Query your own documents with LlamaIndex and Gemini

-

In this article I am going to explain about on creating application for indexing and querying your own documents using LlamaIndex and Gemini. I will provide step by step guide to create application in python.

 


Prerequisites for this example is as follows:

  1. Visual studio code
  2. Python
  3. Api Key of Gemini can be obtain from https://aistudio.google.com  

Open visual studio code and create the file with name "demo.py". Now in visual studio code and go to terminal menu and click on New terminal link it will open new terminal. In terminal enter below command to install the LlamaIndex library and LlamaIndex gemini library in your machine. 

 pip install llama-index llama-index-llms-gemini llama-index-embeddings-gemini

Create the folder named "doc" in root directory of the application as shown in below image and store the documents you want to query.

 

doc folder


 

 Now copy below code and paste in the "demo.py" file.

 from llama_index.embeddings.gemini import GeminiEmbedding  
 from llama_index.llms.gemini import Gemini  
 from llama_index.core import Settings,SimpleDirectoryReader,VectorStoreIndex,StorageContext,load_index_from_storage  
 from llama_index.core.node_parser import SentenceSplitter  
 import os  
 os.environ["GOOGLE_API_KEY"]="your google api key obtained from aistudio.google.com"  
 gemini_embedding_model=GeminiEmbedding(model_name="models/embedding-001")  
 llm=Gemini()  
 Settings.llm=llm  
 Settings.embed_model=gemini_embedding_model  
 Settings.node_parser=SentenceSplitter(chunk_size=512,chunk_overlap=20)  
 Settings.num_output=2080  
 Settings.context_window=3900  
 Persist_dir="./storage"  
 if not os.path.exists(Persist_dir):  
   documents=SimpleDirectoryReader("doc").load_data()  
   index=VectorStoreIndex.from_documents(documents)  
   index.storage_context.persist(persist_dir=Persist_dir)  
 else:  
   storage_context=StorageContext.from_defaults(persist_dir=Persist_dir)  
   index= load_index_from_storage(storage_context=storage_context)  
 query_engine=index.as_query_engine()  
 strInput=""  
 while strInput!="exit":  
   if strInput!="":  
     response=query_engine.query(strInput)  
     print(response.response+"\n\n")  
   strInput=input("Ask the question: ")

 In the above code, code highlighted in yellow is required for tell llamaindex about which LLM model to be used where the Settings object will be used to set the same. If we do not set it then llamaindex will use the default settings of OpenAI LLM model.

Code highlighted in green is required for embedding of the local documents reside inside the "doc" folder, after embedding the vector index will be stored inside the "storage" directory in Json files as shown in figure . if the "storage" directory exists then it wont do embedding again and load the embedding from vector index stored inside "storage" directory.  

local vector index

 

 

Code highlighted in blue is section of code where the application will get the input from user and search in vector store then call the LLM Gemini and get the results to print on screen. the output samples you can see in image below after running the application.

LLM output



 

Comments

Post a Comment

Popular posts from this blog

Implement Logging in CSV file using Nlog in .net core MVC application- part 2

-
Image
In previous article " Implement Nlog in .Net core MVC application part 1 "  I have demonstrated how we can implement basic setup of NLog in .core mvc application. In previous article we have logged the exceptions in flat file system. In this article I will show you how you can write the exception and custom logs in csv file. Writing logs in csv file can enable us better readability of logs in tabular form.       Lets start the implementation of it, I going to use the same example project used in previous demonstration. Modify the Nlog.config file as follows, added new target and rule for logging logs in csv file. <?xml version="1.0" encoding="utf-8" ?> <nlog xmlns="http://www.nlog-project.org/schemas/NLog.xsd" xmlns:xsi="http://www.w3.org/2001/XMLSchema-instance" xsi:schemaLocation="http://www.nlog-project.org/schemas/NLog.xsd NLog.xsd" autoReload="true" throwExceptions="fal...
Read more

Implement Nlog in .Net core MVC application part 1

-
Image
Logging is a very important part of the enterprise application/web application. Setting up the logging correctly not only beneficial for long term maintenance of the application but also it is useful tool to find out the RCA in case of mission critical bug arises. Therefor we will going to discuss the various logging technique implementation using Nlog library package. In this article we will going to use .net core application for demonstration but you can implement it in .net framework application also. So lets start the implementation of Nlog in .net core MVC application. Open visual studio 2019/17 and create the new project as shown below.     Select Asp.net core Web App (Model-View-Controller) and provide the project name in project name text box.   Now on next screen click on create button this will create MVC app with .net core Now go to solution explorer and right click on project and click on Manage Nuget Package. Now in Nuget package manager install below listed ...
Read more

Angular User Session Timeout example step by step

-
Image
This functionality let the user know that their session is about to expire and that they would be logged out if they didn’t take action. I have written this functionality in Angular for my latest application.  If you are new to Angular 6 then read my article about Setting up Angular 6 step by step before jump to this article to get handy in Angular 6. Dealing with timers in Angular can be significantly different when using React components and subscriptions. The basic premises remains the same, though. We’ll have a service with a timer that will provide a Subject to which consumers of the service can subscribe. A React Subject provides an easy mechanism to trigger a “next” subscription to alert consumers that the timer has expired.   import { Injectable } from '@angular/core'; import { Observable, Subject, Subscription, BehaviorSubject} from 'rxjs/Rx'; @Injectable() export class IdleTimeoutService { private _count: number = 0; ...
Read more

Restore the lost focus of Auto post back controls in asp.net update Panel control

-
Problem: We put  autopostback controls like textbox, DropDownList controls inside in update panel to reducing the flickering of the Page.but when user uses tab for entering the details  the focus of the control losts. Web site users who prefer to use keyboard need to use mouse to activate appropriate input box or press TAB multiple times. Solution: Step 1: Save following in js file give any name to it (like MyUpdatePanelFocus.js ) var lastFocusedControlId = ""; function focusHandler(e) { document.activeElement = e.originalTarget; } function appInit() { if (typeof (window.addEventListener) !== "undefined") { window.addEventListener("focus", focusHandler, true); } Sys.WebForms.PageRequestManager.getInstance().add_pageLoading(pageLoadingHandler); Sys.WebForms.PageRequestManager.getInstance().add_pageLoaded(pageLoadedHandler); } function pageLoadingHandler(sender, args) { lastFocusedControlId = typeof (document.acti...
Read more

Disable backspace key using Jquery

-
Disable backspace key in HTML form using Jquery in simple steps. While user enters data into application if user accidentally  press backspace button then it will redirect to previous page and all entered data would be lost. to eliminate this event we can disable the backspace key using Jquery. Following is the JQuery script tag you can add to the web page $(document).keydown(function (e) { var nodeName = e.target.nodeName.toLowerCase(); if (e.which === 8) { if ((nodeName === 'input' && e.target.type === 'text') || nodeName === 'textarea') { // do nothing } else { e.preventDefault(); } } }); The above code can be used seamlessly with any web technology you are using. Hope this will help you in your web application. Happy coding:)
Read more