Automating High Quality Research Reports with LLMs | From Prompts to LangGraph
In this video I cover some interesting ideas for creating research reports. I start with a naive approach with a single prompt to ChatGPT using the OpenAI api, then I follow to break the problem down into multiple prompts. Finally we tackle it using LangGraph, with an agentic workflow involving calling tools to search the web and properly structure the output.
📚 Chapters:
00:00 - Introduction to Creating Research Reports with LLMs
00:05 - Basic Implementation with GPT-40 Mini
00:18 - Lingraph Framework Overview
00:24 - Traditional Research Report Workflow
00:54 - Using AI to Automate the Workflow
01:06 - Initial Prompt Example with GPT-4o Mini
01:24 - Creating a Naive Research Report
04:01 - Breaking the Problem into Smaller Parts
05:00 - Function Calls for Outline, Draft, Review, and Final Report
06:38 - Improved Quality of the Second Version
07:46 - Identified Problems with References
08:04 - Ideal Scenario for Research Reports
09:19 - Using LangGraph for Improved Workflows
10:07 - State Initialization in Langraph
11:07 - Tool for Fetching Archive Papers
12:01 - Web Search Tool Setup
12:46 - Testing the Web Search Tool
14:42 - Final Answer Tool
15:10 - Setting Up the Oracle Prompt
17:11 - Testing the Oracle
18:19 - Defining the Run Tool Function
19:55 - Setting Up the Oracle State
22:14 - Conditional Routing in the Workflow
23:38 - Compiling and Visualizing the Graph
24:46 - Running the Final Example
28:08 - Final Function to Build the Report
35:26 - Displaying the Final Report
35:46 - Summary of Achievements and Improvements
🔗 Links:
Subscribe!: / @automatalearninglab
Tiktok: https://www.tiktok.com/@enkrateialucc...
Twitter: / lucasenkrateia
LinkedIn: / lucas-soares-969044167
AI project tracker from Ben's Bites: https://bensbites.beehiiv.com/subscri...
Support the Channel!
Buy me a cup of coffee: https://tr.ee/7tYsD-tUu2
Paypal: https://paypal.me/lucasenkrateia?coun...