Thank you so much for your feedback on Dify in your video! Truly appreciate your support. By the way, we are working on our documentation and navigation for new users :) So stay tuned!
This was a brilliant episode, thank you so much! One thing that may not have been obvious (or I missed this in the doco), you have to restart the SearXNG docker container after adding the JSON format to the config file. P.S. Not sure if you already know this or if anyone else pointed it out, but the name SearXNG is using Greek CHI to stand in for the missing three letters.
Fun, behind the scenes trivia. for some reason when i recorded this i kept saying dify as if it rhymes with jiffy or that Dutch children's character Miffy. I had to overdub another track where I just said dify as if it rhymes with defy. Doh!
Thank you for the clearest tutorial I've seen about this tool that's made for people who don't already know everything there's to know about these complex installation processes
Thanks for the Video. Awesome idea by Dify. It would be nice to integrate DSPy (to the workflow). In my experience it is a little hard to work without it. I mean it is hard to make a Robust WorkFlow with only Prompts.
I thought I was doing well until you went from cd dify to cd docker, then you completely lost me as my cmd line doesn't recognise the eza -ahl, any thoughts? Thanks
ty for the vid. as for ollama through docker url... that sends the query to the native allama instance running outside of docker, correct? so essentially, this ollama docker part acts as a proxy? i'm asking, because as you had mentioned previously in a reply to a q of me, ollama doesn't / can't use GPU if run inside the docker, and i want to make sure (have seen so via "activity monitor" on macos") that the GPU / acceleration is utilized even when sending ollama API calls to the docker url...
At first, thank you for amazing content. Can you please create a video on connecting SearXNG with Dify or share any source, It does not accept my local host. Looking forward for more course about Dify
You mentioned that the app is clunky in many places. I know nothing about any of this, but theoretically - couldn't they just have the entire app run directly and fully in docker without having to clone a repo on your pc and then building up a container ? Or just have a proper windows installer? Seems like it's neither fully containerized nor actually running on my pc
@@technovangelist yes, i understand the latter part of the argument, but it did seem more logical to me to have everything be inside docker not to go "here's 9 things you'll need to install and then we'll also get 14 images inside docker"
Yup. Strange to build a tool for non devs and require a git clone to get started. That said the easy route is to setup an account and let them host it. It’s not in their best interest to make it too easy to avoid that.
@@technovangelist are there any tools like that? 🤣 Even in a world of perfectly built software there'll always be room for someone will your calming voice, clear explanations and visual examples. Because writing code is one thing, but writing in English, targeting some who's not ok with killing 4h by going through a poorly written documentation - that's very rare. So next time I'll throw my hands in exasperation because a 70k starred project like yt-dlp is convoluted AF, I'll come and watch you, to calm TF down and stop swearing eternal damnation on people who interpret the universe exclusively through CLIs.
Can I generate an installable app to deploy it in own infra? Which code language use it behind? Can I have access to code generated to custom behaviour?
Great, at least very inspiring. And yeah, overwhelming. Tools like this (also Flowise) need trust. Trust in code, trust in other‘s knowledge. Trust in a never changing rule set. Building on top of on top of on top of on top something is some kind of a stomach aching feeling.
grrr, I fixed it in the description. discord.gg/uS4gJMCRH2 - for some reason the links default to expire after 7 days. this should last longer. thanks for pointing it out.
Thanks for the video. Note though that the Dify docker-compose.yaml does not support host.docker.internal out of the box. I ended up adding: extra_hosts: - "host.docker.internal:host-gateway" to several of the service to get your demo to work. Not sure which got me pass the unable to connect to ollama error, but it wasnt nginx. Also, in the searxng container, settings.yaml needs to be updated to include the json format or that connection too will fail.
@@technovangelist I must say I have the same issue. I gett this message when trying to use host.docker.internal "An error occurred during credentials validation: HTTPConnectionPool(host='host.docker.internal', port=11434): Max retries exceeded with url: /api/chat (Caused by NameResolutionError(": Failed to resolve 'host.docker.internal' ([Errno -2] Name or service not known)"))" Doyou know how to fix it?
Took me 10 hours to install dual linex on my new computer (none of the youtube video methods worked for me!) lol... Do you have (or recommend) a channel for COMPLETE NOOBs - ? I am to a NOOB what a NOOB is to Elon Musks complete programming team... so where do I start? Any advice welcome!
