@user-uv3nv2bc6v Thanks! I'm glad you enjoyed it.

Hey Peter, I'm glad your enjoying the agent and that it's working for you. The Flowise embed widget currently does not support HTML rendering for some reason. You have to use another embed widget such as VoiceFlow's embed widget to render HTML. That's the one I'm using in the video. It turns out you can export Voiceflow templates, what I'll do is add the Voiceflow file to the code download later this week. You'll need a Voiceflow account although they come with 2 chat embeds before you have to upgrade to a paid account. You don't have to wait for the template though, you can signup for Voiceflow and use the Flowise api to connect to your Voiceflow widget. That's how I was able to render HTML.

Thanks. I found that if I embed the flowise share chatbot link in an it also works Cheers

That's a good question. Both Flowise and N8N each have their own unique strengths. When it comes to designing AI workflows, Flowise has more AI focused tools as well as a better chat interface. As far as agent orchestration, Flowise also has more options than N8N since it's built from the ground up for AI. For this build though you could also do it in N8N since it's pretty straightforward. I just tend to use Flowise though for designing AI Agentflows because its interface is more intuitive in the same way that N8N's interface is more intuitive when creating automations. They work pretty well together as a team.

@ That’s interesting and thanks for the speedy reply. I recently created a Competitor Analysis tool using a n8n workflow where I used a Flowise workflow as a tool for a LLM tasked with research. The Flowise workflow contained an agentic rag flow using the LlamaIndex nodes to interact with 15 company knowledge bases while the n8n workflow took care of constructing the final reports. Dead easy to setup and super satisfying to see it working.

You could extract data from any identification as long the text is visible. It's pretty similar to the part in the video where I extracted data from an image of a table of survey data. You would just need to adjust your prompt and if necessary use structured JSON outputs.

@OnyxStudiosInteractive Most of LLMs are sensored and personal information in any Identification images are not alloweded to be extracted.

@ I can understand that, they don't want their models to be used to collect sensitive and personal data without user permission.

@@AneeshaRao100 Sure you can do that with an LLM chain. Just connect the output of the LLM Chain module to the input variable of the Custom JS Function module. Alternatively you can also use an agent like the tool agent....you just connect two different Custom JS Functions..one calls the tool agent via API and then passes the output to another Custom JS Function node.

That cool to hear about Xano...there are so many new tools being developed right now and integrated with existing plaforms. There's a lot of options for vector databases that are pretty solid and easy to implement into your application. You can even use the Flowise API to embed documents programmatically into a vector store. As far as using API's in N8N, you can use nodes like the HTTP node in N8N to retreive data using a HTTP Curl request. That's what I did for the Brave Search API in the video. You just need to have the right parameters for the API request. If you'd like to talk more about you project let me know and we can schedule a call.

@@OnyxStudiosInteractiveThanks a lot! I am testing a few things as I build the n8n flow and ingestion flow. Your video already gave me some ideas to test 👍🏻 But will surely come back if I am stuck. Thank you!

Thanks a lot. Your right, Part 1 is more about Flowise and an intro to the N8N tutorial in Part 2. Let me know if you have any questions. Once you get the ranking workflow up it becomes possible to go beyond and even add features not found in the Langchain and enterprise rerankers.

I'm glad you enjoyed the video. We're using SQL because we need a database to store and retrieve the product information. JSON is simply a data format. SQL also gives us an easy way to query the database using AI and automation.

@@OnyxStudiosInteractive Thank you Thomas for the clarification!

This is essentially a single agent flow. In this case the agent uses a variety of tools and training to turn your request into an SQL query and then filter the results to recommend products based on what you're looking for. That's why it's an agentflow. It's not a multi-agent or a sequential agentflow where multiple agents are working together. You could do that but it would be more complicated and potentially take longer. As far as the SQL Prompt template, I used it as a starting point but the template is intended to be more of a concept. It is not plug and play and takes a lot of work and customization for a specific use case.

​@OnyxStudiosInteractive Great video However, one could argue this is a simple RAG system as well. Converting the query into embedding, converting products or their descriptions into embeddings and saving them in a database. Perform a similarity search between the query embedding and the products embeddings, extract most relevant products and recommend them to the client. There is no thinking in this flow, thus no need for an agent.

@@dantshisungu395 I've tried that before and it doesn't scale very well. The issue is that large ecom product databases are constantly changing with inventory updates and you can't really rely on simple embeddings and similarity search when you have potentially thousands of products. It's much more reliable to search the database first and then choose from the products that are currently available.

​@OnyxStudiosInteractive Thanks a lot for the feedback, I will try to implement it and see how it works. However, note that it is not a matter of quantity here, rather how to effectively embed products' descriptions into the right multidimensional space. Maybe trying out different embedding models and see which one performs better

@@dantshisungu395 Sure thing let me know how it works for you. You could also try graph vector databases like neo4j as well however the real challenge is always going to be updating your embeddings and syncing it with your entire product inventory database. That's why SQL query language is so powerful.

@@healthadvice3091 That's right Flowise is Open source and you can deploy on your own computer, server or a service like Railway. Although the cloud version has a few additional features like storage and database backup that you have to provide yourself in the self hosted option.

@@OnyxStudiosInteractive wow i like it as i am from 3rd world country (im a broke tech guy) i just need to find a cheaper alternative hosting

Hi there, that sounds interesting...We may be able to setup a referral program for you. Send me your email and we can discuss it.

You're welcome. Sequential agents are pretty awesome.

Hi Bill. Thanks for your interest. The course is brand new. We produced and released it in August so it's current with the latest version of Flowise. However it will also be updated as new features and versions of Flowise are introduced.

@@OnyxStudiosInteractive Great so it was produced using v2? Do the sample projects in the course include multi or sequential agents or doesn't it go that deep? Thanks.

​@@Bill-ai That's right, it was produced with Flowise version 2. There's actually a project in the course that includes an AgentFlows team along with an N8N integration for downloading KZbin transcripts and web search to research and create different types of content. I wanted to keep it easy to follow while also including some of the latest Flowise features from v2.

You're definitely right about that. It's amazing that we're on the cutting edge of AI development and we haven't even gotten started.😅

Thanks for watching!

Thanks! I'm glad you liked it. haha need to make an animation next time😅

I don't think Agentflows supports custom Session Ids since they're stll pretty new at the moment. You may want to ask in the Discord group.

Thank you! I'm glad you enjoyed it!

They are still pretty new and will go through more development over time. What's most exciting is giving them the ability to add a human in the loop input at any point in the agent flow as well as displaying the text output in the frontend chat window. Amazing to think that we're still in the early days of AI development with the best yet to come!😅

@@OnyxStudiosInteractive

Thanks! I'm glad you enjoyed it. As far as sharing the code, I'm actually going to be doing more of that and making the workflows available for download since it can really save people time and help them develop their own agentflows. The flow for this Agentflow is now available in the store on my website: onyxstudiosinteractive.com/flight-booking-agent-download

You're welcome! That's a great question...you can probably recreate most chatflows in N8N as long as you have Langchain modules. However I think Flowise has a much more intuitive user experience. N8N is primarily an automation tool though so it's more designed to work with Flowise rather than replace it.

I'm glad you enjoyed it!

I am still waiting your reply. just say yes or no.

Hey Thanks! I'm glad you enjoyed the videos! While there are some things I'm unfortunately not able to share, I am working on a membership site where I'll be able to share some of the flows and assets from the episodes. Hopefully it will be finished this month.

To do that you should use the tool agent combined with the retriever tool in Flowsie

@@OnyxStudiosInteractive thanks, I'll try that today.