LangGraph is an amazing approach to developing agents. It really helps to differentiate between the "Agent" and the "Agent Job". So much work went into agents frameworks. LangGraph is a framework for "Agent Job". I think next, would be great to bring Object Oriented Programming concepts to Agent Jobs: making them self-contained, hierarchical, etc. great work. looking forward to next.
Oh wow, you guys are killing it. I was struggling to debug a complex langgraph system and found myself have to remember and know what state flowing through each step. THIS will help tremendously! THANK YOU!
I've been struggling to test my projects with this conversational logic and human feedback. This tool is just going to solve my problem and make the experience even better. You guys are amazing, awesome work!!
Wow! Fantastic work. I'm really excited to try this. I've been learning Langgraph and starting to get up to speed, but it can still be a bit mystifying to know what is happening at any given step, so I'm sure this will provide a great way to understand the flow in a much more transparent way. Yay!
This is a great vision fwd in the development and deployment processes of Agentic applications! Well done! I just wish you had a Linux version also ready to try.
Awesome Demo! Unfortunately I'm not using Apple Silicon though so i cant use the Beta Download yet :( Fingers crossed we get a PC / RTX compatible release soon!
This is a good start but very limited, we need to be able to build agentic apps in a visual way and be able to configure and control every edge and every node like in LangFlow, you guys should get inspiration from their IDE
Have you guys looked at prompt chaining to get optimal fused outputs from multiple agents? The effect is RAG-like refinement but just due to using multiple agent generations simultaneously.
Very cool thank you for making this video and tool. Will this tool allow me to make API calls with Vertex AI and upload video files for analysis? None of the other tools like Langflow, Flowise, have a working function for this.
Can you tell us more about the built in persistence layer that you mention at the start? I've been trying to get the custom checkpointer working using Postgres but it's failing around the pending_writes. Would be good to have a deeper dive into this.
Add a feature to connect to sonnet 3.5 to help in code suggestion for the studio. The APi key can be placed by us even the users can chose the model they way maybe Ollama for local dev or maybe if someone wants or has exo running llama 400b or even api models that there is. Would would make the code iteration even faster
Cool! Thanks for going into better ovsrvability, but I wonder how it would work with a more complex example for a real assistant where you have router and each chain consist of multiple dynamic elements.
When you say “first” or “fastest”, cheapest biggest, smallest. Etc. you’re legally claiming to have sought proven sufficient of research specifically that no IDE, or effectively any input used by a developer target, has existed. Warp it’s already beaten you to this. They are not a CLI because there is literally no difference. In fact, there are more IDE then some of the open source boiler plate ID DE repositories out there. Just, like, don’t do that. Don’t say that. It literallyprovides zero gain and all risk and skepticism if I’m being honest by default for being anything of the first. That is why Apple never releases first, aside for being a brand, promise of completion and not speed.
great job Chase!👍, but is there documents? I encountered : failed to solve: process "/bin/sh -c PYTHONDONTWRITEBYTECODE=1 pip install --no-cache-dir -c /api/constraints.txt -e /deps/*" did not complete successfully: exit code: 1
Langraph and Langchain's abstraction didn't make sense to me at first , Now it makes way more sense with this.
Same!!
I love it but pls. make it available on non-Apple devices too!
LangGraph is an amazing approach to developing agents. It really helps to differentiate between the "Agent" and the "Agent Job". So much work went into agents frameworks. LangGraph is a framework for "Agent Job". I think next, would be great to bring Object Oriented Programming concepts to Agent Jobs: making them self-contained, hierarchical, etc.
great work. looking forward to next.
That´s really cool, when will it be released for windows?
I don't usually comments on videos but this what is really wanted to build to improve llm agents. Awesome work guys
Oh wow, you guys are killing it. I was struggling to debug a complex langgraph system and found myself have to remember and know what state flowing through each step. THIS will help tremendously! THANK YOU!
I've been struggling to test my projects with this conversational logic and human feedback. This tool is just going to solve my problem and make the experience even better. You guys are amazing, awesome work!!
Debugging the agents behavior.. damn, that's something to play with.. good stuff guys
This is excellent. I can foresee the graph to be a node graph where we can put different components for a library. Very cool.
Yes.... now we're getting somewhere! Thx!
The future is so exciting!
Lance for President!
Can't wait for the way you'll shape the future
Great stuff ! Can't wait to try this out - will make prototyping agents way faster
Wow! Fantastic work. I'm really excited to try this. I've been learning Langgraph and starting to get up to speed, but it can still be a bit mystifying to know what is happening at any given step, so I'm sure this will provide a great way to understand the flow in a much more transparent way. Yay!
This is a great vision fwd in the development and deployment processes of Agentic applications! Well done! I just wish you had a Linux version also ready to try.
Well done, thanks! I was wondering when LangChain was going to do something like this.
