Make your AI stop feeling generic.
ReadTheRoom is a local, reviewable behavior-calibration layer for AI assistants. A short guided session tunes how your assistant handles directness, warmth, humor, profanity, corrections, brainstorming, short messages, and tool restraint , without turning every conversation into a settings panel.
One guided session. Zero generic presets. Your AI actually reads the room.
Public Professional v3.4 is a controlled public beta. Calibration state is ephemeral and local to the running process. It is not a hosted account service or a production SLA.
Run the proof, drop in a prompt, and compare Default AI vs ReadTheRoom side by side.
- Guided behavior calibration, tune real behavior, not tone sliders
- Explicit profile and context controls, you decide what the AI knows and how it acts
- Sandbox-style experimentation, test freely, nothing learns silently
- Reviewable calibration receipts, see exactly what changed and why
- Public-only runtime boundaries, loopback by default, no silent data collection
- Responsive desktop and mobile, calibrate anywhere
Requirements: Node.js 22 or newer.
npm test
npm startOpen http://127.0.0.1:8877/read-the-room-public-pro-v3-4/.
No dependency installation is required. The package uses Node's built-in runtime and test runner.
Most "AI personality" settings are a slider and a prayer. ReadTheRoom treats behavior calibration as a first-class problem: a guided session that produces a reviewable, repeatable behavior profile, so your assistant stops sounding like a default and starts sounding like yours.
The standalone server exposes only the public calibration application and its documented profile, artifact, archetype, apply, session, and health routes. Internal MAYA runtime files and private profiles are not included.
node --check scripts/read-the-room/readtheroomPublicServer.mjs
node --check scripts/read-the-room/readtheroomPolicy.js
node --check scripts/read-the-room/readtheroomCalibrationSession.js
node --check scripts/read-the-room/rtrToMemoryLane.mjs
npm testA ready-to-enable GitHub Actions template is included at docs/ci/verify.yml.example; local verification remains the release authority for this beta.
The release has also passed clean-extraction, browser, responsive-layout, session-isolation, malformed-request, and public-boundary audits. See PUBLIC_BETA_LIMITS.md for the precise claim boundary.
- Loopback-only server by default
- Ephemeral in-memory calibration sessions
- No account system or telemetry
- Hardened browser response headers
- Explicit route allowlist and traversal denial
Do not expose the development server directly to the internet. Report sensitive findings privately using SECURITY.md.
Completed calibration sessions can be sealed into a Memory Lane library — the same tamper-evident, chain-verified, searchable memory layer that powers MAYA and Hermes. A session that lives in a local JSON file today becomes a sealed, chain-linked block you can search, resume, and prove hasn't been altered.
node scripts/read-the-room/rtrToMemoryLane.mjs \
--store path/to/readtheroom-calibration-sessions.json \
--ml-base http://127.0.0.1:8770Deterministic and free: facts are written explicitly (no LLM extraction, no token cost). Re-running is idempotent — a session ledger (data/.rtr-ml-sync-ledger.json) prevents double-sealing. Use --dry-run to preview first. The bridge runs against any Memory Lane server (MEMORY_LANE_LIBRARY pointed at your own library).
This is the read side of the same loop: Memory Lane collects itself (write) and answers questions across sessions (read) — ReadTheRoom calibration history becomes durable, queryable memory instead of an editable local file.
ReadTheRoom Public Professional is distributed under the 2ndNatureAi Public Beta Evaluation License 1.0. See LICENSE.txt for the full terms.
Bundled font files remain governed by their own notices: OFL-Inter.txt and OFL-JetBrains-Mono.txt.
Copyright (c) 2026 2ndNatureAi.