Hands on with LangGraph Agent Workflows: Build a LangChain Coding Agent with Custom Tools

In this video, I walk through how to build a LangChain-writing agent using LangGraph. I'll build up from the basics of manually managing a conversation with OpenAI + Tools and then walk through how to handle the same workflow with a custom agent built with LangGraph.
Interested in talking about a project? Reach out!
Email: [e-mail]
LinkedIn: linkedin.com/in/christianerice
Part 1: Running a manual conversation with OpenAI Tools
Part 2: Running a manual conversation with OpenAI plus Tools.
Part 3: Building a LangChain Agent to automate that process.
Part 4: Rebuilding the LangChain Agent with LangGraph.
Part 5: Building a LangGraph agent from scratch that accesses a vector database and writes LangChain code.
Video Chapters:
0:00 Intro
1:19 Manual Conversation
4:54 Manual Conversation w/ Tools
8:39 LangChain Agent w/ Custom Tools
12:10 LangGraph Agent w/ Custom Tools
20:25 LangGraph Developer Agent
Follow along in the code: github.com/christianrice/demo...

[@allenlawson9872]

Hey I'm building a health tech app rn and your videos are so helpful. Thank you !

[@curiouslycory]

I really appreciate all of your videos. For some reason the volume on all of the content that I've seen you post is like 50% more quiet than other videos on youtube. The "why" of how these things are put together is great in your videos and I'm always glad to have a better understanding of the underlying context.

[@deployingai]

Thanks for the feedback, I appreciate it! I'll have to look into the audio issue, I obviously haven't spent any time on production quality but hopefully that's a quick fix.

[@user-pr6nm2di6d]

Ultimate video that i was looking for. Can u make a video on
Building a RAG pipeline on schema of a SQL database like Postgres and chunk it & embedding it on to pgvector extension to load only relavant schema in order to optimise tokens and prompt size & then passing natural language to sql with defog SQL Coder and futher give insights like how this video works. Thanks in advance

[@user-bv2dr1vi2b]

Thank you so much it is great explaination starting from python only, then langchain , then langgraph and at end with real world scenario .. amazing

[@deployingai]

Thanks for the note and thanks for watching!

[@maskedvillainai]

Yeah. Actual machine learning steps ha

[@bwilliams060]

Brilliant! Thanks heaps

[@zyu5354]

This langchain codes work well for one Q&A, but how to modify these to allow interactive conversations, and still able to handle the tools? Thanks!

[@kenchang3456]

Helpful indeed, thank you very much. I'll be looking out for your more sophisticated hierarchical tutorial involving the Agent Supervisor and task-specific Agents. If you could include an Agent that creates Python code on the fly that the Agent would use to fulfill its task, that would be instructive.

[@bertobertoberto3]

That was excellent

[@VitthalGusinge]

what code examples did you pushed in weaviate ?

[@user-dr3cd4bx7l]

Hi, Can I use vector database sample you used in this video?

[@tidymonkey81]

hi thanks for the video. what is the extension or option that provides the coloured blocks for your indents ?

[@deployingai]

I use indent-rainbow, you can find it here: marketplace.visualstudio.com/items?itemName=oderwat.indent-rainbow

[@shnzomar7001]

awesome

[@smehra]

Very helpful. Could you pls post your notebook so we can follow along.

[@deployingai]

Thanks for watching! I added a link to the notebooks in the description.

[@hiranga]

@deployingai Great breakdown here 👌🏾- really appreciate the detail, particularly when LangChain docs are like navigating the desert - everything looks the same, but look again and everything has changed!
I couldn't help but notice the token counts in LangSmith @19:02
>>> your AgentExecutor version cost 102 tokens
>>> the LangGraph version cost 621 tokens😩
I'm guessing this is purely because it goes back to the Agent each cycle in LangGraph - or something else? Can this be avoided?

[@deployingai]

That's a great callout! The much larger token count in my LangGraph version was because I was using function calls as opposed to tool calls (tools can be run in parallel, functions cannot and they are now a legacy option). By switching to tools, we can remove the additional step you noticed that occurs between function calls and it brought the token count down to 46. I have to jump but I'll post the code and maybe a video when I have some time.

[@abhishekshingadiya4778]

if you share your based notebook that will helps alot.

[@deployingai]

Thanks for watching! I added a link to the notebooks in the description.

[@abhishekshingadiya4778]

Thanks, You are the best dude!@@deployingai