Video

Thanks ... I appreciate you POV... I have used CrewAI, AutoGen and more.. I have wasted tons of time learning stuff that I don't use because at some point (usually 3/4 of the way in) I realize the limitations. I mostly just learn why it is not for me. Not sure if I can put a finger exactly on what it is, but it seems to me like most stuff is built these days with training wheels.

Would have been good to cover what ollama is and what problems does it solve, for about 30sec-1min, before going into hardwarre requirements etc.

Interesting Matt, but not enough to draw me in yet, also no windows friendly. Thanks !

Thank you

Great points about Kubernetes being overkill and just the over focus on Python etc. I am building an open source tool inspired by Laravel PHP just to help me see how it is not the language but so much more. And good points about automation, to me I want normal non coding users to see the power and ease of this as they see the power of prompts. Anyways great one!

Not sure if anyone told you, but these hours long streams will be probably skipped by the vast majority of your audience, just because most of the ppl with brains in their heads don't have that much free time to watch all that :)

I love this. Exactly. Spot on. This is why I'm building my new project from scratch.

curious. maybe a bench marking of war and peace using the the various data bases based on size and or based on speed

Hey, I have an 'ask' alias too. Pretty cool! I like piping the output to 'say'.

fascinating. thank you for these videos!

Now, it's 4'16 in the video, but I still can't get any practical tips. I must go now to cut the loss...

Thank you so much for this video. Super helpful and well-presented. Your time and effort is most appreciated!

You saved my day. Thank you Matt.

If the cost of abstraction is always complexity. All these frameworks are transitional. As the new layer becomes standardised, the boilerplate tends to disappear, but so does transparency.

I have watched a few of your videos. I love the way you explain things simply and clearly. Keep it up. Thank you for your work to all of us.

with these frameworks it’s the usual ‘sell shovels during a gold rush’ attitude and I’m getting sick of it tbh. especially with those langchain cowboys on yt…

It would be interesting to see what AI and machine learning could come up with as far as Neuro-training and neuro-therapy protocols.

lol... casual day at the household with regular conversations.

Hi Matt, thanks for this video. But what would you recommend for the following? I want to develop software with agents. The target is not a specific workflow but a actual software that they need to create.

is it possible for long range task with non deterministic input and output

thank you i didn't know iTerm2 exists with ollama copilot plugin we can also chat with vault... i put johnntdt contact info on one of machine learning notes... then ask llama3 who is jhonny, and llama3 give me johnydt it said it's misplaced in out of context note.

Shell-GPT can integrate AI in any terminal. Local models are supported.

No body knows anymore

Darn I missed the live! I'll catch the replay.

Great video. I see now how to use Ollama to get back a specific scheme form the LLM when requesting it. However, how do I provide Ollama with a list of schemes with descriptions of their use, and have the LLM response give a normal message, with a potentially empty list of schemes with their tool names for my code to handle for the LLM to then convert into natural language response? The killer feature of "function calling" is to allow the LLM to assert what functions (if any) need calling. Proper scheme is just icing on the cake.

I am just sick of all the AI applications explicitly using CUDA drivers. Anyone with AMD and Intel GPUs are out of luck.

Very well illustrated! Thanks Matt

I am just getting started on learning about agents. So glad to find this video and hear your perspective.

Hey Matt - love your videos. Not really too relavant to Ollama specifically but can you make a video on model types? Im having a hard time understanding what fill-mask models are. Thanks!

what is kubernetes? me: my nightmare

What a great thing. Temporary embeddings for web research. ChromaDB + python + selenium. Should do it. Thx for inspiring me!

Really? Multiple models? I know a way to do this on different ports. But on the same port? How?

Funny, today I criticized on a channel that I think that everything is blown up for all of the use cases I’ve seen so far. And I do agree on your opinion! I use python and even for complex or more open actions, I avoid libs like CrewAI or LangChain.

The model effectively decides when to call the function, and which function is appropriate at the time by formatting the output for the particular function. If I give it a query where taking a headless web screenshot is appropriate, then the output is formatted for that particular function. And the output does NOT particularly need to be in json either, depending on how the function is described to the openai API and how you're choosing to detect that the model "wants" to call a particular function. Respectfully, I think "function calling" is pretty appropriate, and it really seems like there's a misunderstanding of what function calling does here. Not trying to be negative, just pointing out that "just formatting" is missing the point. Of course the model doesn't have access to actually execute code on your local machine. But the "formatting" decided upon (by the model deciding which function to use) is essentially doing the exact same thing. Based on scrolling through the comments here, disappointingly I don't expect you to be very receptive to this feedback. Just my $0.02, and I stand by it. Function calling has a lot of really great uses, and I think "just formatting," while technically true, sorely misses the point. You can (and I do, not just guessing here,) effectively give the LLM a set of tools, trigger those tools based on output from the LLM, and provide the output from those tools back to the LLM if needed. That's pretty impressive I think. All that being said, I really do enjoy your videos. Thank you for sharing, and I look forward to the next one.

