Hi Matt! thanks for checking it out. I'm working on major features like chat history, discovery, and new search types, and will roll them out eventually. It is still very early for Perplexica, but we're slowly reaching there. The searches are slower because Perplexity directly returns the result to the model instead of performing a similarity search. We do the opposite; after the search is complete, the results are refined and then returned to the model, this ensures that the context is very precise and the answers are some time better than Perplexity. We're working on adding some search types that allow you to disable this option.
hi, great start for perplexica. My use case is creating news bullettin with queries like "top news from bbc news website, translated in italian with a title and a summary". Now: sometimes it works, sometimes it does not at all, others the reply is related to spurious parts of the bbc website (how to subscribe etc). Can you have us have a look at what is actually scraped from the websites (and which websites are selected), in order to make better queries to Perplexica ?
Page Assist, an extension for chrome or firefox Ive been using awhile, does searches and can summarize the page. I forgot to add it uses local ollama models :)
To be honest, if the application only offered a light mode, I would be less inclined to use it. I tend to use everything in dark mode. It’s not because I’m trying to reflect a certain persona or mood, but rather, I find that dark mode is easier on the eyes. This preference becomes particularly noticeable when I’m using the computer for extended periods, up to 18 hours at a time. The reduced strain on my eyes makes the experience much more comfortable.
I feel like this is a video I need to make, since this is flat out not true. It can feel like its easier at a point when you should be looking away, but it can cause issues later on, and its especially bad for 30% of the population. It's kind of like when folks say they are visual learners when most of the science based evidence says there is no such thing.
@@technovangelist while it might not be true, I experience the same, I think its all subjective. but never researched it. You made a genius point about probably touching some grass if strain comes into play xD
Make that video. I use dark mode most of the time and I find it depends on the monitor, surroundings and even the time of day. My laptop is awful to use in dark mode and I find it tiring and harder work. My main PC has four monitors, the two best ones being great in dark mode and the two cheapo ones being awful. I keep that whole system in dark mode and use the secondary monitors for things like open directories, Spotify, etc.
Hi Matt. I just stumbled upon your channel. I am really impressed! Thanks for providing actual knowledge instead of using clickbaity video titels without any real value. Keep it going. I hope your channel will be most successful. Quality over quantity.
For me flowise is already running on port 3000 and ollama webui on 8080, I changed all ports in .yaml files to use other ports than these but the webui for perplexica doesn't load, any idea how to fix this
in the docker file the ports statement for the frontend is 3000:3000. So maybe change that line to 3333:3000. That will forward the container port of 3000 to 3333 on your machine. port 3000 on the container will coincide fine with 3000 on any other container or your machine.
@@technovangelist I made it working, got the flowise to start on port 5000 and have perplexica run on its default 3000. will try your option as well but as its working now will not touch anything until something breaks ;-) Thanks for your reply :-)
Hi. Nice content. I like the pace of your videos. Everything is peaceful and interesting without loud music or dramatic sound effects. Please review fabric ai coupled with agents which are included with it.
Yeah, yeah, yeah. Fabric AI. Gotta do that one. Every now and then, I do add a little bit of music or a few sound effects, but I try not to make them too over the top.
This space is changing quickly, so I'm sure there's going to be another tool to try out for doing everything you could ever want. I think right now my favorite for doing a rag search is still OpenWebUI.
All of the dockers are running, but when I go to port 3000 I just have a loading screen that's been going for like 20 minutes. Does anyone know what I'm doing wrong? also Im running it through wsl.
@@technovangelist I installed and tested it. Msty with real time data enabled provided better search results. Perpelxica took nearly 20 gigabytes of my hard disk storage and results were no where as good as Msty with real time data enabled. Uninstalled it!
What do I think? There are so many cool projects. But so little projects that make significant effort to combine all the cool projects. Some kind of standard that makes it easy to combine tools, use them as agents or whatever, that would be great. That's what I think.
