Using Devika in Local With Ollama

[gorbabor]

I need this help on the configuration to do

[hqnicolas]

$ git clone https://github.com/hqnicolas/devika
$ cd devika
create config file
edit the docker compose to point to your ollama server url
$ sudo docker compose build
$ sudo docker compose up

[hqnicolas]

same as #257 #257 257

[ChanghongYangR]

Any ideas why I couldn't get ollama work?

[hqnicolas]

$ git clone https://github.com/hqnicolas/devika
$ cd devika
create config file
edit the docker compose to ollama url
$ sudo docker compose build
$ sudo docker compose up

[gorbabor]

make it work, you'll need to launch Ollama as it's being described. I'll explain each step assuming you haven't installed Ollama yet. If that's the case:

Go to the Ollama website to download the version that suits you (I have Windows, so I downloaded the Windows version).
After downloading, install Ollama. By default, Ollama will start running on machine startup.
If that's not the case, you'll need to open a console (on Windows, launch the command prompt by typing "cmd" in the search bar) and enter the following command: "Ollama Serve." This will launch Ollama as a server, and it will provide its API through port 11434.

[cardchase]

you would need to open another terminal window. then
cd devika/ui
or if you click the '+' and just
it should be comething like
~/devika$
then run

~/devika$ cd ui
~/devika/ui$ bun install 

then
~/devika/ui$ bun run dev
you should get

(base) user@user:~/devika/ui$ bun run dev
$ vite dev

  VITE v5.2.2  ready in 285 ms

  ➜  Local:   http://localhost:3000/
  ➜  Network: use --host to expose
  ➜  press h + enter to show help

a successful process would look like this:

I have run install and dev twice cause I had restarted the process but this is how you should be able to run.

It will be accessed via http://localhost:3000/ in your web browser

run these steps every time you have disconnected devika or started it

[gorbabor]

To make it work, you'll need to launch Ollama as it's being described. I'll explain each step assuming you haven't installed Ollama yet. If that's the case:

1. Go to the Ollama website to download the version that suits you (I have Windows, so I downloaded the Windows version).
2. After downloading, install Ollama. By default, Ollama will start running on machine startup.
3. If that's not the case, you'll need to open a console (on Windows, launch the command prompt by typing "cmd" in the search bar) and enter the following command: "Ollama Serve." This will launch Ollama as a server, and it will provide its API through port 11434.

[phalexo]

Being able to access the API from ollama does NOT mean it will work. It will encounter all kinds of issues wth the local model itself, not producing the same output as GPT-4.

[trymeouteh]

Would like to see easy setup feature for Ollama

[bjornwed]

running dockerized version I had to set Ollama endpoint to "http://ollama-service:11434" as the container running ollama in devika-subnetwork will be accessible through the host "ollama-service". After that all good, time to play with local llama3 :)

[rajeevchourasiya]

Actually when i run devika, it is not browsing the web although i have entered my google API key, but not sure what to put in the endpoint.

[bjornwed]

The endpoint for Google should already be in the config.toml file, but if not it is: https://www.googleapis.com/customsearch/v1

So in Google you need to create a search engine, get the ID for the engine and create a key for your engine:
https://developers.google.com/custom-search/v1/introduction

To check everything's working correct on Google side, you can simply try the following in a browser:

https://www.googleapis.com/customsearch/v1?key=[insert) your key]&cx=[insert your engine id]&q=lectures

[rajeevchourasiya]

Thanks, but I want to use ollama. Although everything is set up perfectly, it just doesn't output anything, just stepwise plan at most sometimes, that's it.

[bjornwed]

You need to configure both, search engine for the agent to be able to do research for the project, and ollama or other LLM for agents to be able to perform tasks like generate code etc.
Also, depending on your hardware, response may be painfully slow if running everything local, so be patient.

[djpapzin]

Can you explain more what you mean

[albertotorresfoltyn]

still can;t run devika with local ollama

[djpapzin]

Have you tried OpenDevin?

…

On Wed, 14 Aug 2024, 01:16 Alberto Torres, ***@***.***> wrote: still can;t run devika with local ollama — Reply to this email directly, view it on GitHub <#308 (comment)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/A4QSQG2ZAHSWXDJY2G7SX63ZRKHWDAVCNFSM6AAAAABFTTTGHSVHI2DSMVQWIX3LMV43URDJONRXK43TNFXW4Q3PNVWWK3TUHMYTAMZTGEYTEMY> . You are receiving this because you commented.Message ID: ***@***.***>

[mchzimm]

i'd say that's out of scope here, this is about devika