Persistent context
Carry project state, decisions and relevant memory forward instead of treating every conversation as isolated.
Luna is being built as a local-first, persistent AI operator inside EstOS Intelligence, designed to carry context forward, research deeply, use tools, remember what matters and help turn intelligence into action.
The goal is continuity. Luna should know what she is working on, why it matters, what has already been decided and which tools are available, without forcing the user to rebuild the context every time.
Carry project state, decisions and relevant memory forward instead of treating every conversation as isolated.
Investigate source material, market conditions, technical questions and competing evidence before recommending an answer.
Move from analysis into governed action, including research workflows, calculations, document work and future automation.
Luna can research, challenge, recommend and execute within defined permissions. Important decisions remain under human control.
Luna’s personality, operating principles and role are part of the architecture. Her visual identity follows the same rule: consistent, recognizable and grounded in the system we are building.
“The useful answer is not always the first one that sounds right. Sometimes the real work begins when you question it.”
Luna is not intended to be one model pretending to do everything. The runtime can route work across specialized FAST, REASON and, later, RESEARCH roles while preserving one consistent identity, memory layer and operating contract. Models run locally through Ollama; the configured baseline is shown below and has not been live-verified from this website.
qwen3.5:9b
Personality, conversation, simple tasks and routing.
qwen3.5:27b
Thinking, analysis, coding, planning and difficult conversations.
Model not yet selected
Deep research, evidence synthesis, source comparison and long-form investigation.
The development path is deliberately staged. Luna should become useful, reliable and controllable before we spend energy making her look alive.
Reliable local execution, behavioural contracts and model routing.
ACTIVEPersistent project, relationship and decision context with controlled recall.
BUILDINGResearch, structured actions, EstOS workflows and professional automation.
EXPANDINGNatural speech, listening states and an interruptible conversational loop.
PLANNEDA responsive visual avatar with subtle states for listening, thinking and speaking.
PLANNEDThe long-term opportunity is not simply a construction chatbot. Luna can become the persistent operating layer that connects EstOS Intelligence to professional and eventually personal workflows.
Research pricing, challenge assumptions, work with estimating tools, preserve project context and help convert lessons into reusable systems.
Compare sources, investigate technical questions, monitor relevant information and surface what deserves human attention.
Use approved tools to move from answer to action, while respecting permissions, gates and auditability.
With clear separation and governance, the same architecture could eventually support broader personal-assistant workflows across devices.
Luna runs on private local hardware and is not reachable from this website or from the Launch Hub. Remote access will only be considered once a separately designed, authenticated and encrypted path has been reviewed and approved. The local runtime will never be exposed directly to the public internet.
Conversation, voice, project context, research, tools and runtime status will eventually meet behind one protected entry point.
Disabled on purpose. No connection to Luna is attempted from the web.
Luna does not need to pretend to be human. The goal is something more useful: a consistent AI partner that remembers, reasons, researches and operates within a system you control.