OpenHands/docs/documentation/LOCAL_LLM_GUIDE.md

Local LLM Guide with Ollama server

0. Install Ollama:

run the following command in a conda env with CUDA etc.

Linux:

curl -fsSL https://ollama.com/install.sh | sh

Windows or macOS:

• Download from here

1. Install Models:

Ollama model names can be found here (See example below)

Once you have found the model you want to use copy the command and run it in your conda env.

Example of llama2 q4 quantized:

conda activate <env_name>
ollama run llama2:13b-chat-q4_K_M

you can check which models you have downloaded like this:

~$ ollama list
NAME                            ID              SIZE    MODIFIED
llama2:latest                   78e26419b446    3.8 GB  6 weeks ago
mistral:7b-instruct-v0.2-q4_K_M eb14864c7427    4.4 GB  2 weeks ago
starcoder2:latest               f67ae0f64584    1.7 GB  19 hours ago

2. Run Ollama in CLI:

This command starts up the ollama server that is on port 11434

This will show the requests in CLI

conda activate <env_name>
ollama serve

or

This will run with no output in the background

sudo systemctl start ollama

If you see something like this:

Error: listen tcp 127.0.0.1:11434: bind: address already in use

This is not an error it just means the server is already running

To stop the server use:

sudo systemctl stop ollama

For more info go here

3. Follow the default installation of OpenDevin:

git clone git@github.com:OpenDevin/OpenDevin.git

or

git clone git@github.com:<YOUR-USERNAME>/OpenDevin.git

then

cd OpenDevin

4. Run setup commands:

make build
make setup-config

5. Modify config file:

• After running make setup-config you will see a generated file OpenDevin/config.toml.
• Open this file and modify it to your needs based on this template:

LLM_API_KEY="ollama"
LLM_MODEL="ollama/<model_name>"
LLM_EMBEDDING_MODEL="local"
LLM_BASE_URL="http://localhost:<port_number>"
WORKSPACE_DIR="./workspace"

Notes:

• The API key should be set to "ollama"
• The base url needs to be localhost
• By default ollama port is 11434 unless you set it
• model_name needs to be the entire model name
  • Example: LLM_MODEL="ollama/llama2:13b-chat-q4_K_M"

6. Start OpenDevin:

At this point everything should be set up and working properly.

1. Start by running the ollama server using the method outlined above
2. Run make build in your terminal ~/OpenDevin/
3. Run make run in your terminal
4. If that fails try running the server and front end in sepparate terminals:

• In the first terminal make start-backend
• In the second terminal make start-frontend

5. you should now be able to connect to http://localhost:3001/ with your local model running!

Additional Notes for WSL2 Users:

1. If you encounter the following error during setup: Exception: Failed to create opendevin user in sandbox: b'useradd: UID 0 is not unique\n'
   You can resolve it by running:

   export SANDBOX_USER_ID=1000
   

2. If you face issues running Poetry even after installing it during the build process, you may need to add its binary path to your environment:

   export PATH="$HOME/.local/bin:$PATH"
   

3. If you experiencing issues related to networking, such as NoneType object has no attribute 'request' when executing make run, you may need to configure your WSL2 networking settings. Follow these steps:

   • Open or create the .wslconfig file located at C:\Users\%username%\.wslconfig on your Windows host machine.
   • Add the following configuration to the .wslconfig file:

     [wsl2]
     networkingMode=mirrored
     localhostForwarding=true
     

   • Save the .wslconfig file.
   • Restart WSL2 completely by exiting any running WSL2 instances and executing the command wsl --shutdown in your command prompt or terminal.
   • After restarting WSL, attempt to execute make run again. The networking issue should be resolved.