Create a LOCAL Python AI Chatbot In Minutes Using Ollama

[@patrickmateus-iq8bi]

THE 🐐 I became the Python developer I am today because of this channel. From learning Python for my AS level exams in 2020, to an experienced backend developer. From the bottom of my heart, Thank You Tim. I'm watching this video because I have entered a Hackathon that requires something similar. This channel has never failed me.

[@umeshlab987]

Whenever I get a idea this guy makes a video about it

[@ChristianRayNewMoney]

Me too 😂

[@Linuxovert]

You are BRILLIANT @umeshlab987

[@healthadvice3091]

We are one

[@wizurai_engineering]

That's right!

[@omverma5367]

*an idea

[@YangEdward-ex1pw]

already used it to set up a Q&A system to answer customers' common questions. Thank you so much for the sharing and demo.

[@krisztiankoblos1948]

The context will fill up the context windows very fast. You can store the conversations embedings with the messages in a vector database and pull the related parts from it.

[@Larimuss]

Yes but that's a bit beyond this video. But I guess he should quickly mention there is a memory limit. But storing in vector is a whole other beast I'm looking to get into next with langxhain 😂

[@krisztiankoblos1948]

@@Larimuss It is not that hard. I coded it locally and store them in json file. You just store the embeding with the messages then you create the new message embedings and with cosine distance you grab the most matching 10-20 messages. It is less then 100 lines. this is the distance fucntion: np.dot(v1, v2)/(norm(v1)*norm(v2)) . I also summerize the memories with llm too so I can get a shorter length.

[@landinolandese8298]

@@krisztiankoblos1948 This would be awesome to learn how to implement. Do you have any recommendations on tutorials for this?

[@czombiee]

​​​@@krisztiankoblos1948Hi! Do u have a repo to share? Sounds interesting!

[@star_admin_5748]

@@krisztiankoblos1948 Brother, you are beautiful.

[@modoulaminceesay9211]

Thanks for saving the day. i been following your channel for four years now

[@WhyHighC]

New to the world of coding. Teaching myself through YT for now and this guy is clearly S Tier. I like him and Programming with Moshs' tutorials . Any other recommendations? I'd prefer more vids like this with actual walkthroughs on my feed.

[@S1mpai_Official]

idk but I never understood anything from programming with Mosh videos. Tim is a way better explainer for me, especially that 9 hour from beginner to advanced video.

[@M.V.CHOWDARI]

Bro code is GOAT 🐐

[@WhyHighC]

@@M.V.CHOWDARI Appreciate it!

[@joohuynbae5084]

for some Window users, if all of the commands don't work for you, try source name/Scripts/activate to activate the venv.

[@JordanCassady]

The captions with keywords are like built-in notes, thanks for doing that

[@yuvrajkukreja9727]

Awesome that was "the tutorial of the month" from you tim !!! because you didn't use some sponsored tech stack ! they usually are terrable !

[@T3ddyPro]

Thanks to your tutorial I recreated Jarvis with a custom GUI and using llama3 model to make Jarvis, i use it in italian cuz i'm italian, but you can also use it in english and other languages.

[@akhilpadmanaban3242]

these models beiung completely free?

[@leodark_animations2084]

​@@akhilpadmanaban3242with llama yes as they run locally and well you are not using apis. But they are pretty resource consuming..tried it and they couldn't run

[@T3ddyPro]

@@akhilpadmanaban3242 Yes

[@Larimuss]

Wow thanks! This is really simple, straightforward guide to start me getting into writting the python rather than just using peoples UI. Love the explanations.

[@leonschaefer4832]

This just inspired me saving GPT Costs for our SaaS Product, Thanks Tim!

[@CashLoaf]

hey i'm into saas too did u make any project yet?

[@specialize.5522]

Very much enjoyed your instruction style - subscribed!

