Ollama Nextcloud ExApp

A Nextcloud External Application (ExApp) that integrates Ollama for local LLM inference directly into Nextcloud.

Features

• Run LLMs Locally - Complete privacy with on-premise AI
• GPU Acceleration - Support for NVIDIA and AMD GPUs
• Model Library - Easy access to Llama, Mistral, CodeLlama, and more
• OpenAI-compatible API - Drop-in replacement for OpenAI endpoints
• Efficient Memory Management - Run large models with optimized memory usage

Supported Models

Model	Size	Use Case
llama3.2:1b	1.3GB	Fast, lightweight tasks
llama3.2:3b	2.0GB	Balanced performance
mistral:7b	4.1GB	General purpose
codellama:7b	3.8GB	Code generation
phi3:mini	2.2GB	Efficient reasoning

Requirements

• Nextcloud 30 or higher
• AppAPI installed and configured
• Docker with a configured Deploy Daemon (HaRP recommended)
• 8GB+ RAM (16GB+ recommended for larger models)
• NVIDIA GPU with CUDA (optional, for acceleration)

Installation

Via Nextcloud App Store

1. Install and enable the AppAPI app in Nextcloud
2. Configure a Deploy Daemon with GPU support if available
3. Search for "Ollama" in the External Apps section
4. Click Install

Manual Installation

# Register the ExApp with Nextcloud
occ app_api:app:register \
    ollama \
    <your-daemon-name> \
    --info-xml https://raw.githubusercontent.com/ConductionNL/ollama-nextcloud/main/appinfo/info.xml \
    --force-scopes

Configuration

Environment Variable	Description	Default
OLLAMA_DEFAULT_MODEL	Model to pull on startup	llama3.2:3b
OLLAMA_NUM_PARALLEL	Concurrent inference requests	2
OLLAMA_KEEP_ALIVE	Model unload timeout	15m
OLLAMA_MAX_LOADED_MODELS	Max models in memory	1

Usage

After installation, access the Ollama API through Nextcloud:

# List models
curl https://your-nextcloud/index.php/apps/app_api/proxy/ollama/api/tags

# Generate text
curl https://your-nextcloud/index.php/apps/app_api/proxy/ollama/api/generate \
    -d '{"model": "llama3.2", "prompt": "Hello!"}'

# Chat (OpenAI-compatible)
curl https://your-nextcloud/index.php/apps/app_api/proxy/ollama/v1/chat/completions \
    -d '{"model": "llama3.2", "messages": [{"role": "user", "content": "Hello!"}]}'

Pulling New Models

Models can be pulled via the API:

curl -X POST https://your-nextcloud/index.php/apps/app_api/proxy/ollama/api/pull \
    -d '{"name": "mistral:7b"}'

Development

Building the Docker Image

docker build -t ollama-exapp:dev .

Running Locally

docker run -it --rm \
    -e APP_ID=ollama \
    -e APP_SECRET=dev-secret \
    -e NEXTCLOUD_URL=http://localhost:8080 \
    -e OLLAMA_DEFAULT_MODEL=llama3.2:1b \
    -p 9000:9000 \
    -p 11434:11434 \
    --gpus all \
    ollama-exapp:dev

Testing Endpoints

# Health check
curl http://localhost:9000/heartbeat

# List models
curl http://localhost:9000/api/tags

# Generate
curl http://localhost:9000/api/generate -d '{"model":"llama3.2:1b","prompt":"Hi"}'

Architecture

┌─────────────────────────────────────┐
│         Nextcloud + AppAPI          │
└──────────────┬──────────────────────┘
               │
               ▼
┌─────────────────────────────────────┐
│    Ollama ExApp Container           │
│  ┌───────────────────────────────┐  │
│  │  FastAPI Wrapper (port 9000)  │  │
│  │  - /heartbeat                 │  │
│  │  - /init                      │  │
│  │  - /enabled                   │  │
│  │  - /api/* (proxy to Ollama)   │  │
│  │  - /v1/* (OpenAI compat)      │  │
│  └───────────────┬───────────────┘  │
│                  │                  │
│  ┌───────────────▼───────────────┐  │
│  │  Ollama Server (port 11434)   │  │
│  │  ┌─────────────────────────┐  │  │
│  │  │    Model Storage        │  │  │
│  │  │    /data/models         │  │  │
│  │  └─────────────────────────┘  │  │
│  └───────────────────────────────┘  │
└─────────────────────────────────────┘

Integration with Open WebUI

This ExApp works seamlessly with the Open WebUI ExApp to provide a chat interface for your models.

License

AGPL-3.0 - See LICENSE for details.

Links

• Ollama Documentation
• Ollama Model Library
• Nextcloud AppAPI Documentation
• Conduction