Ollama Copilot

Proxy that allows you to use ollama as a copilot like Github copilot

Installation

Ollama

Ensure ollama is installed:

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

Or follow the manual install.

Models

To use the default model expected by ollama-copilot:

ollama pull codellama:code

DeepSeek

To use DeepSeek:

ollama-copilot -provider deepseek -token YOUR_DEEPSEEK_API_KEY -model deepseek-coder

Mistral

To use Mistral:

ollama-copilot -provider mistral -token YOUR_MISTRAL_API_KEY -model codestral-latest

Automatic Installation & Configuration

You can use the install.sh script to automate the build, installation, and configuration of ollama-copilot and your editors:

./install.sh

This script will:

1. Build the binary and install it to ~/.local/bin/ollama-copilot.
2. Optionally install and start a Systemd Service (allowing you to configure num-predict).
3. Automatically detect and offer to configure Neovim, VSCode, and Zed.

Manual Installation

Binary

go install github.com/bernardo-bruning/ollama-copilot@latest

Running

Ensure your $PATH includes $HOME/go/bin or $GOPATH/bin. For example, in ~/.bashrc or ~/.zshrc:

export PATH="$HOME/go/bin:$GOPATH/bin:$PATH"

ollama-copilot

or if you are hosting ollama in a container or elsewhere

OLLAMA_HOST="http://192.168.133.7:11434" ollama-copilot

Configuration

You can configure the server using command-line flags:

Flag	Default	Description
-port	:11437	HTTP port to listen on
-proxy-port	:11438	HTTP proxy port
-port-ssl	:11436	HTTPS port to listen on
-proxy-port-ssl	:11435	HTTPS proxy port
-cert		Certificate file path (*.crt) for custom TLS
-key		Key file path (*.key) for custom TLS
-provider	ollama	Provider to run LLM
-token	TOKEN	Token to pass for provider
-model	codellama:code	LLM model to use
-num-predict	250	Number of tokens to predict (recommended 25 for copilot)
-template	<PRE> {{.Prefix}} <SUF> {{.Suffix}} <MID>	Prompt template for fill-in-middle
-system	You are a helpful...	System prompt to guide the model

Configure IDE (Manual)

Neovim

1. Install copilot.vim
2. Configure variables

let g:copilot_proxy = 'http://localhost:11435'
let g:copilot_proxy_strict_ssl = v:false

VScode

1. Install copilot extension
2. Sign-in or sign-up in github
3. Configure open settings config and insert

{
    "github.copilot.advanced": {
        "debug.overrideProxyUrl": "http://localhost:11437"
    },
    "http.proxy": "http://localhost:11435",
    "http.proxyStrictSSL": false
}

Zed

1. Open settings (ctrl + ,)
2. Set up edit completion proxying:

{
    "features": {
        "edit_prediction_provider": "copilot"
    },
    "show_completions_on_input": true,
    "edit_predictions": {
        "copilot": {
            "proxy": "http://localhost:11435",
            "proxy_no_verify": true
        }
    }
}

PyCharm / JetBrains

1. Open Settings (Ctrl+Alt+S)
2. Navigate to Appearance & Behavior > System Settings > HTTP Proxy
3. Select Manual proxy configuration:
   • HTTP
   • Host name: localhost
   • Port number: 11435
4. Navigate to Tools > Server Certificates
5. Check Accept non-trusted certificates automatically
6. Restart PyCharm

Emacs

(experimental)

1. Install copilot-emacs
2. Configure the proxy

(use-package copilot
  :straight (:host github :repo "copilot-emacs/copilot.el" :files ("*.el"))  ;; if you don't use "straight", install otherwise
  :ensure t
  ;; :hook (prog-mode . copilot-mode)
  :bind (
         ("C-<tab>" . copilot-accept-completion)
         )
  :config
  (setq copilot-network-proxy '(:host "127.0.0.1" :port 11434 :rejectUnauthorized :json-false))
  )

Roadmap

• Enable completions APIs usage; fill in the middle.
• Enable flexible configuration model (Currently only supported llamacode:code).
• Create self-installing functionality.
• Auto-configure IDEs (Neovim, VSCode, Zed).
• Documentation on how to use.
• Windows setup