Create an AI Agent that Writes & Runs Code in FlowiseAI (No Coding Required)

[@leonvanzyl]

Hope you enjoy this video. What other Flowise topics or AI platforms should I cover next?

[@lingzhang2061]

Thanks, Leon, for another awesome tutorial. It seems like e2b code Interpreter can be embedded into a multi-agent system? I'll have to try. Also, I'd appreciate it if you could cover topics like how to set up parameters (e.g., numbers of iterations) for tool usage that can help agents generate the most accurate and reliable content. I created an agent system where multiple agents can use local databases to retrieve information, but sometimes it seems too complex (both time-consuming and costly) for different agents to use different tools (because I used OpenAI API). I'd appreciate your insights into those kinds of challenges. Thank you!

[@leonvanzyl]

@@lingzhang2061 you can definitely add e2b to Agentflows as well. Works exactly the same.

[@ShuaibShahzan]

Thanks for another great video Leon. If you can make a detailed video on accessing Flowise APIs that should be really helpful. LangGraph, LangServe and Make would be great platforms to cover next with an in-depth playlist.

[@dbose_ai]

Hi Leon, your videos are the reason why I have started using this tool. You are just awesome. I am exploring some of the enterprise scenarios and use cases - Integration with Microsoft AD/Entra ID, calling APIs by passing user login tokens, How to use variables in Flowise, Using or Publishing Flowise Agents / Chatbots in Microsoft Teams Channel, Difference between and when to use Agentflow vs Langchain Chatflow, How to create tools (GET/POST) that can work obo of logged in user, Published Chat UI customization for Enterprise use cases, Custom Telemetry / Observability using Enterprise cloud (Azure/AWS), Dynamic Stater Prompts etc.

[@leonvanzyl]

@@dbose_ai Thank you for the amazing feedback!! 🙏 Also, thank you for the many ideas.

[@laflammedenis]

I want to personally thank you for all the great work you’re doing. Your delivery style and simple yet very complete approach is a real treat to consume.

[@leonvanzyl]

Thank you for the wonderful feedback 🙏

[@krisszostak4849]

I'm a simple man. I see a new video from Leon so I press the like button 😆

[@leonvanzyl]

Thank you 😁

[@mjunaidasghar8613]

This is mind blowing man

[@leonvanzyl]

Agreed!

[@appleestrada6810]

This is a sample of a simple idea, well executed. Thanks Leon for another great tutorial!

[@leonvanzyl]

Thank you

[@RenatYaraev]

I join the above mentioned reviews and also want to thank you Leon! I just wanted to add that in order for the function or download button to work, you need to update Flowise to 2.1.3.

[@leonvanzyl]

Tha lnyou very much 🙏. Yes, good point. You need to update Flowise

[@myandrobox3427]

❤As always, huge thanks Leon!! 💯💥

[@leonvanzyl]

You're welcome 🤗

[@Statical_Life_]

Sir Please make a video on Invoice Extracter in csv or excel file format and Thanks Sir from India for another great tutorial!

[@maniecronje]

Thx Leon great FW video ... 👍👍

[@leonvanzyl]

Dankie Manie

[@jkaramk]

🔥🔥🔥🔥

[@HassanAllaham]

As usual .. very useful and very good content.. This channel is like a drug I became addicted to.. I wonder whether FlowiseAI can make the code run locally (I know it might be dangerous but since it can be free when using something like Ollama and I believe we can add the user in the middle so the code can be verified before executed).. Thanks for the very exceptional content.. Leon keeps shining and glowing 🌹🌹🌹

[@leonvanzyl]

Hey there, thank you for the awesome feedback!

[@JoaoMarra-s1r]

Good afternoon, Leon! First of all, congratulations on your tutorials; they have been a tremendous help to me! I'm leaving this comment because I'm having trouble getting my AI to interpret images and describe their content. Is there a way to do this? I would be very grateful if you could make your next video about this topic.

[@WayneBruton]

Hi Leon, Thanks for that. Interesting I find GPT-4o quicker than Claude 3.5 Sonnet.

[@leonvanzyl]

Claude is better at writing code, so works well with a code interpreter. You can try GPT, but if you run into issues then try swapping out for Claude

[@WayneBruton]

@@leonvanzyl Yes thanks, playing with both

[@BadreddineOtky]

Thank you so much for the great content again!! I know it is too much asking, but is it possible to have a video about large PDF embedding and retrieval using a local DB (Milvus, Qdrant or any other)?

[@mikecallahan5441]

Your Flowise content is super valuable and educational! This solution with E2B is just what I was looking for. Any idea why the artifacts aren't displaying for me?