[@franxtheman]

Do you have a video on fine-tuning or prompt engineering? I don't want it to be nameless please.😅

[@kfleming78]

Fantastic explanation - thank you for this

[@KumR]

Hi Tim - Now we can download Llama3.1 too... By the way can u also convert this to UI using streamlit

[@ShahZ]

Thanks Tim, ran into buncha errors when running the sciprt. Guess who came to my rescue, chatGPT :)

[@repairstudio4940]

Awesomesauce! Tim make more vids covering LangChain projects please and maybe an in depth tutorial! ❤🎉

[@konradriedel4853]

Hey man thanks a Lot, could you explain how to implement own Data, PDF, web sources etc. for giving answers when I need to give it some more detailed knowledge about certain internal information about possible questions regarding my use Case?

[@carsongutierrez7072]

This is what I need right now!!! Thank you CS online mentor!

[@arxs_05]

Wow, so cool ! You really nailed the tutorial🎉

[@jagaya3662]

Thanks, super useful and simple! I just wondered with the new Llama model coming out, how I could best use it - so perfect timing xD Would have added that Llama is made by Meta - so despite being free, it's compareable to the latest OpenAI models.

[@timstevens3361]

very helpful video Tim !

[@techknightdanny6094]

Timmy! Great explanation, concise and to the point. Keep 'em coming boss =).

[@build.aiagents]

lol thumbnail had me thinking there was gonna be a custom UI with the script

[@rhmagalhaes]

I love how you make it easy for us. After that we need an UI and bingo. Btw, does it keep the answers in memory after we exit? Don't think so, right?

[@josho225]

based on the code, no. only a single runtime

[@bause6182]

If you combine this with a webview you can make a sorta of artifact in your local app

[@MwapeMwelwa-wn9ed]

Tech With Tim is my favorite.

[@WhyHighC]

Can I ask who is in 2nd and 3rd?

[@tech_with_unknown]

@@WhyHighC 1: tim 2: tim 3: tim

[@davidtindell950]

Adding a context, of course, generates interesting results: context": "Hot and Humid Summer" --> chain invoke result = To be honest, I'm struggling to cope with this hot and humid summer. The heat and humidity have been really draining me lately. It feels like every time I step outside, I'm instantly soaked in sweat. I just wish it would cool down a bit! How about you? ...🥵

[@asharathod9765]

Awesome.....i really needed a replica of chatbot for a project and this worked perfectly....thank you

[@dimox115x9]

Thank you very much for the video, i'm gonna try that :)

[@proflead]

Simple and useful! Great content! :)

[@TechyTochi]

This is Very useful content Keep it up

[@RevanthK-y1l]

Could you please tell us about how to create a fine tunning chatbot using our own dataset.

[@SAK_The_Coder]

This is what i need thank you bro ❤

[@toddgattfry5405]

Cool!! Could I get this to summarize my e-library?

[@weiguangli593]

Great video, thank you very much!

[@taymalsous5894]

hello tim! this video is awesome, but the only problem i have is that the ollama chatbot is responding very slowly, do you have any idea on how to fix this?

[@31-jp6ok]

If you read my message, thank you for teaching and would you mind teaching me more about fine-tune? What should I do? (I want Tensorflow) and I want it to be able to learn what I can't answer by myself. What should I do?

[@pixelmz]

Hey there, is your VSCode theme public? It's really nice, would love to have it to customize

[@Money4Jam2011]

Great video learned a lot. Can you advise me the route I would take if I wanted to build a chatbot around a specific niche like comedy. build an app that I could sell or give away for free. I would need to train the model on the specific niche and that niche only. Then host it on a server I would think. An outline on these steps would be much appreciated.

[@H4R4K1R1x]

This is swag, how can we create a custom personality for the llama3 model?

[@praveertiwari3545]

Hi Tim, I recently completed your video on django-react project , but i need an urgent help from your side if you can make a video on how to deploy django-react project on vercel,render etc. or other known platform (this would really be helpful as there are many users on the django forum still confused on deployment of django-react project to some popular web deployment sites. Kindly help into this.

[@sacv2]

This is great! thanks

[@skadi3399]

Great video! Is there any way to connect a personal database to this model? (So that the chat can answer questions based on the information in the database). I have a database in Postgre, already used RAG on it, but I have no idea how to connect the db and the chat. Any ideas?

[@arunbalakrishnan8978]

Useful. keep doing

[@bsick6856]

Thank you so much!!

