Skip to main content

Documentation Index

Fetch the complete documentation index at: https://mintlify.com/Kevin2523/nextAuditAi/llms.txt

Use this file to discover all available pages before exploring further.

Ollama is the inference engine that powers the AI layer of NextAudit AI. It serves large language models through a local HTTP API, meaning all inference requests from Flowise stay within the stack’s network boundary. No prompts, audit findings, or host data are transmitted to external AI providers.

​
Image and build

The Ollama service uses different sources depending on the environment:
In development, Ollama is built from a local ./ollama context. This lets you customize the image — for example, to pre-bake specific models or apply configuration changes — and rebuild quickly without pulling from a registry.
ollama:
  build:
    context: ./ollama
  container_name: ollama

​
Service configuration

The full service definition (shown here for prod/test) exposes the Ollama API on OLLAMA_PORT on the host, mapping to the fixed internal port 11434: 
ollama:
  image: jjsotom2k4/ollama-ai:${VERSION}
  container_name: ollama
  environment:
    OLLAMA_MODELS: ${OLLAMA_MODELS}
  ports:
    - "${OLLAMA_PORT}:11434"
  volumes:
    - ollama_data:/root/.ollama
  restart: unless-stopped

​
Environment variables

​
OLLAMA_MODELS

environment:
  OLLAMA_MODELS: ${OLLAMA_MODELS}
OLLAMA_MODELS specifies which models Ollama pre-loads when the container starts. Set this to a comma-separated list of model names (using Ollama’s model tag syntax, e.g. llama3.2,nomic-embed-text). Pre-loading avoids the latency of on-demand pulls when Flowise first requests inference.
Include both a chat model and an embedding model in OLLAMA_MODELS if you plan to use Flowise’s vector store features. The embedding model dimension must match the EMBEDDING_SIZE set on the PostgreSQL service.

​
VERSION

The VERSION variable is used in the prod and test compose files to pin the jjsotom2k4/ollama-ai image tag: 
image: jjsotom2k4/ollama-ai:${VERSION}
Set VERSION in your environment file (prod.env, test.env) to control which image revision is deployed. This ensures that infrastructure updates are explicit and auditable.

​
Volume

volumes:
  - ollama_data:/root/.ollama
The ollama_data volume stores downloaded model weights. Models can be several gigabytes each; persisting this volume prevents re-downloading on container restart.

​
Internal API for Flowise

Ollama listens on port 11434 inside the container. Flowise connects to it using the Docker service hostname: 
http://ollama:11434
This connection never leaves the Docker network. To configure it in Flowise, add an Ollama chat model or embeddings node and set the base URL to http://ollama:11434.
The host-facing OLLAMA_PORT is available for direct API calls during development (e.g., testing a model with curl), but production traffic between Flowise and Ollama uses the internal Docker network exclusively.

​
Privacy and data isolation

Because Ollama runs inside the same Docker Compose stack as the rest of NextAudit AI, inference requests never leave the host. Audit findings, osquery results, policy data, and host metadata processed by AI flows remain within the infrastructure boundary defined by the stack. This is particularly relevant for deployments subject to data residency or confidentiality requirements.

Build docs developers (and LLMs) love

Get started for freeTalk to us