- 10
- 19 799
Akshay Pachaar
Приєднався 28 кві 2016
Here to simplify LLMs, AI Agents, RAGs and Machine Learning for you! • Co-founder DailyDoseOfDS • BITS Pilani • 3 Patents • ex-AI Engineer @LightningAI • Educating a community of 500k+ across my X, LinkedIn and Newsletter.
RAG using DeepSeek R1 | 100% local
I just created a RAG app with the latest DeepSeek-R1⚡️
It runs 100% locally, powered by DeepSeek, which delivers OpenAI-o1 level intelligence at 96% less cost.
The video below shows a demo of how it works!
I have published a Lightning AI Studio using DeepSeek.
Tech stack:
- DeepSeek AI R1 served as LLM.
- LlamaIndex for orchestration.
- Streamlit for the UI.
We'll cover more advanced techniques with R1 very soon! ✨
Launch the studio now: lightning.ai/akshay-ddods/studios/rag-using-deepseek-r1?view=public§ion=featured
#rag #agenticai #llm #agents #ai
It runs 100% locally, powered by DeepSeek, which delivers OpenAI-o1 level intelligence at 96% less cost.
The video below shows a demo of how it works!
I have published a Lightning AI Studio using DeepSeek.
Tech stack:
- DeepSeek AI R1 served as LLM.
- LlamaIndex for orchestration.
- Streamlit for the UI.
We'll cover more advanced techniques with R1 very soon! ✨
Launch the studio now: lightning.ai/akshay-ddods/studios/rag-using-deepseek-r1?view=public§ion=featured
#rag #agenticai #llm #agents #ai
Переглядів: 1 342
Відео
Evaluate and monitor Agentic AI apps and RAG
Переглядів 271День тому
Let's build a dashboard to evaluate and monitor your Agentic and RAG apps! In this video, I'll guide you through creating an evaluation and observability pipeline for your AI apps using a 100% open-source tool! Tech Stack: - Comet's Opik to eval and monitor - LlamaIndex to build a RAG pipeline - Ragas for synthetic datagen You'll learn: - Setting up Opik - Building a RAG pipeline - Creating an ...
Build an agentic RAG step by step, powered by Crew AI and a local Llama 3.2
Переглядів 3,3 тис.14 днів тому
I just created a 100% local Agentic RAG app! It's powered by a locally running Llama 3.2 and has ability to search through your docs and fallback to web search incase it doesn't find the answer there. Tech Stack: - CrewAI for multi-agent orchestration - Qdrant to self-host a vector DB - Firecrawl for web search Primarily the app features two agents: 1️⃣ Retriever Agent The retriever agent is re...
Step-by-step CrewAI flows agentic-workflow tutorial to write socials posts, using a local Llama 3.2
Переглядів 702Місяць тому
Create an agentic-workflow to automatically write and publish content! It's powered by CrewAI Flow and Llama 3.2, running 100% locally. Tech stack: - CrewAI to build an agentic workflow - FireCrawl for web scraping - Typefully for scheduling Here's how it works: - You provide a link to a website. - It scrapes and saves the data as markdown. - A router triggers the desired Crew of agents. - The ...
ReAct AI Agents, clearly explained!
Переглядів 997Місяць тому
#agent #ai #llm #gpt ReAct (reasoning and action) agents, clearly explained! ReACT, is a framework that combines the reasoning power of LLMs with the ability to take actions! In this video, I'll clearly explain what ReAct agents are and how to build one using Dynamiq ✨ Dynamiq is like a Swiss army knife for AI Engineers, it supports: - RAG - Multi agent orchestration - Managing complex LLM work...
Build a multi-agent financial analyst using Microsoft's AutoGen and Lllama3-70B
Переглядів 3,5 тис.2 місяці тому
Let's build a multi-agent financial analyst using Microsoft's Autogen and Llama3:70B! Tech stack: - Qualcomm's Cloud AI 100 ultra for serving Llama 3:70B - Microsoft's Autogen for multi-agent collaboration Corresponding twitter thread: x.com/akshay_pachaar/status/1859575920912429464 Before we dive into coding, Get your API keys and free playground access to run Llama 3.1-8B, 70B: qualcomm-cloud...
Building RAG over complex, real-world documents.
Переглядів 6122 місяці тому
The biggest hurdle is document complexity. Real-world docs can be messy, with tables, images, and intricate flow charts. Traditional parsing and chunking methods struggle to handle these. So, what’s the solution? We need smart techniques that can intuitively chunk relevant content and understand what’s inside each chunk, whether it's text, images, or diagrams. In this video, I’ll walk you throu...
Make your RAG 32x faster and memory efficient! #rag #llm
Переглядів 4776 місяців тому
Make your RAG 32x faster and memory efficient by leveraging binary quantization of vector embeddings. It reduces memory over head as well as vector comparison becomes extremely fast. I'll be using a self-hosted Qdrant Vector DB for this, stay tuned for another video.