The cleanest is always going to be to code it yourself. There are a number of other tools out there, but a lot of them are messier. flowise is a favorite but its rough. n8n or zapier all have problems. rivet is nice, but they need to handle so many edge cases that its hard to smooth the edges.
I have used Dify for about 4 months but its not that easy it looks , I rather prefer use Flowise AI instead, Dify has much more in built Integrations as tools and easier to create custom tools that flowise AI but Dify has harder learning curve and difficult to create agent flow
Pretty Interesting. But I'd say n8n's new Advanced Ai stuff is also cool and pretty similar, but not that Ai centiric and has more of a zapier like feeling
I'm struggling between Dify and Flowise. They're both brilliant, though Dify seems more refined and pricier. The kicker to me is Dify doesnt allow custom tools for free, since it's completely cloud-based and managed, which makes sense, but is a bummer. update: yeah, the local version of Dify is totally free. With the workflow tooling now, tough call between it and Flowise.
Yeah, I'm gonna give the local one a shot. My app cant be stuck behind a request paywall, so I'll trudge through it before deciding. I will say off the bat, Dify seems more flexible and "visible". The built-in python coding is a nice touch, though technically flowise can port into anything you code as well. Thanks for turning us on to it. I'd say these 2 tools are the best in this class for sure.
@@technovangelist Node-Red is working on an ai workflow-framework, but it's not mature for many folks' needs, and too kitchen-sinky? I'd use it if I was a node-red (or flowfuse, etc.) guy already though. Big learning curve. n8n is essentially "Make" (and not the same thing)? Rivet looks interesting, I'll check it out. Note: with the new/lower openai token costs, a lot of these 3rd party tools need to adjust their pricing.
Too much configuring?!? I would suggest you find another goal if you don’t want to build anything. Or pay them to host it if you don’t want to do the work.
Interesting video. Dify is aimed at no coders, yet you need to have a developers brain to understand it and set it up. Docker does my head in. Too confusing and not user friendly at all.
Slowing down the playback speed a little might help also, I often use this for tutorials I am following along with and maybe using a different device to follow along.
Thank you so much for your feedback on Dify in your video! Truly appreciate your support. By the way, we are working on our documentation and navigation for new users :) So stay tuned!
Please check also your RAG functionality
@@dify_ai can i use dify on pc，but not use docker。
nowdays it's quite hard to find something useful, that is not clickbait or advertisement of service, thank you!!!
This was a brilliant episode, thank you so much! One thing that may not have been obvious (or I missed this in the doco), you have to restart the SearXNG docker container after adding the JSON format to the config file.
P.S. Not sure if you already know this or if anyone else pointed it out, but the name SearXNG is using Greek CHI to stand in for the missing three letters.
I kinda knew it but didn't make the connection. Working on a video now that mentions it and credits you for pointing me to it
Fun, behind the scenes trivia. for some reason when i recorded this i kept saying dify as if it rhymes with jiffy or that Dutch children's character Miffy. I had to overdub another track where I just said dify as if it rhymes with defy. Doh!
I think it _should_ rhyme with jiffy, that's how I hear it in my head
😆
Thank you for the clearest tutorial I've seen about this tool that's made for people who don't already know everything there's to know about these complex installation processes
I'm a huge fan of aider, so I'd like to know if it can similarly generate interconnected project files and modify them collectively during changes
Dify is by far the best tool for AI out there. I absolutely love it
Thanks for this short overview. Interesting !
Any recommendations on how to add a Stripe payment link to a workflow ? This even possible? Love to start testing a chatbot monetization strategy. Thx
Really cool video.
Just out of interest do you know if Llama 3 can be used inside Blender?
i don't know if there is any integration for any ai in blender. though it would be cool
@@technovangelist For diffusion models there is but I have not seen LLMs like Llama3. I know there was a GPT plug once but it didn't work for me.
I Like your Style, straight to the topic, well done 👍
What do you prefer? Dify vs n8n vs Flowise?
Thank you for sharing
Oh that’s easy. N8n. Used it for 5+ years and it’s great.
@@technovangelist Thank you
Thanks for making this. Seems similar to Flowise and a few others but perhaps easier to publish. I'll keep an eye on this.
Thanks for the Video. Awesome idea by Dify. It would be nice to integrate DSPy (to the workflow). In my experience it is a little hard to work without it. I mean it is hard to make a Robust WorkFlow with only Prompts.
