On this article, we are going to create a Chatbot to your Google Paperwork with OpenAI and Langchain. Now why do we’ve to do that within the first place? It could get tedious to repeat and paste your Google Docs contents to OpenAI. OpenAI has a personality token restrict the place you may solely add particular quantities of knowledge. So if you wish to do that at scale otherwise you need to do it programmatically, you’re going to wish a library that will help you out; with that, Langchain comes into the image. You’ll be able to create a enterprise influence by connecting Langchain with Google Drive and open AI so to summarize your paperwork and ask associated questions. These paperwork may very well be your product paperwork, your analysis paperwork, or your inner information base that your organization is utilizing.
- You’ll be able to learn to fetch your Google paperwork content material utilizing Langchain.
- Discover ways to combine your Google docs content material with OpenAI LLM.
- You’ll be able to study to summarize and ask questions on your doc’s content material.
- You’ll be able to learn to create a Chatbot that solutions questions based mostly in your paperwork.
This text was printed as part of the Data Science Blogathon.
Load Your Paperwork
Earlier than we get began, we have to arrange our paperwork in google drive. The important half here’s a doc loader that langchain supplies referred to as GoogleDriveLoader. Utilizing this, you may initialize this class after which move it an inventory of doc IDs.
from langchain.document_loaders import GoogleDriveLoader import os loader = GoogleDriveLoader(document_ids=["YOUR DOCUMENT ID's'"], credentials_path="PATH TO credentials.json FILE") docs = loader.load()
You will discover your doc id out of your doc hyperlink. You will discover the id between the ahead slashes after /d/ within the hyperlink.
For instance, in case your doc hyperlink is https://docs.google.com/doc/d/1zqC3_bYM8Jw4NgF then your doc id is “1zqC3_bYM8Jw4NgF”.
You’ll be able to move the checklist of those doc IDs to document_ids parameter, and the cool half about that is you too can move a Google Drive folder ID that comprises your paperwork. In case your folder hyperlink is https://drive.google.com/drive/u/0/folders/OuKkeghlPiGgWZdM then the folder ID is “OuKkeghlPiGgWZdM1TzuzM”.
Authorize Google Drive Credentials
Allow the GoogleDrive API through the use of this hyperlink https://console.cloud.google.com/flows/enableapi?apiid=drive.googleapis.com. Please guarantee you’re logged into the identical Gmail account the place your paperwork are saved within the drive.
Step 2: Go to the Google Cloud console by clicking this link . Choose “OAuth consumer ID”. Give utility kind as Desktop app.
Step 3: After creating the OAuth consumer, obtain the secrets and techniques file by clicking “DOWNLOAD JSON”. You’ll be able to comply with Google’s steps when you have any doubts whereas making a credentials file.
Step 4: Improve your Google API Python consumer by operating under pip command
pip set up --upgrade google-api-python-client google-auth-httplib2 google-auth-oauthlib
Then we have to move our json file path into GoogleDriveLoader.
Summarizing Your Paperwork
Ensure you have your OpenAI API Keys obtainable with you. If not, comply with the under steps:
1. Go to ‘https://openai.com/ and create your account.
2. Login into your account and choose ‘API’ in your dashboard.
3. Now click on in your profile icon, then choose ‘View API Keys’.
4. Choose ‘Create new secret key’, copy it, and put it aside.
Subsequent, we have to load our OpenAI LLM. Let’s summarize the loaded docs utilizing OpenAI. Within the under code, we used a summarization algorithm referred to as summarize_chain supplied by langchain to create a summarization course of which we saved in a variable named chain that takes enter paperwork and produces concise summaries utilizing the map_reduce strategy. Exchange your API key within the under code.
from langchain.llms import OpenAI from langchain.chains.summarize import load_summarize_chain llm = OpenAI(temperature=0, openai_api_key=os.environ['OPENAI_API_KEY']) chain = load_summarize_chain(llm, chain_type="map_reduce", verbose= False) chain.run(docs)
You’ll get a abstract of your paperwork for those who run this code. If you wish to see what LangChain was doing beneath the covers, change verbose to True, after which you may see the logic that Langchain is utilizing and the way it’s considering. You’ll be able to observe that LangChain will routinely insert the question to summarize your doc, and all the textual content(question+ doc content material) will likely be handed to OpenAI. Now OpenAI will generate the abstract.
Beneath is a use case the place I despatched a doc in Google Drive associated to a product SecondaryEquityHub and summarized the doc utilizing the map_reduce chain kind and load_summarize_chain() perform. I’ve set verbose=True to see how Langchain is working internally.
from langchain.document_loaders import GoogleDriveLoader import os loader = GoogleDriveLoader(document_ids=["ceHbuZXVTJKe1BT5apJMTUvG9_59-yyknQsz9ZNIEwQ8"], credentials_path="../../desktop_credetnaisl.json") docs = loader.load() from langchain.llms import OpenAI from langchain.chains.summarize import load_summarize_chain llm = OpenAI(temperature=0, openai_api_key=os.environ['OPENAI_API_KEY']) chain = load_summarize_chain(llm, chain_type="map_reduce", verbose=True) chain.run(docs)
We are able to observe that Langchain inserted the immediate to generate a abstract for a given doc.
We are able to see the concise abstract and the product options current within the doc generated by Langchain utilizing OpenAI LLM.
Extra Use Instances
1. Analysis: We are able to use this performance whereas doing analysis, As an alternative of intensively studying all the analysis paper phrase by phrase, we are able to use the summarizing performance to get a look on the paper rapidly.
2. Training: Academic establishments can get curated textbook content material summaries from intensive knowledge, tutorial books, and papers.
