If you aren't already, will you please look into integrating knowledge graphs into the app, along the same lines as the graphRAG project? Thanks for everything you do!
Actually have a development branch from long ago with KGs in them. I didn't find the performance much more remarkable, but the issue is never made it to prod was some OSS llms were performing horribly trying to make node/relationships with the graph dB I was using
Hi @TimCarambat, I just installed it in my Mac M2, and when I try to load a CSV file to my workspace, it keeps "loading" until it stops and presents an error msg, but the files is very small (less than 0.5mb). Is there any trick that I'm missing? thank you so much for your work!!
Sounds like the native embedder is not loading. It is downloaded on first embed of a file. You can also load the program in debug mode to get full error logs which help a ton docs.useanything.com/debug
Just running the 8B because it works OK for me and I get faster inferencing as opposed to the 30B. Also, the size on disk is smaller - was just a preference for the video, no real particular reason one way or another
"stop paying for gpt" - shows local llms such as mistral and gemma as alternatives, ok buddy. Have you every tried them? dont get me wrong they are ok, and some at around 24gb of vram are better than ok, but not comparable to gpt, wtf? youtube clicks much?
Just can't get the https given by ngrok on my windows machine to load the models on my macbook pro. I keep getting a 403 forbidden showing up in the ngrok http requests list. I do want to say that other than the remote connection attempt, I followed everything else and it does run the models on my windows machine, for the most part, just fine. the q8 version seemed to slow everything down. I am a newbie at this so don't slam me. Lol. I just think there are a few steps of how to get it all going on the windows machine without running it there, and just serve it to run remotely.
Are you sure it responded immediately, because you failed to hide the status bar below that shows time. you don't neet to lie just tell people it's superslow on average machines
I have a 4090 in the computer and get a time to first token of around 300ms on average with 30 tokens/sec. Yeah, it'll run slower on weaker machines. That's pretty common knowledge and why would an app layer interacting with an LLM provider like lmstudio improve that performance?
@@TimCarambat It wouldn't; that's the issue. "It just kinda pops up instantly!" Apparently, five minutes passed in that instant, and when asked for clarification, you instead deflect. Obviously (to us), local models take that kind of time; the issue is the intentional misinformation toward people who don't yet know that. It gets your software on more machines I suppose?
Hi! Nice video! can you share the specs of the windows computer? I would like to know if mine needs an upgrade. Thanks! (ok, just got that you have a 4090. RAM and CPU is important?)
Thank you very much. As soon as I saw that RAG was built in and it was simple to use, I immediately started finding readme pdfs on various topics to ensure I could use this tool as efficiently as possible. After my targeted pdfs are found, I plan on grabbing data from how-to and wiki.
I have a little problem, I use AnythingLLM with the TextGenerationWebUI as a backend and when I try to use @agent feature, I always get the error msg "Could not respond to message. OpenAI API key must be provided to use agents." But I want it to use my local llm, isn't that possible?
🎯 Key points for quick navigation: 00:13 *🤔 The speaker asks if you can tell him exactly what data is in your vector database, and guarantees that you can't.* 00:39 *🔓 VectorAdmin is a universal GUI that allows you to manage your vector data like any other database.* 01:05 *💼 VectorAdmin focuses on application, while other companies focus on infrastructure.* 02:11 *📁 VectorAdmin can connect to multiple types of databases, including Pinecone and Chroma.* 04:42 *🔑 When you first log in to VectorAdmin, you'll see a page with a red connection indicator because it hasn't been filled out yet.* 06:03 *⏱️ The sync Pinecone data button indicates that there's data in Pinecone that VectorAdmin doesn't know about.* 07:12 *📁 You can add documents directly into your Pinecone instance using an embedding service like OpenAI.* 09:04 *🔍 You can see what's in the vector database, including text chunks and embedded documents.* 10:02 *💪 You can edit or delete vectors atomically without affecting other workspaces or namespaces.* Made with HARPA AI
So I downloaded Ollama on my Mac and all, but when I got everything LLM, it promoted me to download Ollama too so it runs on it rather than having to run both at the same time. What's the difference?
I couldn’t get the web browsing feature to work. I put in my api key and search engine ID but still nothing. I’m attempting to do this with llama 3 dolphin, has anyone been successful getting this feature to work on the dolphin model?
Bro this has to be the most comprehensive, simple, engaging and all around entertaining video on AI I've ever watched. Your presentation, explanations, and exert level knowledge base are all 'S' tier! Bra-freakin'-vo! Subscriber well earned and deserved! 🏆👏🏽👏🏽
I was diving deep into CrewAI the other day because I would love to have CrewAI users be able to "port" their work into AnythingLLM but found CrewAI doesn't have like a "server" or REST API, its like just a library like LangChain is - so scripts and static code :/ CrewAI+ (their hosted and paid model) is the only thing we could possibly integrate with - which is not ideal. AutoGenStudio is probably the only other AI Agent framework/tool we could integrate with. SuperAgent as well,
I love this tool, I already made several Workspaces, each with its own LLM and RAG. This video was good how with an explanation. I am a Python developer and I would like to create my own agents
@TimCarambat I have stallednin my progress when trying to run ollama It sits for about 5 minutes and then the powershell tells me that it " timed out waiting for llama runner to start - progress 1.00" Suggestions?
Great video abd software thanks. I am unable to get agent commands to do anything. They just hang indefinitely with no indication that it's doing anything. The non agent commands seem to work fine but as soon as i try to use the @ agent command, it does nothing. I am pretty sure I have everything configured correctly. Ideas?