Pre-existing things installed

[NZ-Wanderer]

This may be a silly question, but how do I install Odysseus if I already have pre-existing things installed..
All the YT videos I have watched etc are presuming users have noting installed.
I already got these installed:
Local Docker
Local Ollama (and models)
Local Searxng (on port 8888)
and I am assuming that using the official way mentioned on here it going to install something I already got installed (unless I wrong)
If I am correct is there some kind person that could tell me what I need to do to install Odysseus on my linux (cachyos) system please??

[MarufHasan-dev]

Not a silly question at all, and you are spot on with your assumption. The out-of-the-box Docker setup will try to spin up a duplicate SearXNG container, which will immediately clash with your existing port 8888.

Before breaking down the exact configuration steps for CachyOS, it helps to know which deployment path you prefer for your workflow, as you have two clean ways to handle this:

1. The Containerized Path: You keep Odysseus isolated inside a Docker container, but we edit the docker-compose.yml to strip out the duplicate SearXNG service and use a host-gateway mapping so the container can talk back out to your existing host-managed Ollama and SearXNG.
2. The Native Path: You run Odysseus natively on your system using a Python virtual environment and uvicorn. Since your heavy lifting (Ollama and SearXNG) is already running locally on the host, running the UI natively completely sidesteps any container-to-host networking friction.

Which approach sounds better for your setup?

[NZ-Wanderer]

Thank you for the reply :)
I’ll go with the native path (Python venv + uvicorn).
I already have Python 3.14.5 and uvicorn 0.49.0 installed, and I’m running Ollama and SearXNG locally, so native integration should fit my setup better.

[MarufHasan-dev]

@NZ-Wanderer Better choice, the native path makes perfect sense since your stack is already alive on the host. Running it natively completely eliminates container isolation issues, meaning you can just point to localhost across the board without messing with bridge networks.

Plus, there is a major functional benefit to going this route: because Odysseus features deep agentic workflows (like running shell commands, executing python scripts, and writing local files), running natively allows the agent to actually interact with your real system files and workspace directly. If you ran it inside Docker, the agent would be entirely sandboxed inside an isolated container instance, meaning it wouldn't have any real visibility into your actual host files or system tools unless you manually mapped volumes.

Here is the exact setup to match your existing environment:

1. Clone and prepare the repository:

git clone https://github.com/pewdiepie-archdaemon/odysseus.git
cd odysseus

2. Spin up your venv and pull dependencies:
   Since you have Python 3.14.5 ready to go, build the virtual environment and install the package requirements:

   python3 -m venv venv
   source venv/bin/activate
   pip install -r requirements.txt

3. Configure the environment for your SearXNG:
   Copy the example config file to initialize your overrides:

cp .env.example .env

Open .env in your text editor. According to the configuration table in the docs, the exact variable name to override is SEARXNG_INSTANCE (not URL). Update it to your local port:

SEARXNG_INSTANCE=http://localhost:8888

4. Initialize and Launch:
   Run the project's setup script to seed the database, then kick off the application using your local uvicorn:

python setup.py
python -m uvicorn app:app --host 127.0.0.1 --port 7000

Connecting your Ollama:
You don't actually need to pre-seed your Ollama configurations inside the .env file for a native run. Once uvicorn boots up, open your browser to http://localhost:7000, log in with the temporary admin password printed in your terminal, head over to Settings, and add your local endpoint directly inside the UI: http://localhost:11434/v1.

One quick tip for CachyOS: If you plan on using the built-in "Cookbook" feature to scan your hardware or handle background model serving, make sure you have tmux installed on your system (sudo pacman -S tmux). The native backend relies on it to run background downloads and serves outside the main process.

[NZ-Wanderer]

Many thanks, worked perfectly :) - Up and running, now I can have a play with it...