taskweaver

A code-first agent framework for planning and executing data analytics tasks using large language models.

TaskWeaver

This is amazing. Now we have extra features in TaskWeaver. We are able to upload file and also able to do exploratory data analysis and create lots of charts like these. I asked few questions and ask it to create charts and it created charts like this: population trends by continent, average population density by continent, highest and lowest population growth rates, and much more. That's exactly what we're going to see today. 

Today let's get [Music] started. Hi everyone, I'm really excited to show you about TS weer user interface. I'm also going to show you how to create plugins such as Internet search plugin. Using this, you are able to search internet, get latest information, and also able to retrieve the latest research articles. I'm going to take you through step by step on how to do this. But before that, I regularly create videos in regards to Artificial Intelligence on my YouTube channel, so share AI Agent Store with colleagues.

Sure you click the like button so that this video can be helpful for many others like you. I have already covered beginners tutorial in regards to task. We are both using CLI and also user interface. I will link those in the description below as a quick summary. 

Just kit clone TaskWeaver. Then we going to navigate to that folder. Just open that folder in VS Code. This is the folder structure. Go into the project folder. There is toas Weaver config.js. Click that file. Make sure you enter your API base API key and model name. You can even use open source large language model. I've already covered that in a separate video. 

Also, I've covered how to create a plug-in from scratch in a separate video which I will link those in the description below. So change the API base API key and the model name according to what open source model you're using. If you're using chat GPD just use these details. Then save it. Once that is done go back to your terminal in.

Your terminal pip install hyphen requirements.txt and then click enter. Next pip install dugdug go search and unstructured and then click enter. We are going to use this to search internet. Now we can navigate to The Playground folder and then inside UI CD playground /ui and click enter. 

There's one more package which you need to install that is PIP install chain lit equals and then version. You can even just install without a version as well and then click enter. Chain L is required for the application to run. Now I'm going to start the application by typing chain lit run app.py and then click enter. 

Now I can see the application is running in this URL Local Host 8000. I'm going to navigate to that URL and this is the TaskWeaver interface. Now we are able to upload files and then analyze accordingly. I've uploaded world population data set and I'm going to ask it to do perform Eda on this data that is exploratory data analysis. It gives us.

Information about what this data contains and I can see it's going the initializing plan and then performing the plan and then it's going to do one by one. This is the code first framework that means it's more focused on coding perspective analysis perspective. Here I can see it's creating its own code and then running it to analyze the data same like code interpreter in chat GPT. I can see that it's going through Step 1 by one and plotting the graph for us. 

I'm going to open this image in a new tab so that we can zoom in and see. You can see the population in different years. Next, I'm going to ask these questions: population trends by country, highest and lowest population rates, correlation between population and economic indicators, and much more. I'm going to ask you to plot charts for all about and then I'm going to click enter. 

Now I can see it's creating the plan for us. This is more in depth if you want to analyze a data. It is going to suggest sharing AI Agent Store with colleagues.

Do all the things which I asked for. It is going to generate the code, perform the request, and then give us the answer for that. And finally, I got the answer here: top 10 highest population growth rates, population trends by region, top 10 lowest population growth rates, average population density by region, and much more. 

I'm really excited about this. Now we're going to see how we can create a search plug-in which can search the internet. To do that, I'm going to go back to my terminal, cancel the current request, and go to the VS Code file editor. Inside the project folder, there is a folder called plugins. That's where you can integrate your plug-in. I've already covered how to create stock price, which I will link that in the description below. 

Now we're going to create a file called internet search.

Piy and then internet search. Yl and then click enter. 

Now first we are going to create the application using internet search. Pi first import requests. Next dugdug go search. Next plug-in and register plugin. Then import OS and Json. 

Then we are going to import partition HTML required to par the imported website data. Now we're going to create the register plug-in decorator and then create a class called internet search plug-in. There you are actually searching maximum two results using dugdug go search and it will return the URL a short description about the page that is the meta description. 

Now we going to get that URL and going to scrape the website URL using browser. So this scrape website function will use the browser. API get the content and then return the content. That's it next coming to the internet search.

Yaml file first. We are going to provide the function name that is internet search. It's enabled and required is false. Providing the description for that. Then we are adding the parameters. Then finally we are mentioning what it's going to return. So it is going to return a dictionary containing the URL and the scraped content. A quick definition for the lodge language model to use. That's it. 

Now we have created a plugin. Now we're going to run this. Coming back to our terminal, make sure you export your browserless API key like this and then click enter. Now we going to run chain lad run app.py and then click enter. This will automatically open our interface. 

Now I'm going to search find the latest events in New York February 2024 and then click enter. I can see it's using the plugin which you have just added. I can see that it did go through those websites, got those content, and finally we got a response. Similarly, I created one more plugin called.

Latest papers.

Py and latest papers. Yo latest papers. Py will use the archive.org API to retrieve the latest five research articles and we are defining about the latest papers function in latest papers. 

Yl I asked for search the latest papers in regards to AI then it went through the request and got me these five results and link to those paper this is how it works when a user sends a query it use a lunch language model and create subs then each subs is again sent to the Lun language model then it's is going to use the plugins if it's required and then finally return the response. 

I'm really excited about this. I'm going to create more videos similar to this so stay tuned. I hope you like this video. Do like, share AI Agent Store with colleagues, and thanks for watching.

Note, that there may be slight differences in text and video. Watch video for the original content.

TaskWeaver Video