Cool. This saves my the time using an open source flow chart tool and export the nodes in json, and have Gemini turn into the agent. ❤
Wow. This is amazing! Can't wait to try it out. 👏👏
Fantastic new feature!! Keep working as you are! Thanks
🤯 breakthrough in human in the loop
Awesome Demo! Unfortunately I'm not using Apple Silicon though so i cant use the Beta Download yet :( Fingers crossed we get a PC / RTX compatible release soon!
🎉 enjoyed have fun
Great when you guys post , keep it up
This is amazing. Would love to use it 🎉
Edit: Typescript support on roadmap?
This is AMAZING, keep up the good work !
This is a good start but very limited, we need to be able to build agentic apps in a visual way and be able to configure and control every edge and every node like in LangFlow, you guys should get inspiration from their IDE
Have you guys looked at prompt chaining to get optimal fused outputs from multiple agents? The effect is RAG-like refinement but just due to using multiple agent generations simultaneously.
A visual way to quickly create nodes/edges would be incredible.
Wow, that's increadible!
Feels like Jupyter and Docker multi-stage builds, where you can benefit from maintaining state without starting over from scratch each time.
Hi, this is really useful. Possible to have LangGraph be used with a simple mobile device (e.g.mobile phone)?
how to run with local LLMs?
The moment you can add agents usings this IDE, it would be awesome
Great tool. Thank you.
When will windows version be available?????
Replays from logs, this would be a huge stuff.
Very cool thank you for making this video and tool. Will this tool allow me to make API calls with Vertex AI and upload video files for analysis? None of the other tools like Langflow, Flowise, have a working function for this.
Feels like early GPT when the model was Ada, etc. Very cool.
Can you tell us more about the built in persistence layer that you mention at the start? I've been trying to get the custom checkpointer working using Postgres but it's failing around the pending_writes. Would be good to have a deeper dive into this.
Add a feature to connect to sonnet 3.5 to help in code suggestion for the studio. The APi key can be placed by us even the users can chose the model they way maybe Ollama for local dev or maybe if someone wants or has exo running llama 400b or even api models that there is. Would would make the code iteration even faster
Plans for major platforms?
Oh wow, more lock-in into the usage of paid LangSmith
This is awesome! I tried the same example but I had this issue: "unknown flag: --watch" any tips to get around it?
Need to upgrade docker desktop - we're adding better documentation and error handling for this
@@LangChain Ok nice! Thanks ! :)
Thanks! this defly gonna help me a lot.
Is this available for Ubuntu or just Apple?
Is there a pdf for the documentation that I can look at? Thank you!
Weird this was released only for mac. I mean its just a node editor right?
this is a game changer
Cool! Thanks for going into better ovsrvability, but I wonder how it would work with a more complex example for a real assistant where you have router and each chain consist of multiple dynamic elements.
try it out and let us know :)
Super super thanks!
Use it to create a set of Agents that will create a set of agents for a given natural language prompt ;-)
Hopefully you'll make this available for Linux or Windows soon, I know its probably a pain though. Just asked someone for their old macbook
How does this differ from Flowise's Sequential Agents?
Why go back into my code and edit the prompt why is that not part of this UI?
Windows version please?
huge!
What can we build using this ?
Real life scenarios
my boy chase is really up to something here.
This is EPIC!
When do we get to use this in windows?
Now we're cooking!
Can't run it keep getting Docker not installed or not running: ExecaError: Command failed with ENOENT: docker info -f json
spawn docker ENOENT
How to handle router logic if written between multiple agents
Windows please
amazing but why only macs tho...
insane
Damn this is good
The language change should be useful to be done via UI...
Did anyone find a way to try this on windows?
It will be more convenient If it is integrated to tradition IDE, VS code, pycharm, as plugins
Oh damn…. 😮
I'm building something like this for TS, I'll make it open source as soon as it's ready
Can I sign up somewhere to be notified when it's ready? Really cool btw!
Seems like unnecessary overhead when all this can be done directly in python. Am I missing something? How is this better than just writing code?
Looking for windows release nıw
I love it but that's not Old English
Fork and Retry is a very unfortunate name for a button most especially if you're scouse and saying it...
wr
unfortunately linux is out of the game...
Are you one who watched whole video and looking for Windows setup? 🤣
hasn't n8n solved this already?
Putting those shackles on your AI slaves now huh.
When you say “first” or “fastest”, cheapest biggest, smallest. Etc. you’re legally claiming to have sought proven sufficient of research specifically that no IDE, or effectively any input used by a developer target, has existed.
Warp it’s already beaten you to this. They are not a CLI because there is literally no difference. In fact, there are more IDE then some of the open source boiler plate ID DE repositories out there. Just, like, don’t do that. Don’t say that. It literallyprovides zero gain and all risk and skepticism if I’m being honest by default for being anything of the first. That is why Apple never releases first, aside for being a brand, promise of completion and not speed.
Please release linux or windows pleaaaseee
great job Chase!👍, but is there documents? I encountered : failed to solve: process "/bin/sh -c PYTHONDONTWRITEBYTECODE=1 pip install --no-cache-dir -c /api/constraints.txt -e /deps/*" did not complete successfully: exit code: 1
Windows please