local ollama

[SimulationApe]

How to use the local ollama model? I didn't see any instructions on connecting to a local model in the wiki. Could you please guide me? thx.

[Vermino]

I followed my guide I asked Claude on building it out deepwiki-open. I personally changed my top_k to 20 instead of 5.

Prerequisites

1. Install Ollama: Download from https://ollama.com/download
2. Pull the embedding model:

   ollama pull nomic-embed-text
   

Configuration

1. Configure the Embedder

Edit api/config/embedder.json:

{
"embedder": {
"client_class": "OllamaClient",
"model_kwargs": {
"model": "nomic-embed-text"
}
},
"retriever": {
"top_k": 5
}
}

2. Set Environment Variables

Create/edit your .env file:

Ollama configuration

OLLAMA_HOST=http://localhost:11434

You still need an LLM API key for wiki generation

(Ollama is used for embeddings only)

GOOGLE_API_KEY=your_google_api_key_here
OR
OPENAI_API_KEY=your_openai_api_key_here

Server configuration

PORT=8001
SERVER_BASE_URL=http://localhost:8001

3. Docker Setup (if using Docker)

If running via Docker Desktop, change OLLAMA_HOST in .env to:

OLLAMA_HOST=http://host.docker.internal:11434

This allows the Docker container to access Ollama running on your host machine.

4. Verify Ollama is Running

Check if Ollama is running

curl http://localhost:11434/api/tags

Test the embedding model

ollama run nomic-embed-text

What Does Ollama Do in DeepWiki?

• Embeddings: Ollama generates vector embeddings of your code for semantic search (via nomic-embed-text)
• LLM Generation: You still need an API key (Google Gemini or OpenAI) for the actual wiki content generation

The nomic-embed-text model is optimized for code embeddings and works well for repository indexing.

Troubleshooting

Issue: "Connection refused" errors

• Solution: Ensure Ollama is running (ollama serve or check system tray)
• For Docker: Use host.docker.internal:11434 instead of localhost:11434

Issue: Model not found

• Solution: Pull the model first: ollama pull nomic-embed-text