Infrastructure

AI Lab

ClairVise private AI infrastructure for Finance & IB workflow: local-first reasoning, knowledge, and automation.

Architecture

AI Stack Overview

Input

Voice (Superwhisper)

Text / Documents

Reasoning

Claude

Codex

Gemini CLI

Open WebUI

Memory

Open WebUI KBs

CACS Graphify

Engine

Apple Silicon

LM Studio

Qwen 3.5 Local LLM

Local Embeddings

Automation

n8n (Native Node)

Fireflies Webhooks

Ruthos (Always on private AI agent)

Philosophy

Design Principles

Local-first for sensitive data

Sensitive documents and proprietary models stay on-device. LM Studio runs on Apple Silicon for local chat and embeddings, while Open WebUI knowledge bases can be run with cloud connectors disabled.

Persistent knowledge graphs

CACS Graphify maintains structured memory across domain graphs for infrastructure, content, learning, and other knowledge areas, enabling compounding intelligence rather than ephemeral chat.

Research-first AI

AI is deployed for deep research synthesis and first-pass analysis. Judgement, valuation conclusions, and client-facing outputs remain entirely human.

Workflow Integration

This platform's workflow combines Claude, Codex, Gemini CLI, and Open WebUI for reasoning and local research, LM Studio on Apple Silicon for private Qwen chat and embeddings, CACS Graphify for persistent knowledge, and Ruthos (Always on private AI agent) for bounded Telegram-side orchestration. Cloud escalation is deliberate and cost-aware; sensitive work stays on local rails.