Hi Matt, I wanted to ask if you could create a python script that will allow ollama +open webui to be able to search the Internet. I want to know if this idea will work? I'm new and learning this😊😊😊😊
I think it is great, or in fact great potential. Would love to have a more stable retrieval from the web (see comment to the author down below), but in general most programs that work with ollama seems to be very good, so it is hard to settle to one solution (a good thing, in this case)
plus having an offline option is great cuz i got to the perplexity website just now and it says there are too many users asking questions and yeahhhhhhhh great lol
I think there is a space for a robust knowledge curation app, so we can save the valuable results and tag them. Got a whole bunch of ideas around that. You always have fresh stuff, man. Wish I could pick your brain sometime. Planning to do something in all golang.
Lightmode? Pleasing? C’mon, ever heard of circadian rhythms? Going out of whack due to monitor light. Let’s widen our knowledge base into more than just the digital world.
Your comment goes in two directions. First talk about rhythm, accepting that light mode is important, then you make up stuff about going out of whack due to monitors suggesting that dark mode is any better. Choose the argument and stick with it. Perhaps widen your knowledge to facts and not just made up stuff.
Hi Matt! thanks for checking it out. I'm working on major features like chat history, discovery, and new search types, and will roll them out eventually. It is still very early for Perplexica, but we're slowly reaching there. The searches are slower because Perplexity directly returns the result to the model instead of performing a similarity search. We do the opposite; after the search is complete, the results are refined and then returned to the model, this ensures that the context is very precise and the answers are some time better than Perplexity. We're working on adding some search types that allow you to disable this option.
That sounds great! Is it possible to use its services as API as well? Or just through the UI?
yeah, sound amazing! it is a open source project?
hi, great start for perplexica. My use case is creating news bullettin with queries like "top news from bbc news website, translated in italian with a title and a summary". Now: sometimes it works, sometimes it does not at all, others the reply is related to spurious parts of the bbc website (how to subscribe etc).
Can you have us have a look at what is actually scraped from the websites (and which websites are selected), in order to make better queries to Perplexica ?
being able to add docs would be a huuuge plus!
Page Assist, an extension for chrome or firefox Ive been using awhile, does searches and can summarize the page. I forgot to add it uses local ollama models :)
I covered that in an earlier video. Perplexica does it in a very different way and is far more successful.
@@technovangelist I may have missed that vid, and I thought I had seen them all. I am a big fan of ollama. :)
Hello there I am wondering to see a video using ollama+ webui with web search + local llm
To be honest, if the application only offered a light mode, I would be less inclined to use it. I tend to use everything in dark mode. It’s not because I’m trying to reflect a certain persona or mood, but rather, I find that dark mode is easier on the eyes. This preference becomes particularly noticeable when I’m using the computer for extended periods, up to 18 hours at a time. The reduced strain on my eyes makes the experience much more comfortable.
I feel like this is a video I need to make, since this is flat out not true. It can feel like its easier at a point when you should be looking away, but it can cause issues later on, and its especially bad for 30% of the population. It's kind of like when folks say they are visual learners when most of the science based evidence says there is no such thing.
@@technovangelist while it might not be true, I experience the same, I think its all subjective. but never researched it.
You made a genius point about probably touching some grass if strain comes into play xD
Make that video. I use dark mode most of the time and I find it depends on the monitor, surroundings and even the time of day. My laptop is awful to use in dark mode and I find it tiring and harder work. My main PC has four monitors, the two best ones being great in dark mode and the two cheapo ones being awful. I keep that whole system in dark mode and use the secondary monitors for things like open directories, Spotify, etc.
I simply love your reviews. Clearly highlighting pros & cons. This is how it should be done.
Hi Matt. I just stumbled upon your channel. I am really impressed! Thanks for providing actual knowledge instead of using clickbaity video titels without any real value. Keep it going. I hope your channel will be most successful. Quality over quantity.
For me flowise is already running on port 3000 and ollama webui on 8080, I changed all ports in .yaml files to use other ports than these but the webui for perplexica doesn't load, any idea how to fix this
in the docker file the ports statement for the frontend is 3000:3000. So maybe change that line to 3333:3000. That will forward the container port of 3000 to 3333 on your machine. port 3000 on the container will coincide fine with 3000 on any other container or your machine.
