Great tutorial,just to add,problem with smaller models like Mistral 7b,is the lack of knowledge in various topics.In general Mistral 7b is amazing conversational model but if you start asking questions about presidents,egyptian history or specific car it would just halucinate and tell something that is funny and not true. With anything LLM you can build massive databases on every topic you like,anatomy,behavior,social networking,just feed the model with wiki-s or scientific documents and it would become a monster that is capable to do anything. I personally build a database for human anatomy,i feed it with medical documents and web books,and now its capable to elaborate complicated studies and giving answers about anything you can imagine,any organ,surgery procedures or even forensic science. You can feed it with books on human psychology and it would become your best friend free therapist,with knowledge you can only dream about. I tested this software and it's just a dream,forget about chat GPT and just use this to build Database with topics you are interested. Anything LLM supports PDF,text files,web documents,HTML,web books formats and even audio files which is super crazy. You can just scrape entire web sites and built into your database,and which each document model is becoming more powerfull.
I was expecting more, I just downloaded the Anything LLM and it can connect to the outside world using like a Google API / would that help with RAG to get more accurate info ? without attaching anything ?
He shows the interface at ~ 15:00 but it's just a input area to drag or click to upload documents underneath the files he shows. It says "Click to upload or drag and drop" or you can load a website URL in the same interface. It says "Fetch Website" It's pretty straightforward but he doesn't show every step, yes.
@@GovBotics Tried that but could never get them to talk to each other. Regardless of what I did everything llm could not connect to the ollama instance
Great tutorial,just to add,problem with smaller models like Mistral 7b,is the lack of knowledge in various topics.In general Mistral 7b is amazing conversational model but if you start asking questions about presidents,egyptian history or specific car it would just halucinate and tell something that is funny and not true.
With anything LLM you can build massive databases on every topic you like,anatomy,behavior,social networking,just feed the model with wiki-s or scientific documents and it would become a monster that is capable to do anything.
I personally build a database for human anatomy,i feed it with medical documents and web books,and now its capable to elaborate complicated studies and giving answers about anything you can imagine,any organ,surgery procedures or even forensic science.
You can feed it with books on human psychology and it would become your best friend free therapist,with knowledge you can only dream about.
I tested this software and it's just a dream,forget about chat GPT and just use this to build Database with topics you are interested.
Anything LLM supports PDF,text files,web documents,HTML,web books formats and even audio files which is super crazy.
You can just scrape entire web sites and built into your database,and which each document model is becoming more powerfull.
Awesome！ Thanks for the sharing
RAG: retrieval augmented generation
Thank you.
Thank you for the great tutorial!
I was expecting more, I just downloaded the Anything LLM and it can connect to the outside world using like a Google API / would that help with RAG to get more accurate info ? without attaching anything ?
Is there an API on the AnythingLLM side? I'd like to give it a folder full of PDFs and let it run overnight making summaries.
Yep, there is... It's a good combo with n8n automation tool...
Yea, I'm interested in giving a CrewAI Agent this as a tool for local document store knowlege. So I'm asking the same question about API access.
@@hiddenkirby
Does not privateGPT work for that goal?
thanks but you dont really show how to load document, do you? I found it but What is the embedding settings. All default or ollama embedding..
He shows the interface at ~ 15:00 but it's just a input area to drag or click to upload documents underneath the files he shows. It says "Click to upload or drag and drop" or you can load a website URL in the same interface. It says "Fetch Website" It's pretty straightforward but he doesn't show every step, yes.
Is this something I can run on my server without using a desktop app?
yes but you run ollama in the server mode.
@@GovBotics Tried that but could never get them to talk to each other. Regardless of what I did everything llm could not connect to the ollama instance
@@GovBotics
You type very fast on your keyboard ⌨️.
Your keyboard I probably from 1996
Just a mechanical keyboard, which are still used a lot.
Impossible to read a single word in smartphone