Private Obsidian AI: Add DeepSeek to your Obsidian with Ollama (GUIDE + SETUP)

so if we want to privatize your copilot experience we need actually to install two models now let's change to the QA all right um the same question I will

like to ask now you will have a totally private obsidian compilot in your computer hi friends this is Tony in my last two videos I've demonstrated the incredible

potential of combining of CDM with powerful AI tools you can ask AI to quickly locate what you need from thousands of your past notes you can collaborate with AI to build personalized learning maps and knowledge

graphs tailored to your interests but one important question keeps popping up in the comments is there a local and a private solution I completely understand

this concern after all one of the main reason I choose obsidian over notion is because it offers complete local storage allowing you to keep your nodes 100%

private in this video I will show you exactly how to install your own private local AI cilet directly into obsidian okay this video will be split

into three parts the first part is how to set up obsidian compiler and uh what are those key features the second part is to show you how to install your private AI model deep seek with AMA

lastly you'll be able to integrate your private AI with obsidian copilot to have a true local and private AI in your obsidium okay this is a typical obsidium

interface I have some example notes on the left so uh to install your obsidian compilot you will need to go to the

setting and find your community plugin and search for obsidian copilot this is the co-pilot plugin you will need to download it it was developed by

Logan okay all right after you install it enable your plugin so on the right hand side you find that the API key is not ready

yet if you happen to have an open AI or anthropic API key you can definitely set up here so this is really simple if you

have an open AI API key okay I've already copied my key you need to do is to verify it everything is verified and set up now

you can simply use it on the left hand side you will find your cilet and open that and you can

start chatting say hi there and Tony hello Tony okay the chat mode works and for the vault QA mode uh this is the

mode that you will be able to query your own nerds and uh ask your AI to help you find your nerds okay your copilot already find two cases

and it also showed you the relevant resources AI Implement cases customer agents customer experience support and let's check it

out this is this case was confirmed click therapeutic on clinical use Gemini from Google workspace to

trans to transform complex operation data into actionable Insight good one uh beside ask any questions about

your notes you could also find that if you click the relevant notes let me drop it if you find a note say like um case

report okay the relevant note function showed you the most relevant notes compare with what you write in the drafter note is uh something from is

something from the Enterprise case report uh if you check closely you'll find that this is the part I copied this is also a very important

function I found useful because when you are writing your report or notes or draft the obsidian compilot behind it will actively help you find the most

relevant notes that you may find interesting even though you did not type any links here okay now we find three key obsidian compiler features the first one is chat

mode you can chat with your AI with any topics the second one is Vault QA mode that you can ask your AI to find a relevant notes and ask relevant questions about your notes the third one

is relevant notes when you are reading and writing a notes ciled behind your OB CDM will actively find relevant notes for you

so before we jump into uh how to install your private AI thing I would like to explain a little bit more about how this is possible this will definitely help you understand how can we totally

privatize your AI for your obsidian so behind the Cella actually there are two Motors that is working for you the first one of course is GPT 40 the open air

motor the second one which is also key one is the embedding motor what is embeding embedding is simple it use neuron networks to

translate your neres into numbers those are vector numbers that your machine will be able to calculate the relevance between your thousands of

ndes so when you were creating your new notes the machine behind it will actually use those Vector numbers to calculate similarities between your new notes and your old noes if they find a

similar if they find similar it is high it will pop up if you ask any questions about your notes the machines will calculate and find the most similar and

relevant notes and send it to your AI and answer your question so embedding is the key so if we want to privatize the

copilot experience we need actually to install two models to help you totally privatize it how let's come to the second part

okay let's come to the second part install deep seek AI model with AMA okay what is ama ama is a free software that can help you run large language model in

your computer no matter it is Linux Windows or Mac OS you can simply download and install it then what is dpse currently dsek R1

is the best open source motor you can have on the market right now you can simply download it from llama and have it install to your computer the only one

thing you need to keep in mind is your computer memory if you have a relatively small Ram computer for example 16 GB you

are suggest to install Motors like uh 7B 8B and 14b okay once you have your AMA download what you do is very simple you

copy this sentence here and have it run on on your terminal uh you can directly uh interact with your private model say

how are you today deeps R1 is a sying motor what you will have is a sying process and its final response back to AMA you can also find your ined motor

onama what I recommend is this one currently BGE M3 is also an open source and free motor that even surpass the

performance of O AI of open AI eding motor so BG M3 is here and it is small so it's rly only 1 gig um you copy it

and uh pull your motor here now we have a trap motor ready and embeding motor ready but there's one more thing you need to

do if you are familiar with larger language Motel you know that each motor has its context window um the context window is the amount of tax you can ask

the Moto to help you process for for each Moto download from AMA the default context window is limited by AMA to ensure your Moto's speed since we need

this motor to answer questions about our notes we better to maximize the context window of your local AI otherwise you might find it to be frequently

hallucinating from your notes that's something we don't want to see what you need to do is to copy this command at the terminal here set the

parameter to 128k and the parameter is set then you type

save you can Simply Save the same name as the old Motel all right new model created now

you can check the parameters of your new model show parameters all right your context window is saved as

128k last step link AMA to your obsidian what you need to do is to ensure your AMA is running at the back end and paste the

command and R it what does this mean this means to link your AMA address with obsidian application the application will be able to use AMA model

directly now quick your AMA and type AMA serve your obsidian address is prepared here right let's

quit the terminal now we open AMA again and everything is ready to use okay now we have everything ready

let's integrate your private AI motor with obsidian Copilot I've already removed open thei there's no motor in mycd pallette

what you need to do is to in your motor setting to add your motor here add custom mode uh if you don't remember the exact motor name you should look in your

uh terminal cuz the name should be exactly the same as the AMA motor then pick your provider AMA then verify it

cool everything is ready verification successful since um the murdery is a reasoning motor you can pick it here it's doesn't matter all right uh

remember that we also need an embedding motor you should add your embedding model below here uh don't add here this is chart mode add your embedding model

here at your custom embedding model remember uh you will need to copy the exact the same name of the motor

inama pick AMA and verify it great job let's add the motor once you have your local motor

created you will select it here deep seek R1 and uh yes AMA in Med model everything is ready now let's try

it out okay hi there um how are you view oh cool the thinking motor will think a lot even the simplest

question but the speed is good have any question about sub cidian great job now we can check the vot QA as

well oh before Oh before we use the vote mode uh this's one thing I need to remind you um if we set a new eding motor it will ask you to Reb the index

what you need to do is to click continue now let's change to the QA all right on the same question I will like to

ask great seems like the motor has already find it WS let's search it to see if the motor

is right or not yes we have war war use Gemini to enhance employment confidence in their communication for customer support

effective customer interaction good job now the last function you will find a relevant note here now you will have a totally private obsidian copilot in your

computer and no one can look into your data if you you like my content please hit subscribe and give me a thumbs up now let's wrap it

up so compare with the last two videos um there are two things that currently uh local compil can do which is agentic file management and deep research the

strategy I suggest is that you can have a totally private vot with private Ai and ask questions about your own writings and notes um for information

from the website you are welcome to use those AI providers from open AI Google and the search online um since this is information from public you don't really

need to worry about your privacy for information private you can have a private AI to process with it okay that's it for today if you have any ideas you would like me to try please

leave your comments below and thank you very much and happy hacking bye-bye