@@technovangelist I made it working, got the flowise to start on port 5000 and have perplexica run on its default 3000. will try your option as well but as its working now will not touch anything until something breaks ;-) Thanks for your reply :-)
Hi. Nice content. I like the pace of your videos. Everything is peaceful and interesting without loud music or dramatic sound effects. Please review fabric ai coupled with agents which are included with it.
Yeah, yeah, yeah. Fabric AI. Gotta do that one. Every now and then, I do add a little bit of music or a few sound effects, but I try not to make them too over the top.
Thanks. So what should we use for searching our documents.
This space is changing quickly, so I'm sure there's going to be another tool to try out for doing everything you could ever want. I think right now my favorite for doing a rag search is still OpenWebUI.
All of the dockers are running, but when I go to port 3000 I just have a loading screen that's been going for like 20 minutes. Does anyone know what I'm doing wrong? also Im running it through wsl.
Holy smokes this is cool. Great stuff as always Matt!
You can add Groq API in Msty and that will enable the real-time search mode. How is this better than Msty?
Same reason as wanting to use Ollama... comes down to security and privacy and sometimes cost.
@@technovangelist I installed and tested it. Msty with real time data enabled provided better search results. Perpelxica took nearly 20 gigabytes of my hard disk storage and results were no where as good as Msty with real time data enabled. Uninstalled it!
Really nice! Is there a way to use it as api as well? Or just using the UI?
What is currently the best RAG solution for docs?
That’s like asking whats the best model or whats the best car. There is no best and there can never be. It depends on your requirements.
The end was really appreciated.
What do I think? There are so many cool projects. But so little projects that make significant effort to combine all the cool projects. Some kind of standard that makes it easy to combine tools, use them as agents or whatever, that would be great. That's what I think.
Hi Matt, I wanted to ask if you could create a python script that will allow ollama +open webui to be able to search the Internet. I want to know if this idea will work? I'm new and learning this😊😊😊😊
What a great thing. Temporary embeddings for web research. ChromaDB + python + selenium. Should do it. Thx for inspiring me!
nice
Appreciate your shares and balanced reviews.
They have light/dark modes now (12/2024). I still like dark mode more.
self hosted AI search engine ? didn't know about it. Very interesting, will definitely try
Amazing work!
I think it is great, or in fact great potential. Would love to have a more stable retrieval from the web (see comment to the author down below), but in general most programs that work with ollama seems to be very good, so it is hard to settle to one solution (a good thing, in this case)
Another awesome video! Can't wait to try this out.
Which branch are you using?
The main one
plus having an offline option is great cuz i got to the perplexity website just now and it says there are too many users asking questions and yeahhhhhhhh great lol
You da man! That's exactly what I wanted to know!
I think there is a space for a robust knowledge curation app, so we can save the valuable results and tag them. Got a whole bunch of ideas around that. You always have fresh stuff, man. Wish I could pick your brain sometime. Planning to do something in all golang.
Ping me on the discord or if you are near Seattle we can meetup.
Cool thanks @technovangelist. On what server? A little new to discord. Can’t find a way to search. I’ve only joined from invites so far
Discord.gg/ollama. I’m mattw tvl.st/y
The profile shots are a bit weird. Not sure you should add them.
The awkward silence at the end is hilarious though.
I love it when I find something like this that im like, ooh I like that. Lets give it a try. And its broken....seigh.
Searxng = Searching. Clever, I know.
Msty, make simple app to search, use api, and rag
All this AI technology is still pretty green ...
Lightmode? Pleasing? C’mon, ever heard of circadian rhythms? Going out of whack due to monitor light. Let’s widen our knowledge base into more than just the digital world.
Your comment goes in two directions. First talk about rhythm, accepting that light mode is important, then you make up stuff about going out of whack due to monitors suggesting that dark mode is any better. Choose the argument and stick with it. Perhaps widen your knowledge to facts and not just made up stuff.
⭐️⭐️⭐️⭐️⭐️