2.1.8 Frontend: BionicGPT

BionicGPT

Handle: bionicgpt
URL: http://localhost:33901

Starting

# [Optional] BionicGPT requires quite a
# few services. You might want to pull them
# ahead of starting the service
harbor pull bionicgpt

# Will start bionicgpt alongside
# the default webui
harbor up bionicgpt

Harbor can't configure BionicGPT ahead of time as connectivity configuration is stored in Postgres and can be adjusted via the UI after service starts.

BionicGPT supports ollama and OpenAI-compatible backends out of the box.

Ollama

When configuring Ollama in BionicGPT, use Harbor's internal Ollama URL because BionicGPT runs inside the Harbor Docker network. Start Ollama with BionicGPT, then get the internal URL:

harbor up bionicgpt ollama
harbor url -i ollama

Use that value with /v1 appended as the OpenAI-compatible base URL. On current Harbor, it resolves through Docker Compose service DNS as:

http://ollama:11434/v1

BionicGPT also expects a non-empty API key for OpenAI-compatible providers; use a placeholder such as sk-ollama. For the model name, choose a model already available in Ollama:

harbor ollama list
harbor ollama ls

If the model is missing, pull it first:

harbor ollama pull llama3.2:3b