[@leonvanzyl]

I think there's a small bug at the moment. Simply close and reopen the chat

[@laflammedenis]

Very well done. What a great way to let our agents show data. I need a tool that I can count how many milliseconds a user takes to say a word as soon as it appears on the screen. I hope you guys can help me at some point.

[@leonvanzyl]

I would build something outside of Flowise itself to generate the word and store it in the database with a timestamp and then expose APIs to generate or send the word. These APIs can be called from Flowise custom tools. I know that's a simplified explanation, but hope it points you in the right direction 😁

[@altonjones5271]

Fantastic tutorial as always Leon - I have been trying with GPT-4o mini (Claude 3.5 is a bit pricy at the moment) - but the results are pretty bad. I guess you get what you pay for lol

[@leonvanzyl]

Yeah, you'd have to use a more advanced model. Have you tried GPT4o?

[@altonjones5271]

@@leonvanzyl Will try now - its a pity because GPT 4o mini is very fast also

[@DevinHenkel]

Thanks for all the great content! Any idea why the artifacts are coming through as broken links to FILE_STORAGE png files?

[@abcthegreat1]

@leonvanzyl have you figured out how to enable the PDF support beta for Claude? Apparently greatly increases PDF understanding vs straight OCR

[@leonvanzyl]

I'll look into this

[@micbab-vg2mu]

thanks I will try it:)

[@leonvanzyl]

It's a super useful feature

[@MH-xx6df]

Great video sir! Probalem: when I hosted flowise on the free tier with Render, I'm unable to upload files. Know why? Also, how could I hook up a API to this rather than file upload please?

[@shovonjamali7854]

I have a pipedream workflow, that is processing some data from an external source. the pipedream flow is being called by a tool in flowise. now if i want to generate the chart using this code interpreter, how can i do that like the data and all the things are in the pipedream flow.

[@luisortiz9540]

Hello Leon, im just update flowise, but for example don't show me the option "load file" in configuration, i don't understand. Show me until "Leads"

[@k1r0vsiii]

any idea why im missing the file upload function in config? flowise doc says u gotta have a vector store present and a document store... from what i quickly searched.. but u dont have in tise workflow? thx flor the videos as always

[@leonvanzyl]

I'm not using vector stores in this tutorial either. Perhaps you need to update your Flowise instance?

[@MisCopilotos]

Which is better: E2B code interpreter or OpenAI assistants code interpreter?

[@eric-xn7ll]

Hi! Can you do a tutorial on how to extract Structured Data (json) from documents? Imagine i have a template where extract data from a curriculum and return it as json using structured output parser. Right now i dont know how to do it, its not that i will send every time what i want to extract, that would be in the prompt template. Thanks

[@leonvanzyl]

Check out my latest video 😁. kzbin.info/www/bejne/m2KoZHWqeZ55q5Y

[@WayneBruton]

Hi Leon, quick question, when deployed, I don't see the code interpreter, is it possible to show it?

[@leonvanzyl]

You need to update your Flowise instance 👍

[@WayneBruton]

@@leonvanzyl Hi, I did because I needed the file upload, still no joy

[@DataAI-k9x]

The file loader is not optimized for large files or even small file. my 20kb file gave error - context error GPT 4 context is smaller then the content of the file + the prompt . Best way can be using VDB or a RAG kinda solution with E2B Code Interpreter I think :(

[@leonvanzyl]

You need to use a model with a large context window, like the Claude 3.5 Sonnet model as shown in this video.

[@Obinna-ai]

Great tutorial. I noticed it uses quite a lot of tokens/credits. Is that normal?

[@leonvanzyl]

I would imagine it depends on the size of the file. Remember, the entire file is loaded into the context. In a production version of this you might want to clean / filter the data before running the prediction.

[@kirubhakaran6576]

if you could add more videos in langflow it will much helpful.

[@leonvanzyl]

Will do. Just waiting for Langflow to get a little bit more stable. Been having issues with it lately.

[@kirubhakaran6576]

@@leonvanzyl i completly agree with that, but he it has huge potential in deployment flexibility than flowise.

[@BadreddineOtky]

why, oh why are there dislikes about this video??? 🤣🤣🤣

[@leonvanzyl]

I know 🤷

[@bodyguardik]

Why not running code locally in some sandbox, like dev container?

[@leonvanzyl]

That's possible as well. Since e2b is open source it's probably possible to self host it locally.

[@genx65535]

@@leonvanzyl I can only find e2b cli packages and they insist on connecting to e2b's cloud service. However posing the question to perplexity shows it's pretty easy to stitch together a few python packages (docker-py, jupyter-kernel-gateway, and dask) that achieve the same core functionality.