Compare Llama3 and Phi3 using RAG
Переглядів 2,4 тис.9 місяців тому
Link to all the code & detailed blog: lightning.ai/lightning-ai/studios/compare-llama-3-and-phi-3-using-rag?
RAG using MetaAI's Llama-3
Переглядів 7 тис.9 місяців тому
Here’s all the code and detailed blogpost: x.com/akshay_pachaar/status/1781299361600995516?s=46&t=mJEhrJTdTG_ZhCgPDaBxhA
Hey Akshay, thanks for making this video for us, may I ask what tools do you use to create those amazing demo videos?
Hey Akshay ! from Linkedin , thanks for the crisp n clear tutorials.
Glad you liked it! :)
Hello, I am new to this. is anyone getting this error? RuntimeError: Event loop is closed ERROR:root:LiteLLM call failed: litellm.APIConnectionError: OllamaException - [Errno 61] Connection refused
Awesome but if possible if you make full long explanation video tutorial sir
How do you edit your zoom ins and outs? Thanks!
asking for .edu mail
A video how GenAI capture the job market
Bro we want playlist on Gen AI Fine Tuning Arc
hey thanks for info even i tried to create my own tool which was vector search using mongodb vector db integrated with agentic rag worked so flawless
Is a full step by step video coming ?
Working on it. :)
whats ur cell number?
Yo! akshay great content. I followed you from linkedin though. Love to see more from you.
Happy to hear this! I plan to put more videos this year!
this project doesn't even work. so many errors on my environment. can you help me out with the issues. it says [ImportError: cannot import name 'FireCrawlWebSearchTool' from 'src.agentic_rag.tools.custom_tool' ]
This is awesome! I have a question though. Why is it that every CrewAi video I see ops for using a notebook like app like G. Collab or Juipyter Notebook (.ipynb format) instead of an IDE? I think it's because it's easier to not install all dependencies in your environment but I could be wrong. Thanks!
This is the recommend way, once you get comfortable with it, it's really great in term of modularity, reusing agents and tasks... To create a new CrewAI project, run the following CLI (Command Line Interface) command: crewai create crew <project_name> This command creates a new project folder with the following structure: my_project/ ├── .gitignore ├── pyproject.toml ├── README.md ├── .env └── src/ └── my_project/ ├── __init__.py ├── main.py ├── crew.py ├── tools/ │ ├── custom_tool.py │ └── __init__.py └── config/ ├── agents.yaml └── tasks.yaml
@ yeah I’ve tried this and never worked even though I had the dependencies, hence why I also went and used .ipynb lol
😂
Thank you for your efforts! However, the code seems to be breaking. I recommend uploading only tested and working code, as this will help attract more interest from people.
👍
Is it possible to integrate an Api instead of fire crawl
@@dhirajda...6316 yes, just like I created a custom tool for firecrawl you can do for any API
Awesome! I came across this from X and wanted to ask if this project would be a good addition to my resume as a fresher.
Go ahead! You can find more on our GitHub.
@akshay_pachaar okay, thanks
keep going💥
Cheers! :)
Hi Akshay, great video! I would like to know how would it respond to queries like Hi, Hello, Bye or something generic which doesn’t require fetching something from the internet or from a document. How would you deal with this?
Couple of things can be done here, either we can add this to the description of the retriever agent or create another agent that first tries to get the answer from the llm itself, then self-reflects on it if it's correct then respond back to the user with this else take a decision to delegate the task to retriever agent. But it should doable. I would avoid creating new agent and try to add this instruction in the retriever agent first.
Can i get an idea of what the example_thread.txt looks like pls? I think that could be the reason why I'm getting error scheduling with typefully.
@akshay_pachaar - which python library do the ImagineLLM and ImagineChat come from ? I'm getting an error on the CustomModelClient imagine module not found.
I am trying to replicate this but I can't seem to install the scheduler package. I can see the scheduler package is a custom package as you did code it. How did you integrate the scheduler package into your Jupyter notebook or which IDE did you use?
It a utility script for scheduling tweets that is at the same level as you main.ipynb, just import and use it
great, you did justice with the word "intuition" especially the picture example
simply excellent... short sweet to the point... no unnecessary ... Akshay could you come up with a video using python how we can achieve without groundx... Many Thanks...
Great explanation - the image comparison sold it. Thinking of doing this myself using 'Xenova/all-MiniLM-L6-v2' sentence transformer. My other consideration is calculating the binary vector in the front end and sending 1/0s to the back end on astra db. That should be a faster/more light-weight than sending an array of 600+ floating point numbers. Fingers crossed.
what is app in demo?
Phi-3 还是有好处, 假设的少, 就没有那么多幻觉.
你这个问问题的方法对Phi-3不是很有利, 因为回答你的问题需要调动背景知识解释, 这方面Phi-3就寄了, 但是如果回答问题来源都需要依据paper的话, 那Phi-3可能反而有些可靠性的优势.
hi. it would be better if you could share your github repo link.
Added to the video description
@@akshay_pachaar i dont see github repo link on your video