LangFlow Tutorial #2: Agents and Tools

Рет қаралды 9,501

Күн бұрын

#langflow #langchain #openai
Langflow makes it easy to create AI applications without writing a single line of code.
In this video we will compare agents to chains, and build our first agent chatbot.
☕ Buy me a coffee:
www.buymeacoffee.com/leonvanzyl
📑 Useful Links:
OpenAI API Key: platform.openai.com
SerpAPI: serpapi.com
💬 Chat with Like-Minded Individuals on Discord:
/ discord
🧠 My AI Automation Agency:
www.cognaitiv.ai
TIMESTAMPS:
00:00 - Intro
00:26 - LLM Chain Limitations
02:00 - Building our first agent
03:11 - Agent Types
03:44 - Adding Tools
05:39 - Adding Google Search
06:19 - SERP API Key
08:19 - Conversational Agent
09:53 - Shameless plug

Пікірлер: 66

@leonvanzyl 11 ай бұрын

Apologies for the scratchy audio at certain points in the video. I only noticed this during editing and didn't want to delay getting a video out to you guys.

@shrvn110 11 ай бұрын

it doesnt matter! ther is value all over the video! thank you

@leonvanzyl 11 ай бұрын

Thank you kindly 🙏

@koen.mortier_fitchen 11 ай бұрын

It's all good

@lifestyleblink9184 11 ай бұрын

Leon, I'm a huge fan of your tutorials! Your explanations and simple demonstrations have completely transformed my understanding of AI. Thank you so much for sharing all your amazing AI tutorials. I truly appreciate your hard work and dedication!

@leonvanzyl 11 ай бұрын

Incredible comment. Thank you very much 😊

@edgecrush3r 3 ай бұрын

Huge fan of your videos (todays discovery). Hope to see more Langflow tutorials.

@leonvanzyl 3 ай бұрын

Coming soon! Waiting for V1 to drop

@davidborges4623 10 ай бұрын

Keep going! Very good videos really practical! Hug from portugal

@leonvanzyl 10 ай бұрын

Thank you!

@koen.mortier_fitchen 11 ай бұрын

My favorite langchain learning channel because it's so efficiently explained. And I follow practically all AI tutors. Channel is very underrated at this point but it will exponentially grow its subs I'm sure. Leon, any thoughts on the limit of the number of tools? Meaning, would a model malfunction if ALL tools were added🤔 I heard another tutor say it would slow it down because the code would be too large.. but I'd like to hear it from you before I believe that. Thanks for your work.

@leonvanzyl 11 ай бұрын

Thank you for the feedback 🙏. I would recommend that you only load the tools that are required by the agent. I personally haven't tested the performance impact of assigning lots of tools, but I did see the bots get "confused" and started using tools when it wasn't needed.

@C.Radchaneeporn 9 ай бұрын

very informative video, Thank you so much !

@leonvanzyl 9 ай бұрын

You're welcome

@karthikeyakuncham6929 4 ай бұрын

amazing......we want more videos from you on langflow.

@leonvanzyl 4 ай бұрын

Will do 👍

@ProminentFox 3 ай бұрын

When's the next langflow tutorial? 😊

@leonvanzyl 3 ай бұрын

@@ProminentFox I've been in contact with the Langflow team. They're releasing a major update any day now and asked me to hold off on doing the tutorial. Should release the new series soon 👍

@somag6810 8 ай бұрын

Awesome Leon

@leonvanzyl 8 ай бұрын

Thank you

@altonjones5271 11 ай бұрын

Great content as always Leon - have you found any way to manage API keys in Langflow , like env file we would normally use in VS code?

@leonvanzyl 11 ай бұрын

Hey Alton, I am fully onboard with that idea. Unfortunately the nodes are not defaulting the keys based on environment variables at this stage, but it would make sense that these tools implement that option at some point.

@bambanx 4 ай бұрын

thanks you

@leonvanzyl 4 ай бұрын

You're welcome 🤗

@thecooler69 3 ай бұрын

what is the difference in connecting Calculator to a new openAI node instead of the existing one? does using a separate openAI node give it its own conversation instance?

@musumo1908 11 ай бұрын

Hey another great one. look forward to the next langflow. Very interesting after flowise. I do like to see whats going on at the backend! Can I ask does langflow support API calls into other apps e.g. botpress? If so whats the easiest way to do this. I can do this in flowise via POST but not sure in langflow. Am a low coder! Thanks a lot

@leonvanzyl 11 ай бұрын

Glad you liked the video 👍. LangFlow does provide API support, but there is no way to protect these API routes yet using API keys. I've been holding off from creating tutorials on Langflow APIs (and deployment) because of this.

11 ай бұрын

Thanks for this tutorial. It is very usefull. One question: when I click on "chat icon" my screen shows only the chat, there isn´t that left part with options (input variables). What do I need to do?

@leonvanzyl 11 ай бұрын

Thanks for the feedback. That's a bit odd.. I'll look into it

@YourGuy4AI 9 ай бұрын

@@leonvanzyl any update? I jave teh same issue

@YourGuy4AI 9 ай бұрын

I also have several mistakes whem tyring to run my application: 1. Error: Could not parse LLM output: `Hello! How can I assist you today?` 2. Error: One output key expected, got dict_keys(['output', 'intermediate_steps']) Any ideas why and how to solve? Thank you!

