Skip to main content

Documentation Index

Fetch the complete documentation index at: https://mintlify.com/TrinaxCode/TrinaxAI/llms.txt

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

By the end of this guide you will have TrinaxAI running on your machine, the PWA open in your browser, your first chat sent to a local Ollama model, the CLI installed, and your first project indexed and queryable with RAG — all without touching a cloud service or entering a credit card.
1

Install TrinaxAI

Run the one-command installer for your OS. It downloads Ollama, pulls the right models for your hardware, sets up the RAG API and PWA, and configures auto-start.
curl -fsSL https://raw.githubusercontent.com/TrinaxCode/TrinaxAI/main/install.sh | bash
Before piping any script to your shell, review it first:
curl -fsSL https://raw.githubusercontent.com/TrinaxCode/TrinaxAI/main/install.sh | less
Alternatively, clone the repo and run ./install.sh locally so you can inspect every line before execution.

Install flags

Pass flags to tailor the install to your environment:
FlagDescription
--non-interactiveAutomatic install for CI/scripts — accepts all defaults
--no-modelsSkip downloading Ollama models (useful when re-installing)
--no-visionSkip the vision model download
--no-autostartDo not enable boot auto-start
--no-startDo not start TrinaxAI after the install finishes
--profile 8gb|16gb|max|ultraOverride the auto-detected hardware profile
--lan-systemEnable LAN system-control endpoints and auto-generate an admin token
# Example: non-interactive install on a 16 GB machine, no boot auto-start
./install.sh --non-interactive --profile 16gb --no-autostart
Platform-specific guides: Linux · macOS · Windows
2

Open the PWA

Once the installer finishes, open your browser and navigate to:
https://localhost:3334
The PWA is served over HTTPS using a self-signed certificate generated at install time. Your browser will show a security warning on first visit — this is expected. Click Advanced → Proceed to localhost (Chrome/Edge) or Accept the Risk (Firefox) to continue. You only need to do this once per browser.From a phone or tablet on the same Wi-Fi, visit https://[YOUR-LAN-IP]:3334 and install the app from the browser’s Add to Home Screen prompt.
The 7-step onboarding wizard will guide you through selecting a default model, language preference, and dark/light theme. You can revisit any setting later from the Settings panel.
3

Send your first chat message

In the PWA chat interface, type a question and press Enter (or the send button). TrinaxAI’s auto-router picks the best local model for your query — no configuration needed.Try asking:
Explain how async/await works in Python with a short example.
The response streams in real time from your local Ollama instance. No request leaves your machine.
Switch between the Ollama and RAG engines using the engine toggle in the chat toolbar, or force an engine from the CLI with --engine ollama or --engine rag.
4

Install the CLI

The trinaxai CLI is a Python package in the same repo. Install it in editable mode from the repo root:
pip install -e .
Verify the entry point is on your PATH:
trinaxai --help
Available commands: ask, chat, index, browse, research, memory, collections, watch, export, obsidian, doctor, start, stop
5

Run the health check

Confirm that all services are running and models are loaded:
trinaxai doctor
doctor checks Ollama connectivity, the RAG API health endpoint, model availability, storage layout, and embedding readiness. It prints a colour-coded report and exits non-zero if anything needs attention.
Any time something feels wrong — slow responses, missing citations, a blank PWA — run trinaxai doctor first. It surfaces the most common issues (Ollama not running, missing model, full disk) in seconds.
6

Index your first project

Point TrinaxAI at a directory of code or documents. The indexer uses AST-aware chunking for 15+ programming languages and is incremental — re-running only processes changed files.
trinaxai index ~/my-project
You can also index from the PWA: open Settings → Indexing and enter the path, or drag-and-drop files onto the Knowledge Browser.Watch the progress bar as chunks are embedded with bge-m3 (or nomic-embed-text on the 8gb profile) and persisted to storage/. Large repositories finish in seconds on subsequent runs thanks to the manifest-based change detection.
7

Ask a question with RAG

Once indexing is complete, query your codebase with the RAG engine. TrinaxAI retrieves the most relevant chunks via hybrid vector + BM25 search, reranks them, and synthesises a grounded answer with source citations.
trinaxai ask "How does authentication work?" --engine rag
The CLI prints the answer followed by a Sources block showing the file path, collection, and relevance score for each cited chunk. You can browse and explore chunks interactively with:
trinaxai browse
In the PWA, citations appear as expandable Source cards beneath each RAG response.

What to explore next

TopicLink
RAG indexing options, collections, and the file watcherRAG & Indexing
Voice mode and vision (image analysis)Voice & Vision
Persistent memory and knowledge collectionsMemory & Collections
PWA install on iOS, Android, and desktopPWA
Security model, LAN hardening, and .env referenceSecurity Model
Full CLI command referenceCLI Reference

Build docs developers (and LLMs) love