Thank you Matt for this great quick walk thru, will try it out
"get some coffee... and then some water..." - yup, same as me! ☺
your tutorial is very clear, thank you Matt :)
In your opinion would Dify be appropriate for creating paid consumer apps? Looks better for intranet or personal use tools.
I think it’s a great prototyping tool because it’s fast to get up and running. But best to build from scratch if going to charge for it.
I thought I was doing well until you went from cd dify to cd docker, then you completely lost me as my cmd line doesn't recognise the eza -ahl, any thoughts? Thanks
Just change directories into dify then docker
ty for the vid.
as for ollama through docker url... that sends the query to the native allama instance running outside of docker, correct?
so essentially, this ollama docker part acts as a proxy?
i'm asking, because as you had mentioned previously in a reply to a q of me, ollama doesn't / can't use GPU if run inside the docker, and i want to make sure (have seen so via "activity monitor" on macos") that the GPU / acceleration is utilized even when sending ollama API calls to the docker url...
ollama won't use the gpu in docker IF and ONLY IF on MacOS with Apple Silicon.
@@technovangelistIs that an ARM thing? Are the new Snapdragon X Elite Windows laptops potentially going to have the same issue?
macOS on apple silicon.
At first, thank you for amazing content. Can you please create a video on connecting SearXNG with Dify or share any source, It does not accept my local host. Looking forward for more course about Dify
You mentioned that the app is clunky in many places. I know nothing about any of this, but theoretically - couldn't they just have the entire app run directly and fully in docker without having to clone a repo on your pc and then building up a container ? Or just have a proper windows installer? Seems like it's neither fully containerized nor actually running on my pc
But it is running on your machine and fully containerized. Having a decent windows installer means extra work and only deals with windows users.
@@technovangelist yes, i understand the latter part of the argument, but it did seem more logical to me to have everything be inside docker not to go "here's 9 things you'll need to install and then we'll also get 14 images inside docker"
Yup. Strange to build a tool for non devs and require a git clone to get started. That said the easy route is to setup an account and let them host it. It’s not in their best interest to make it too easy to avoid that.
If I only made videos about tools that get everything right I wouldn’t have a channel.
@@technovangelist are there any tools like that? 🤣 Even in a world of perfectly built software there'll always be room for someone will your calming voice, clear explanations and visual examples. Because writing code is one thing, but writing in English, targeting some who's not ok with killing 4h by going through a poorly written documentation - that's very rare.
So next time I'll throw my hands in exasperation because a 70k starred project like yt-dlp is convoluted AF, I'll come and watch you, to calm TF down and stop swearing eternal damnation on people who interpret the universe exclusively through CLIs.
WOW this is going to revolutionize my work with market research!
Really neat video, would love to see video on making llama 3 api with say a Mac Studio self hosted api to see what kind of user load it could handle
Hi matt, I love your lecture very much. you are a good teacher.
Can I generate an installable app to deploy it in own infra? Which code language use it behind? Can I have access to code generated to custom behaviour?
For that you would need to build something yourself. This is for web based apps that you want help building
I am from the curacao, can i used all those tools here? Thank you.
Sure. And I have been to curacao. A long time ago on a cruise that went there and Grenada and Martinique and Caracas and more.
Thank you for your time and energy amazing tutorial i wanna see how this helps with building apps and chatbots
My favourite AI uncle
Matt, love the information. Are there ohter apps like DIFY_?
Flowise is the closest.
Great, at least very inspiring. And yeah, overwhelming. Tools like this (also Flowise) need trust. Trust in code, trust in other‘s knowledge. Trust in a never changing rule set. Building on top of on top of on top of on top something is some kind of a stomach aching feeling.
The Technovangelist discord link is dead. Is there an updated one?
grrr, I fixed it in the description. discord.gg/uS4gJMCRH2 - for some reason the links default to expire after 7 days. this should last longer. thanks for pointing it out.
Thank you! Really appreciate your content.
Thanks for the video. Note though that the Dify docker-compose.yaml does not support host.docker.internal out of the box. I ended up adding: extra_hosts:
- "host.docker.internal:host-gateway" to several of the service to get your demo to work. Not sure which got me pass the unable to connect to ollama error, but it wasnt nginx. Also, in the searxng container, settings.yaml needs to be updated to include the json format or that connection too will fail.
Didn’t have to do anything like that. Just worked as is.
