Connect a Local self hosted LLM via Ollama LLM to Instant BI AI module of open source BI Helical Insight

For data security and other concerns, in many cases companie’s would like to host the LLM and AI on their own servers and not use the AI provided by BI provider. Most of the other BI products, open soruce and proprietary, bind and force you to use their LLM only.

Not Helical Insight. We even give you an option to self host the LLM (GenAI) locally also, and then connect to it via Ollama. You have full control and ownership of the data and applications. Not even a single bit of data will go outside. In this blog we are going to cover how to do the same.

How it fits together

Instant BI sends your question to the LLM provider you configure. In this guide, that provider is an Ollama container running on the same machine.

Setup roadmap

Before you begin

Prerequisites

• A running instance of Helical Insight is there. (Installation via Docker of Open Source BI Helical Insight – Helical Insight)
• Ollama installed and configured, via docker. (https://dev.to/ajitkumar/the-complete-guide-to-ollama-run-large-language-models-locally-2mge)
• Memory and disk: enough RAM and free disk space for your chosen model. On machines with limited memory, start with a small model (about 3 billion parameters).
• Server file access: permission to edit files on the Helical Insight server (needed in Step 5).

Two separate downloads

People often mix these up. You will download two different things:

Step 1 | Start Ollama in Docker

Open Command Prompt or PowerShell and run the following two commands.

1. Download the Ollama image

docker pull ollama/ollama

2. Start the container

docker run -d --name ollama -v ollama:/root/.ollama -p 11434:11434 ollama/ollama

What the options mean:

Step 2 | Download a model

Pick a model that fits your available memory. This guide uses qwen2.5-coder:3b, a small coding-oriented model that suits generating SQL and runs on modest hardware. Download it into the running container:

docker exec -it ollama ollama pull qwen2.5-coder:3b

When the download finishes, list the installed models:

docker exec ollama ollama list

Step 3 | Verify Ollama on the host

Confirm the container is running, then test the API from Windows.

Check the container is running

docker ps

List models through the API

curl.exe http://127.0.0.1:11434/api/tags

Expected result: HTTP 200 and a JSON response containing a models array that includes your model.

Now test that the model can actually generate text. Use the version for your shell:

Command Prompt

curl.exe http://127.0.0.1:11434/api/generate -H "Content-Type: application/json" -d "{\"model\":\"qwen2.5-coder:3b\",\"prompt\":\"Write SELECT * FROM employees;\",\"stream\":false}"

PowerShell

Invoke-RestMethod -Uri http://127.0.0.1:11434/api/generate -Method Post `

  -ContentType "application/json" `

  -Body '{"model":"qwen2.5-coder:3b","prompt":"Write SELECT * FROM employees;","stream":false}'

Step 4 | Verify Helical Insight can reach Ollama

Helical Insight must be able to reach Ollama from its own runtime, which is not always the same as your browser or Windows. If Helical Insight runs in Docker and Ollama’s port is published on the Windows host, use the special hostname host.docker.internal.

First, find the name of your Helical Insight application container:

docker ps

Then test the connection from inside that container:

docker exec <HELICAL_APPLICATION_CONTAINER> curl -sS http://host.docker.internal:11434/api/tags

Expected result: A JSON response listing your model confirms that Helical Insight can reach Ollama.

The Base URL to use for this arrangement (Docker Desktop on Windows) is:

http://host.docker.internal:11434

Step 5 | Register the model in the catalog file

The list of models shown in Instant BI comes from a server-side catalog file, not only from the web interface. To make a new Ollama model appear, add its exact tag to this file.

1. Open the Instant BI configuration folder on the Helical Insight server (path shown is an example and may differ in your version):

cd <HELICAL_INSTALL_DIR>/hi/hi-repository/System/InstantBI

ls

nano llm_models_catalog.yaml

2. Find the langchain-ollama section and add your model on a new line, matching the indentation and format of the entries around it. For example:

Example (your list may differ)

langchain-ollama:

  - deepseek-coder-v2

  - llama3.2

  - llama3.1

  - mistral

  - qwen2.5-coder:3b      # <- your model, exact tag from 'ollama list'

  - codellama

Figure 1. The catalog file in the nano editor. The langchain-ollama section lists the available Ollama models, including qwen2.5-coder:3b.

3. Save and exit. In nano: press Ctrl+O, then Enter to save, then Ctrl+X to exit.

4. Reload if required. If the model list does not refresh on its own, follow the restart or reload procedure approved for your Helical Insight deployment.

YAML is sensitive to indentation. Use spaces, not tabs, and copy the pattern of the existing entries. Entry structure can vary by version, so follow an existing entry from your own file.

Step 6 | Configure the Ollama provider in the web UI

Open the provider settings

1. Sign in to Helical Insight as an administrator.
2. Open the Instant BI tab in the administration area.
3. Select LLM Providers.
4. Click the Ollama provider card to edit it.

Figure 2. LLM Providers page. Provider types and labels vary by version and configuration. Click the Ollama card to configure it.

Fill in the provider fields

In the Edit Provider (ollama) dialog, set the following:

Figure 3. The Edit Provider (ollama) dialog. The Base URL shown (localhost) is only an example; use the one that works for your setup.

Choose the right Base URL

The correct Base URL depends on where Helical Insight and Ollama each run. Pick the row that matches your setup:

Click Save when done.

Step 7 | Confirm Instant BI is ready

1. Open Instant BI from the Helical Insight home page.
2. Make sure a semantic model is selected or connected (the cube icon in the prompt box). Click the link to learn how to create semantic model.
3. Ask a question about your data, or click one of the suggested questions.
4. Check that Instant BI returns an answer using your Ollama model.

Figure 4. Instant BI ready for use. Suggested questions depend on the connected semantic model and its metadata, and the interface may vary by version.

Troubleshooting

Quick reference

Key points to remember

• The model name in Helical Insight must match the Ollama tag exactly, including suffixes like :3b or :latest.
• localhost does not work across containers. Always test from the Helical Insight runtime.
• Keep the ollama Docker volume so downloaded models survive container re-creation.
• Add catalog entries by following the existing YAML structure. Never replace the whole file.
• Before production use, review your organization’s security policies, resource allocation, model licensing, and data-handling requirements.