While I agree with the sentiment that simpler is better. Sometimes you don't know that you could benefit from the ability to let the LLM talk back and forth between other LLMs until you've tried it on several workflows. If your plan is super well defined then it makes sense to simply the design, but for now I'll continue using crewai to scaffold out AI based workflows till I can determine the true limits of the models that are available. Good video though! Thank you for putting it together. It made me think!

I remember that Charles Petzold book . . . learned a lot from that. He's a cool guy. You should check out his book on Turing or his book on Code. Neat guy.

NLU and LLM the way to go, NLU still king as is assured "CERTAINTY", intersplice with tactical LLM agents, LLM with Rag Educate an Agent GUI. This is the model that is forging ahead. If you want a real service there's about another 50 things you need to consider, and most tools, especially codey tools cover <10%. Multi-Customer/Sass, Analytics/Timezones, Chat Client, Skinnable, Configurable, Plug and Go, Voice/Speech/Multi-Modal, AI Agents (LLMS), NLU /LLM Hybrids. Escalations and Triggers. Communication API's, IPAAS back office Services. Multi-Messaging Agents Groups / Messaging Cockpit, or Call centre. Multi- CC Connectors, Pro-Active Tuning, Auto LLM Tuning, and Live Spech Rec tuning. and more.

I think for external you go to settings-> API Keys -> Add New Key -> Open AI Compatible, but it didn't work for me 😕

function __fish_complete_ollama_list ollama list 2>/dev/null | tail -n +2 | string replace --regex "\s.*" "" end complete -c ollama -f -a "(__fish_complete_ollama_list)" --condition '__fish_seen_subcommand_from run'

Thanks Matt for the video! great introduction to Msty

Very good

You mentioned Charles Petzold, when i first read it i was 18. A lot of time has passed.

i did not see any manager agent in this video. what did i miss?

Subscribed and linked. Definitely an underrated video. Yet to check out your other videos, and will check those out too. You mentioned that you were using a m1 based laptop (assuming, due to the reference to battery life). Not sure if you sped up the inference part of the video, but if that is the native inference speed, it really is great. Is there any video where you've covered the specs of this laptop ?

having a RAG will be cool

VID SUGGESTION ~ (Resolve Error Response: Invalid Volume Specification) - Thanks test@xz97:~$ docker run -d --gpus=all -v /home/test/models/:root/.ollama -p 11434:11434 --name ollama ollama/ollama docker: Error response from daemon: invalid volume specification: '/home/test/_models/:root/.ollama': invalid mount config for type "bind": invalid mount path: 'root/.ollama' mount path must be absolute.

old skool....loads of good frameworks make embedding VERY VERY simple

You can close a Dialog with the ESC key. I had the exact same response.... especially because the window border is so faint. Took me a while to realize was in a modal.

very good !!! qell understood.... (quick advice)... Tempreture is related to the training also (as things which were not trained deeply will need higher tempreture ... and things deeply embedded will be ok with lowest tempreture: how do people train thier odels and what are thier acceptable levels? as some are .=0.5 and under whilst other dont care and let the model complete an epoch on large dataset and assume the data took .. as long as thier final output was preferable : when in fact all the data which did not go in at the loss below .0.5 did not take and is not retrivable perhaps its there ephemeallly ... as it is like a pretraining ... its just used for next word prediction... but we are doing tasks ! which is whole sequece prediction/recall so when we train for a task we expect the whole of the data set to be fit in range .... so low temptrture 1 should be acceptable losses ... Some say tha this effect the soft max of possiblisty chosen byu the topk sample as well as the topP percentage of cutt of... but this is when there are many sample chosen... but this also depicts the values that were trained at thatr rate of loss .... so it will be collecting sample from the level under the temptrture rate of 1 ( a lot ) so this will need constraining with topP (selecting the highest of probablitys ... but the softmax will also spread them alowing for more random also, when the model has been over trained.) .. so an over trained model can be loosened by raising the temptretue and a wild model tamed ! lol...

i hope add option for change folder or c: to d: for download models or install olama