@thewarrior7570 3 ай бұрын

Can you explain how to make a custom tool and thank you

@maverick1901 11 ай бұрын

can you also provide multi-agent examples with local gpt4all/huggingface llm's (including one for the Calculator)? The OpenAPI examples are all not feasible for professional usage (data privacy etc.)

@leonvanzyl 11 ай бұрын

Hey there. I personally haven't seen good results using local models with Agents, but I'll definitely create a tutorial once I get it working 👍.

@TheShierl 4 ай бұрын

Great Video Series Leon! I have started playing with Langflow, wondering if it is only meant to create chatbots? My use case is feeding data into a LLM (i.e. 100 Products from a list) enrich it via the LLM and then write the results back to the list. If possible, can you do a video how to support this, writing output back to a list? Keep on the great work!

@leonvanzyl 4 ай бұрын

I would like to find out more about your use case, if you'd be interested in sharing. I'm working on an updated series on Langflow, but I think n8n might be a perfect match for you. Happy to create a tutorial on your scenario

@TheShierl 4 ай бұрын

@@leonvanzyl I am very happy to share my use case in detail - I think it is relevant for so many scenarios => enriching existing data with "LLM generated insights at scale".

@TheShierl 4 ай бұрын

@@leonvanzyl Not sure how to directly connect - I sent you and invite on LinkedIn ;). Looking forward to exchange

@user-ez5nm9sx7e 11 ай бұрын

Leon, quick question : Can we cite source URL link after every answer to flowise and Langflow?

@leonvanzyl 11 ай бұрын

Good question. I'll look into it.

@CarlosChang11 11 ай бұрын

Very good video! is it possible to create different type of agents and chain them? for example create Agent "Chef" that is really good at answering cooking questions and Agent "Hotel" which can answer any question regarding Hotel question. Chain them and have a bot that can be really good at both... this would make an amazing video! looking forward to it!

@leonvanzyl 11 ай бұрын

Thank you Carlos. I've only used one agent per project, but I'll look into this for you. As a rule, if I need two distinct agents, I would create them as seperate projects. I then tie them together using bot builders like Botpress or Voiceflow. I clearly need to create series on these bot builders 😁

@user-rr5ic6ts7p 11 ай бұрын

Thank you for the tutorial! By the way, do you happen to plan on making a video about AutoGPT in Flowise or LangFlow? It would be very timely, as there is absolutely no precise information anywhere on how to use it.

@leonvanzyl 11 ай бұрын

I'd be happy to make a tutorial on it. Do you have an example of why you would want to use AGPT in Flowise / Langflow?

@user-rr5ic6ts7p 11 ай бұрын

@@leonvanzyl Great! I'm developing an application that selects products from the internet based on a request, provides a link to them, and then compares them. I tried using other agents, connected them to the internet, they did a search, but did not provide very high-quality answers due to the number of iterations. It would be great if you show how to do it in Flowise

@josephn7392 11 ай бұрын

An though on how to call API via flowise ? Theres a Get API option but no documentation around it. For example calling google places API via flowise

@leonvanzyl 11 ай бұрын

Since this is a series on Langflow, I assume you are referring to the API tool on Langflow. I'll look into it and create a video 👍

@gankam 11 ай бұрын

o you think doing a short video to compare flowise and langflow would be a good idea?

@leonvanzyl 11 ай бұрын

I am most likely going to create a comparison video at some point.

@j0hnc0nn0r-sec 3 ай бұрын

Please do more videos on Langflow and how to incorporate chatbot into website

@leonvanzyl 3 ай бұрын

Will do. Just waiting for the latest release

@VincenzoManzoni Ай бұрын

Also, how to integrate the bots into Telegram or WhatsApp

@manthanpatel2024 11 ай бұрын

need more flowise video please

@leonvanzyl 11 ай бұрын

They're coming 😄.

@practical-skills-school 11 ай бұрын

Looks cool. But there's no Auth. So how to host it on a server without any password protection? Can't understand that. And leaving it on a local device is useless for production needs.

@leonvanzyl 11 ай бұрын

You're jumping the gun a bit 🤣. We'll cover production deployment in a future video in this series. I trust you build and test in dev before deploying to prod.

@practical-skills-school 11 ай бұрын

@@leonvanzyl looking forward! In flowise I can easily protect my account with password, but I failed to find it in Langflow

@leonvanzyl 11 ай бұрын

Hmm.. this seems to be a major oversight on the Langflow side 🤣. I was hoping that we could set the credentials in the environment variables (like with Flowise), but that doesn't seem like an option at this stage. I will create a video as soon as this feature becomes available.

@musumo1908 11 ай бұрын

Its a huge oversight! Someone said they are adding this soon!

@karthikeyakuncham6929 4 ай бұрын

@@leonvanzyl please create

@mishaelo689 11 ай бұрын

Please what's the difference between langflow and flowise?

@leonvanzyl 11 ай бұрын

Fundamentally, Flowise uses the JS version of Langchain and Langflow uses the Python version of Langchain. Other than that, they are very similar. The Python library is slightly more advanced, but the JS library seems to be catching up.