And I mentioned the need for the config to allow searxng to output json
@@technovangelist I must say I have the same issue. I gett this message when trying to use host.docker.internal
"An error occurred during credentials validation: HTTPConnectionPool(host='host.docker.internal', port=11434): Max retries exceeded with url: /api/chat (Caused by NameResolutionError(": Failed to resolve 'host.docker.internal' ([Errno -2] Name or service not known)"))"
Doyou know how to fix it?
Loving your content Matt ❤
awesome ... thanks
Took me 10 hours to install dual linex on my new computer (none of the youtube video methods worked for me!) lol... Do you have (or recommend) a channel for COMPLETE NOOBs - ? I am to a NOOB what a NOOB is to Elon Musks complete programming team... so where do I start?
Any advice welcome!
I just started a full ollama course here on this channel
@@technovangelistWhat prerequisite knowledge is needed to benefit from your course?
Nothing really
Thanks Matt, nicely done very clean video for an unfortunately "messy" tool. Is there any challengers on the market ?? Thx again 🙏👍
The cleanest is always going to be to code it yourself. There are a number of other tools out there, but a lot of them are messier. flowise is a favorite but its rough. n8n or zapier all have problems. rivet is nice, but they need to handle so many edge cases that its hard to smooth the edges.
@@technovangelist Thanks Matt!Flowise is indeed a bit more clean and yes code is still king for a while I guess 😁
I have used Dify for about 4 months but its not that easy it looks , I rather prefer use Flowise AI instead,
Dify has much more in built Integrations as tools and easier to create custom tools that flowise AI but Dify has harder learning curve and difficult to create agent flow
Pretty Interesting. But I'd say n8n's new Advanced Ai stuff is also cool and pretty similar, but not that Ai centiric and has more of a zapier like feeling
I'm struggling between Dify and Flowise. They're both brilliant, though Dify seems more refined and pricier. The kicker to me is Dify doesnt allow custom tools for free, since it's completely cloud-based and managed, which makes sense, but is a bummer.
update: yeah, the local version of Dify is totally free. With the workflow tooling now, tough call between it and Flowise.
I didn’t really try the hosted variant. Some folks say dify is the refined one and other think flowise is the refined one.
Yeah, I'm gonna give the local one a shot. My app cant be stuck behind a request paywall, so I'll trudge through it before deciding.
I will say off the bat, Dify seems more flexible and "visible". The built-in python coding is a nice touch, though technically flowise can port into anything you code as well.
Thanks for turning us on to it. I'd say these 2 tools are the best in this class for sure.
There are some others to consider too. Rivet, n8n, and there may even be plugins for node red
@@technovangelist Node-Red is working on an ai workflow-framework, but it's not mature for many folks' needs, and too kitchen-sinky? I'd use it if I was a node-red (or flowfuse, etc.) guy already though. Big learning curve. n8n is essentially "Make" (and not the same thing)? Rivet looks interesting, I'll check it out.
Note: with the new/lower openai token costs, a lot of these 3rd party tools need to adjust their pricing.
Thank you 🎉
Hope you review and compare Dify with Coze
Never heard of that one. Don’t see it on the list of tools that integrate with ollama.
Thanks!
thanks so much for that
Great video
Thank you !!
Too much configuring.. remember, I want to build an ai app. Is there an alternative
Too much configuring?!? I would suggest you find another goal if you don’t want to build anything. Or pay them to host it if you don’t want to do the work.
your videos are TOP
Not yet cause i'm need better knowledge off some point. But it's great. Thanks a lot
2 drinking vessels? Doesn't your coffee have water in it?
It looks very similar to Langflow.
Using this information, while interesting, with all of the caveats mentioned in getting this to work there's nothing easy about it.
this is about as easy as it gets to create new apps that leverage AI.
Interesting video.
Dify is aimed at no coders, yet you need to have a developers brain to understand it and set it up.
Docker does my head in. Too confusing and not user friendly at all.
but it doesn't seem to be free
It 100% free when hosting yourself. That’s what I showed in the video.
This is almost like flowise or I don't understood well this Dify
I tend to focus on apps that have indicated an integration with ollama by submitted to the repo. Flowise hasn't
interesting,
You go too fast! Slow down your instructions.
You do know there is a pause button, right?
Slowing down the playback speed a little might help also, I often use this for tutorials I am following along with and maybe using a different device to follow along.
interesting, it seems node-red + llm