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.

Raw fleet data — device inventories, policy results, vulnerability counts — only becomes useful when you can ask questions of it and get clear answers. NextAudit AI embeds a fully self-hosted AI stack that lets your security and IT teams query fleet telemetry in natural language, surface patterns across thousands of devices, and generate audit-ready summaries, all without sending a single byte to an external AI provider.

​
How the AI stack is structured

The AI layer in NextAudit AI is composed of three coordinated services:
  • Ollama — runs large language models locally on your own hardware
  • Flowise — provides a visual AI agent builder for constructing analysis workflows
  • PostgreSQL with pgvector — stores vector embeddings for semantic search over historical fleet data
These services are wired together so that Flowise agents can call Ollama for inference and query the vector store for context retrieval, all within your network perimeter.
Because Ollama runs inference locally, you choose which model is loaded via the OLLAMA_MODELS environment variable. This means you can select a model sized appropriately for your hardware — from lightweight models on a single server to larger models on GPU-equipped infrastructure.

​
Flowise AI agent flows

Flowise is the orchestration layer where you build and run AI-assisted analysis workflows. Using its visual interface, operators can construct agent flows that:
  • Accept a natural-language question about fleet state
  • Retrieve relevant device records and policy results from the database
  • Call the Ollama LLM to reason over that context
  • Return a structured, human-readable answer
Typical questions an operator might pose through a Flowise agent flow include:
  • “Which endpoints are missing critical patches from the last 30 days?”
  • “Summarize the current compliance posture across the Windows fleet.”
  • “Which devices have failed more than three consecutive policy checks?”
  • “Generate an executive summary of vulnerability exposure for this week’s audit.”
Flowise persists its agent flow definitions in PostgreSQL, so your analysis workflows are version-tracked alongside your other operational data and survive container restarts.

​
Ollama local LLM inference

Ollama provides the inference runtime — it loads and serves LLM models from local storage and exposes them via a standard API that Flowise queries at analysis time. Because inference is local, response latency depends on your hardware rather than external API availability, and there are no per-token usage costs or rate limits imposed by a third party. The Ollama service persists downloaded models in a dedicated volume (ollama_data), so models are downloaded once and reused across restarts without re-fetching.
Ollama supports a range of open-weight models. The specific models available to your deployment are configured via OLLAMA_MODELS. Consult the Ollama documentation for the current list of supported models and their hardware requirements.
NextAudit AI uses a PostgreSQL instance with the pgvector extension enabled. The EMBEDDING_SIZE configuration controls the dimensionality of the vector embeddings stored in the database. This setup enables Flowise agents to perform semantic similarity searches over historical fleet data — finding devices, events, or audit records that are conceptually related to a query even when they don’t share exact keyword matches. In practice, semantic search allows agents to:
  • Find fleet events that are contextually similar to a described incident
  • Retrieve past audit findings relevant to a current investigation
  • Surface related compliance failures across different policy categories

​
Privacy and data residency

Every component in the AI stack runs within your own infrastructure. Ollama inference, Flowise agent execution, and vector storage all operate on-premises. Fleet telemetry, audit findings, and the questions you ask of your data never leave your environment.
This architecture is particularly important for organizations with regulatory data residency requirements or internal policies that prohibit sending device inventory or vulnerability data to external cloud services.

Flowise AI service

Service configuration for Flowise, including database connection and port settings.

Ollama service

Service configuration for Ollama local LLM inference and model management.

Build docs developers (and LLMs) love

Get started for freeTalk to us