[@kinuthiastevie4031]

Nice one

[@AlexTheChaosFox1996]

Will this run on an android tablet?

[@sean_vikoren]

thank you.

[@davidtindell950]

You may find it 'amusing' or 'interesting' that when I (nihilistically) prompted with "Hello Cruel World!', 'llama3.1:8b' responded: " A nod to the Smiths' classic song, 'How Soon is Now?' (also known as 'Hello, Hello, How are You?') " !?!?!🤣

[@siddhubhai2508]

Please Tim help me how to add long term (infact ultra long) memory to my cool AI agent using only ollama and rich library. May be memgpt will be nice approach. Please help me!

[@birdbeakbeardneck3617]

not na ai expert so i could said something wrong: you mean the ai remeber things from messagrd way back in the conversation? if so thats called context of the ai, and is limited by the training and is also an area of current developpement, on the other hand tim is just making an intrface for already trained ai

[@siddhubhai2508]

@@birdbeakbeardneck3617 I know that bro but I want custom solutions for what I said, like vector database or postgre, the fact is I don't know how to use them, the tutorials are not streight forward unlike Tim's tutorial also docs are not able to provide me specific solution. Yes I know after reading docs I will be able to do that but I have very little time (3 Days), and under these days I will have to add 7 tools to the AI agent. Otherwise I'm continuously trying to do that. ❤️ If you can help me through any article or blog or email, please do that 🙏❤️

[@davidtindell950]

Thx. Tim ! Now, llama3.1 is available under Ollama, It generates great results and has a large context memory !

[@siddhubhai2508]

@@davidtindell950 But bro my project is accordingly that can't depend on the LLM's context memory. Please tell me if you can help me with that!

[@davidtindell950]

​@@siddhubhai2508 I have found FAISS vector store provides an effective and large capacity "persistent memory" with CUDA GPU support. ...

[@swankyshivy]

how can this be moved from locally to on an internal website

[@alexandresemenov8671]

Hello! Tim when i run ollama directly there is no delay in response but using script with langchain some delay appear. Why is that? How to solve it?

[@tengdayz2]

Thank You.

[@ccKuang-ziqian]

should i install ollama in a virtural env?

[@sunhyungkim5764]

Amazing!

[@Eyuel3256]

I had been using the program Ollama on my laptop, and it was utilizing 101% of my CPU's processing power. This excessive usage threatened to overheat my device and decrease its performance. Therefore, I decided that I would discontinue using the program.

[@БогданСірський]

Hey, Tim! Thanks for your tutorial. A haver a problem. Bot isn't responding to me? Mabe someone else have the same problem. Give me some feedback, please

[@jorgeochoa4032]

hello do you know if its possible to use this model as a "pre trained" one, and add some new let say.. local information to the model to use it for an specific task?

[@AmitErandole]

Can you show us how to do RAG with llama3?

[@andhika277]

How much ram required to make this program running well? Cause i have 4GB ram only

[@TanujSharma-d9o]

Can You teach us how to implement it in GUI form, i don't want to run the program every time i want help of this type things

[@m.saksham3409]

I have not implemented myself, but I have doubt, you are using langchain where the model is llama 3.1, langchain manages everything here, then what's the use of Ollama ?

[@gunabaki7755]

the langchain simplifies interactions with LLM's, it doesn't provide the LLM. We use Ollama to get the LLM

[@TigerBrownTiger]

Why does microsoft publisher window keep popping up saying unlicensed product and will not allow it to run?

[@lyndonyang1269]

how coincidental, i made this project just 2 days ago

[@okotjakimgonzalo2270]

Where do you get all this stuff from

[@trevoro.9731]

If you need to work with large amounts of data OpenAI performance still can't be matched locally, unless you spend a ridiculous amount on your computer build.

[@hirthikbalajic]

it can be matched by running llama 3.1 405B model !.

[@abuhabban-tz8xj]

what's your pc specs sir?

[@thegamingaristocrat7615]

Is there any way to make python script to automatically train a locally-ran model?

