The LQ.AI Atlas LQ.AI's documentation, bound to the code it describes
234 documents

The LQ.AI Atlas

LQ.AI's documentation, bound to the code it describes

Journeys

First hour with LQ.AI

The orientation path: what LQ.AI is and why, how to run it, how decisions are routed, the system shape, and how to tell what's real.

7 stops →

The Inference Gateway: the security boundary

The component that holds the keys and guards the egress: its pipeline, the threat model around it, what anonymization does to outbound prompts, how model routing stays inspectable, and where the inference-tier boundary sits.

6 stops →

Document review & citations, end to end

How a contract becomes reviewed, cited, and trustworthy: the verification cascade, where citations are stored, the playbook and tabular surfaces that ride on it, and the honest edges where verification stops.

5 stops →

The Autonomous Layer (M4)

The opt-in background executor and the discipline that makes unattended agency acceptable: its design lineage, the five-phase walk, the single chokepoint, the brakes, the honest caveats, and what shipped.

6 stops →

Getting Started

The front door. What LQ.AI is, how to stand it up, and the agent-orientation file that routes every implementation decision. Start with README, run it via QUICKSTART, then keep CLAUDE.md open as you work.

Product & Vision

The canonical specification and the live work list it generates. The PRD is the reference of record (transparency as a founding principle, the five-tier inference spectrum, the DE-XXX backlog); HONEST-STATE is its conscience — what actually ships today vs. what's plumbing — and ROADMAP turns the gap into sized, tagged contributor tasks.

  • Honest State Whenever you need the truth about what works today vs. what's plumbing — with the file or test that proves it.
  • LQ.AI Open Roadmap When you want to pick up open work sized and tagged to your background.
  • LQ.AI — Product Requirements Document The reference of record for any product, capability, or architectural decision — jump to the relevant § via its table of contents.

Architecture & Decisions

Why the system is shaped this way: the self-hosted topology, the single gateway security boundary, the Postgres 16 + pgvector schema, the observability surface, the opt-in autonomous executor — and the 13 ADRs that record each load-bearing decision and the tradeoff it accepted.

Capabilities & Services

The features a legal team actually touches and the services that run them: the four-stage Citation Engine, Playbooks, Tabular Review, the Word add-in, the Slack/Teams intake bridges, and the README for each service (api, gateway, word-addin, the two bridges). The docs are honest about what's shipped vs. scaffold.

  • Citation Engine You're touching citation verification, the message_citations table, ensemble judging, or the anonymization skip boundary.
  • LQ.AI Backend API You need the backend's local run/test commands or health endpoint.
  • LQ.AI Inference Gateway You need the gateway's run/test commands, its endpoint paths, or where its config schema lives.
  • LQ.AI Slack Bridge (M3-D1) You need the Slack bridge's env config, run command, manifest scopes, or token-handling posture.
  • LQ.AI Teams Bridge (M3-D3) You need the Teams bridge's env config, the Azure AD registration steps, or why Teams has no per-tenant encryption.
  • LQ.AI Word add-in You need to build, validate, run-in-Word, or sideload the add-in, or want the per-surface roadmap.
  • Playbooks You're working on playbook execution, the Easy Playbook pipeline, or the seeded built-ins.
  • Slack / Teams Light Intake Bridges You're working on the Slack/Teams OAuth bridges or need their honest verification state before claiming they work.
  • Tabular / Multi-Document Review You're working on the tabular grid, table-mode skills, cell extraction, or export.
  • Word Add-In You're touching Word add-in install/manifest, the OAuth dialog, or the version handshake — or assessing what it actually does.

Security

The security posture as inspectable artifacts: the STRIDE threat model, the Presidio anonymization layer, append-only audit logging, the cryptographic inventory, encrypted provider keys, the Greenwood boundary-registers map, and the disclosure policy. The Inference Gateway is the trust boundary throughout.

Skills

Skills are the canonical artifact of value — open, attorney-attested legal work product, not hidden prompts. This category holds the authoring guide; the 14 starter skills themselves are indexed for search and deep-link into the sibling Skills Explorer for their curated, diagrammed experience.

Contributing & Governance

How to contribute and how the project is governed: the engineering PR path (DCO, ruff/mypy, the test taxonomy), the Code of Conduct, the upstream-skill provenance ledger, the skill-acceptance-testing framework, and the eight short-cycle "easiest contributions" mini-PRDs, each a pre-written maintainer decision.

Operations & Compliance

Running, deploying, and selling the thing: the optional OpenTelemetry observability recipes, the procurement-readiness pack (SIG Lite / CAIQ), the compliance-alignment stubs (SOC 2 / ISO / GDPR / HIPAA / FedRAMP), and the upstream-research analysis behind the skill library.

Developer Tools

The documentation tooling itself — the static Repository + Skills Explorer that ships under docs/explore. (This Atlas lives alongside it and is not indexed in its own corpus.)

Planning & History

The working record: milestone implementation plans, dated session handoffs, the superpowers plan/spec set, the LQVern (M4) design and handoff docs, and the sample NDAs/MSAs used as fixtures. Indexed for search; not curated or drift-checked, because handoffs reference code as it stood that day.