables, vault, or 5. Search for "Google Docs API" in the search bar and select it from the results list.

whatever secrets management service your infra can access. 6. Click the "Enable" button to enable the API for your project.

Let's construct a function to help us authenticate against our Google account to 7. Click on the OAuth consent screen menu and create and give your app a name,
discover Google Docs. e.g., "mychatbot," then enter the support email, save, and add scopes.

def authorize_gdocs():
google_oauth2_scopes = [
"https://www.googleapis.com/auth/documents.readonly"
]
cred = None
if os.path.exists("token.pickle"):
with open("token.pickle", 'rb') as token:
cred = pickle.load(token)
if not cred or not cred.valid:
if cred and cred.expired and cred.refresh_token:
cred.refresh(Request())
else:
flow = InstalledAppFlow.from_client_secrets_file("credentials.json"
cred = flow.run_local_server(port=0)
with open("token.pickle", 'wb') as token:
pickle.dump(cred, token)

You must also add test users since this Google app will not be approved yet. This can
To enable the Google Docs API and fetch the credentials in the Google Console, you
be your own email.
can follow these steps:

1. Go to the Google Cloud Console website (console.cloud.google.com).

2. Create a new project if you haven't already. You can do this by clicking on the
"Select a project" dropdown menu in the top navigation bar and selecting "New
 Example folder structure with google credentials in root

Once you have set up your credentials, you can access the Google Docs API from
your Python project.

Go to your Google Docs, open up a few of them, and get the unique id that can be
seen in your browser URL bar, as illustrated below:

You will then need to set up credentials for your project to use the API. To do this, go
to the "Credentials" section from the left-hand menu and click "Create Credentials."
Select "OAuth client ID" and follow the prompts to set up your credentials.

Gdoc ID

Copy out the gdoc IDs and paste them into your code below. You can have N number
of gdocs that you can index so ChatGPT has context access to your custom
knowledge base. We will use the GoogleDocsReader plugin from the LlamaIndex
library to load your documents.

# function to authorize or download latest credentials

authorize_gdocs()

# initialize LlamaIndex google doc reader

GoogleDocsReader = download_loader('GoogleDocsReader')
Once your credentials are set up, you can download the JSON file and store it in the
root of your application, as illustrated below: # list of google docs we want to index
gdoc_ids = ['1ofZ96nWEZYCJsteRfqik_xNQTGFHtnc-7cYrf0dMPKQ']

loader = GoogleDocsReader()
 # load gdocs and index them
documents = loader.load_data(document_ids=gdoc_ids)
index = GPTSimpleVectorIndex(documents)

LlamaIndex has a variety of data connectors covering services such as Notion,

Obsidian, Reddit, Slack, etc. You can find the compressive list of available data
connectors here.

If you wish to save and load the index on the fly, you can use the following function We will interact directly with vanilla ChatGPT first to see what output it generates
calls. This will speed up the process of fetching from pre-saved indexes instead of without injecting a custom data source.
making API calls to external sources.

# Save your index to a index.json file

index.save_to_disk('index.json')
# Load the index from your saved index.json file
index = GPTSimpleVectorIndex.load_from_disk('index.json')

Querying the index and getting a response can be achieved by running the following
That was a little disappointing! Let's try again.
code below. Code can easily be extended into a rest API that connects to a UI where
you can interact with your custom data sources via the GPT interface.

INFO:google_auth_oauthlib.flow:"GET /?state=oz9XY8CE3LaLLsTxIz4sDgrHha4fEJ&code
INFO:googleapiclient.discovery_cache:file_cache is only supported with oauth2cl
# Querying the index INFO:root:> [build_index_from_documents] Total LLM token usage: 0 tokens
while True: INFO:root:> [build_index_from_documents] Total embedding token usage: 175 token
prompt = input("Type prompt...") Type prompt...who is timothy mugayi hint he is a writer on medium
response = index.query(prompt)
print(response) INFO:root:> [query] Total LLM token usage: 300 tokens
INFO:root:> [query] Total embedding token usage: 14 tokens
Timothy Mugayi is an Engineering Manager at OVO (PT Visionet Internasional), a
last_token_usage=300
Given we have a Google Doc with details about me, information that's readily Type prompt...

available if you publicly search on google. Type prompt...Given you know who timothy mugayi is write an interesting introdu

Timothy Mugayi is an experienced and accomplished professional with a wealth of

last_token_usage=330
It can now infer answers using a new custom data source, accurately producing the # define anoter LLM explicitly
llm_predictor = LLMPredictor(llm=OpenAI(temperature=0, model_name="text-davinci
following output.
# define prompt configuraiton
We can take things further. # set maximum input size
max_input_size = 4096
# set number of output tokens
num_output = 256
# set maximum chunk overlap
Type prompt...Write a cover letter for timothy mugayi for an upwork python proj
max_chunk_overlap = 20
INFO:root:> [query] Total LLM token usage: 436 tokens
prompt_helper = PromptHelper(max_input_size, num_output, max_chunk_overlap)
INFO:root:> [query] Total embedding token usage: 30 tokens

index = GPTSimpleVectorIndex(
Dear [Hiring Manager],
documents, llm_predictor=llm_predictor, prompt_helper=prompt_helper
)
I am writing to apply for the Python project to build a custom ChatGPT bot with

I am currently an Engineering Manager at OVO (PT Visionet Internasional), a sub

I am confident that I can deliver a high-quality product that meets the require
If you want to keep tabs on your OpenAI free or paid credits, you can navigate to the
Thank you for your time and consideration.
OpenAI dashboard and check how much credit is left.
Sincerely,
Timothy Mugayi Creating an index, inserting into an index, and querying an index will use tokens.
last_token_usage=436
Type prompt...
Hence, it's always important to ensure you output token usage for tracking purposes
when building your custom bots.

last_token_usage = index.llm_predictor.last_token_usage
LlamaIndex will internally accept your prompt, search the index for pertinent
chunks, and then pass both your prompt and the pertinent chunks to the ChatGPT print(f"last_token_usage={last_token_usage}")
model. The procedures above demonstrate a fundamental first use of LlamaIndex
and GPT for answering questions. Yet, there is much more you can do. You are only
limited by your creativity when configuring LlamaIndex to utilize a different large
Final Thoughts
language model (LLM), using a different type of index for various activities, or
ChatGPT combined with LlamaIndex can help to build a customized ChatGPT
updating old indices with a new index programmatically.
chatbot that can infer knowledge based on its own document sources. While
Here is an example of changing the LLM model explicitly. This time we tap into ChatGPT and other LLM are pretty powerful, extending the LLM model provides a
another Python package that comes bundled with LlamaIndex called langchain. much more refined experience and opens up the possibility of building a
conversational-style chatbot that can be used to build real business use cases like
customer support assistance or even spam classifiers. Given we can feed real-time
from langchain import OpenAI
data, we can evaluate some of the limitations of ChatGPT models being trained up to
from llama_index import LLMPredictor, GPTSimpleVectorIndex, PromptHelper a certain period.

...
For the complete source code, you can refer to this GitHub repo.
If you are looking to build custom ChatGPT bots that understand your domain, drop a
message in the comments section and let's connect.

Programming Chatgpt Software Development AI Python

Enjoy the read? Reward the writer.Beta

Your tip will go to Timothy Mugayi through a third-party platform of their choice, letting them know you appreciate
their story.

Give a tip

Sign up for Coffee Bytes

By Better Programming

A newsletter covering the best programming articles published across Medium Take a look.

Emails will be sent to [email protected]. Not you?

Get this newsletter