[@sharanvellore9016]

Hi, I have tried this and its working, but the model is taking long response time anything I can do for reducing that?

[@cyrilypil]

How do you get Local LLM to show? I don’t have that in my VS Code

[@ruthirockstar2852]

is it possible to host this in a cloud server? so that i can access my custom bot whenever i want?

[@felipemachado8311]

can i train this model? give him information that he can answer to me before?

[@乾淨核能]

what's the minimum hardware requirement? thank you!

[@gunabaki7755]

8GB RAM

[@乾淨核能]

@@gunabaki7755 no discrete GPU needed?

[@VatsalyaB]

thx ;)

[@bhaveshsinghal6484]

Tim this ollama is running on my cpu and hence really slow can I make it run on my GPU somehow?

[@TotoyBabes]

do i need to install longchain?

[@734833]

Nice

[@kingkd7179]

It was a great tutorial and I follow it properly but still I am getting an error : ConnectError: [WinError 10061] No connection could be made because the target machine actively refused it I am running this code on my office machine which has restricted the openai models and Ai site

[@vivekanandl8798]

Does respose speed of AI bot depend on gpu like llama ?

[@gunabaki7755]

YES

[@Hrlover205]

i dont know what is happening when i run python file in cmd it shows me hello world then the command ends

[@antoniosa]

A dummy question.. Where is used the template ?

[@ParanoidNotAndroid]

Does anybody know what type of data the llama software is exchanging ?

[@aviralshastri]

how can we stream output ??

[@khushigupta5798]

Hey how can i show this as a ui I want to create a chatbot which can provide me the programming related ans with user authentication otp Please tell me how can i create this by using this model And create my UI i am a full stack developer ane new to ml please reply

[@pixelmz]

Sadly even though I have 32GB of ram, the 7B "llama3" takes up to 1 minute to answer

[@silasknapp4450]

Hi. Is there a way to uninstall llama3 again?

[@mit2874]

do i need vram 4 this ?

[@helloansuman]

Great. Now link it to front end and create the chatbot.

[@GeneKim-g1w]

How do I activate this on windows ??

[@opita_opica]

this context thing is not working, the bot does not know what was earlier in the conversation

[@PaulRamone356]

PS C:\Windows\system32> ollama pull llama3 Error: could not connect to ollama app, is it running? what seems to be wrong? (sorry for hte noob question)

[@gunabaki7755]

you need to run the ollama application first, it usually starts when u boot up ur pc

[@PaulRamone356]

@@gunabaki7755 will try ths thanks bro!

[@annismehdi343]

Hello there ... Can i e-mail you ? I am facing a problem and really could use some help

[@nomannosher8928]

Anyone knows Tim HW specs?

[@shadowtrell9961]

Anyone else read Ollama as Obama? No? Just me? Darn.

[@saberlite8809]

Do you know what's Obama's last name?

[@TDark1]

where is the script bro?

[@ahmedthebest]

Can you please refer back to the old way of recording and produce your content I really really don't like this new way

[@TechWithTim]

What’s wrong with this? If you give me some better feedback I’d be happy to take it into consideration!

[@ahmedthebest]

@@TechWithTim First of all, thank you very much for your quick response and your desire to do the best for us, the viewers. In the beginning, the video speed and the text that appears every short period of time is something I really don’t like. In the past, you used to record the video and explain it directly without exaggerated editing. I like your calm voice while you explain and also when you write the code in front of us. Sometimes some errors may occur, and this makes us feel that the video is more realistic. This is the really special thing about your channel. I hope that you return to the old way that you are accustomed to because I really love following your channel and it is like an entrance for me into the world of programming. I have been following it for a long time. I thank you in advance for reading the comment and I hope that you return to the old way as soon as possible.

[@ahmedthebest]

@@TechWithTim And of course I can't forget the legendary sound of the keyboard as you type the code. This really helps in entering death and learning better.

[@SnowballOfficials]

does it require gpu ?

[@gunabaki7755]

depends on size of the model, smaller models don't require

[@CrocodilesDen]

How do i deploy it to my website?

[@gunabaki7755]

I think u can try to convert it into API using FASTAPI and call the API from frontend