Augmented AI Studio

A compound, self-hosted intelligence workspace: long-form reasoning, source-grounded Deep Research, project documents, isolated Python analysis, visual/street-scene reasoning, persistent session memory and specialist/custom agents. The web app never falls back to a commercial AI provider.

Independent intelligence layer

Reason, research, inspect, calculate and verify in one workspace

General reasoning is separated from retrieval, code execution and evidence. That means a citation comes from a source, a number can come from reproducible code, and the model is not allowed to invent either.

Checking engine…

Web interface ready; owned model host not connected. The deterministic planning checker/calculators continue to work independently. This studio intentionally does not route around an offline self-hosted engine by sending your content to Lovable, OpenAI, Anthropic, Perplexity or another paid AI API.

Intelligence mode

AgentMode

Allow web research
Uses the self-hosted research/search layer. Leave off for local-only reasoning.

Ask the project, not a generic chatbot

Attach evidence, select an agent/mode, and the engine can retrieve local sources, use its project memory, calculate, run isolated Python or research the web when you explicitly allow it.

The studio is an evidence and productivity system, not a substitute for planning, legal, tax, valuation, structural, QS or lending advice. Web research, when enabled, reads external public sources through the self-hosted research layer; the language/vision/code models themselves are owned open-weight deployments rather than commercial AI APIs.