3. Enterprise Intelligence: Information analysts should undergo a big set of paperwork to extract insights from paperwork. Utilizing this performance, they will scale back the massive quantity of effort.
4. Authorized Case Evaluation: Regulation training professionals can use this performance to rapidly get important arguments extra effectively from their huge quantity of earlier comparable case paperwork.
Let’s say we needed to ask questions on content material in a given doc, we have to load in a unique chain named load_qa_chain . Subsequent, we initialise this chain with a chain_type parameter. In our case, we used chain_type as “stuff” This can be a simple chain kind; it takes all of the content material, concatenates, and passes to LLM.
- map_reduce: In the beginning, the mannequin will individually seems into every doc and shops its insights, and on the finish, it combines all these insights and once more seems into these mixed insights to get the ultimate response.
- refine: It iteratively seems into every doc given within the document_id checklist, then it refines the solutions with the current info it discovered within the doc because it goes.
- Map re-rank: The mannequin will individually look into every doc and assigns a rating to the insights. Lastly, it should return the one with the very best rating.
Subsequent, we run our chain by passing the enter paperwork and question.
from langchain.chains.question_answering import load_qa_chain question = "Who's founding father of analytics vidhya?" chain = load_qa_chain(llm, chain_type="stuff") chain.run(input_documents=docs, query=question)
If you run this code, langchain routinely inserts the immediate together with your doc content material earlier than sending this to OpenAI LLM. Beneath the hood, langchain helps us with immediate engineering by offering optimized prompts to extract the required content material from paperwork. If you wish to see what prompts they’re utilizing internally, simply set verbose=True, then you may see the immediate within the output.
from langchain.chains.question_answering import load_qa_chain question = "Who's founding father of analytics vidhya?" chain = load_qa_chain(llm, chain_type="stuff", verbose=True) chain.run(input_documents=docs, query=question)
Construct Your Chatbot
Now we have to discover a technique to make this mannequin a question-answering Chatbot. Primarily we have to comply with under three issues to create a Chatbot.
1. Chatbot ought to bear in mind the chat historical past to know the context relating to the continuing dialog.
2. Chat historical past must be up to date after every immediate the person asks to bot.
2. Chatbot ought to work till the person needs to exit the dialog.
from langchain.chains.question_answering import load_qa_chain # Perform to load the Langchain question-answering chain def load_langchain_qa(): llm = OpenAI(temperature=0, openai_api_key=os.environ['OPENAI_API_KEY']) chain = load_qa_chain(llm, chain_type="stuff", verbose=True) return chain # Perform to deal with person enter and generate responses def chatbot(): print("Chatbot: Hello! I am your pleasant chatbot. Ask me something or kind 'exit' to finish the dialog.") from langchain.document_loaders import GoogleDriveLoader loader = GoogleDriveLoader(document_ids=["YOUR DOCUMENT ID's'"], credentials_path="PATH TO credentials.json FILE") docs = loader # Initialize the Langchain question-answering chain chain = load_langchain_qa() # Record to retailer chat historical past chat_history =  whereas True: user_input = enter("You: ") if user_input.decrease() == "exit": print("Chatbot: Goodbye! Have an ideal day.") break # Append the person's query to speak historical past chat_history.append(user_input) # Course of the person's query utilizing the question-answering chain response = chain.run(input_documents=chat_history, query=user_input) # Extract the reply from the response reply = response['answers']['answer'] if response['answers'] else "I could not discover a solution to your query." # Append the chatbot's response to speak historical past chat_history.append("Chatbot: " + reply) # Print the chatbot's response print("Chatbot:", reply) if __name__ == "__main__": chatbot()
We initialized our google drive paperwork and OpenAI LLM. Subsequent, we created an inventory to retailer the chat historical past, and we up to date the checklist after each immediate. Then we created an infinite whereas loop that stops when the person provides “exit” as a immediate.
On this article, we’ve seen how you can create a Chatbot to offer insights about your Google paperwork contents. Integrating Langchain, OpenAI, and Google Drive is likely one of the most helpful use circumstances in any subject, whether or not medical, analysis, industrial, or engineering. As an alternative of studying total knowledge and analyzing the info to get insights which prices a variety of human time and effort. We are able to implement this expertise to automate describing, summarizing, analyzing, and extracting insights from our knowledge recordsdata.
- Google paperwork may be fetched into Python utilizing Python’s GoogleDriveLoader class and Google Drive API credentials.
- By integrating OpenAI LLM with Langchain, we are able to summarize our paperwork and ask questions associated to the paperwork.
- We are able to get insights from a number of paperwork by selecting acceptable chain varieties like map_reduce, stuff, refine, and map rerank.
Continuously Requested Questions
A. To construct an clever chatbot, that you must have acceptable knowledge, then that you must give entry to ChatGPT for this knowledge. Lastly, that you must present dialog reminiscence to the bot to retailer the chat historical past to know the context.
A. One of many options is you should utilize Langchain’s GoogleDriveLoader to fetch a Google Doc then, you may initialize the OpenAI LLM utilizing your API keys, then you may share the file to this LLM.
A. First, that you must allow Google Drive API, then get your credentials for Google Drive API, then you may move the doc id of your file to the OpenAI ChatGPT mannequin utilizing Langchain GoogleDriveLoader.
A. ChatGPT can’t entry our paperwork immediately. Nevertheless, we are able to both copy and paste the content material into ChatGPT or immediately fetch the contents of paperwork utilizing Langchain then, we are able to move the contents to ChatGPT by initializing it utilizing secret keys.
The media proven on this article shouldn’t be owned by Analytics Vidhya and is used on the Writer’s discretion.