Lattice AI

Your model is the voice you use today. Your Brain is the asset you keep.

모델은 갈아타도, 내 지식은 내 컴퓨터에 남는 로컬 우선 AI 브레인.

Chat, files, folders, notes, and web pages all flow into one durable knowledge graph on your computer. Any model — local MLX or cloud — can speak with that memory. Nothing leaves your machine without explicit consent.

대화·파일·폴더·웹페이지가 전부 내 컴퓨터 안의 지식 그래프로 쌓이고, 어떤 모델이든 그 기억을 이어받아 대화합니다.

What You Can Do

See your Brain's story in time — a growth curve, an activity heatmap, and each day's story, rewindable to any past moment	Chat with a Brain that remembers — every conversation grows durable, source-linked memory
See how knowledge connects — a real relationship graph, not a file list	Capture anything — files, whole folders, notes, screenshots, web pages
Automate with review — agent changes become proposals you approve first	Pick a model in one click — recommended local models for your hardware
Watch a file become memory — three named steps, not a pipeline diagram	Say how much it may do alone — one dial in plain words; dangerous actions stay blocked either way

Why Lattice AI

• Own your memory — knowledge lives in a local SQLite Brain you can back up, export, inspect, and restore (.latticebrain encrypted archive).
• Model-independent — switch between local MLX models and cloud models without rebuilding context from zero.
• Honest by design — the Brain tells you when retrieval context is limited, when captured pages extracted poorly, and when the vector index is catching up.
• Safe automation — automations are consent-first drafts; edits to existing content always pass through a reviewable proposal with a diff.

매번 AI에게 프로젝트 맥락을 다시 설명하고 있다면, 지식이 여러 서비스에 흩어져 있다면, 그 기억을 특정 회사가 아니라 내가 소유하고 싶다면 — Lattice AI가 그 브레인입니다.

Quick Start

pip install ltcai        # or: npm install -g ltcai
LTCAI                    # then open http://127.0.0.1:4825/app

Apple Silicon local models: pip install "ltcai[local]". Desktop app (Tauri) ships as a dmg on each GitHub Release.

First-run flow — wake the Brain, pick the owner, load a recommended model:

Screenshot index and capture notes: output/release/v11.5.2/SCREENSHOT_INDEX.md

Current Release

The current release is 11.5.2 — Tight Ship:

A cleanup-and-correctness release built from three audits of the settled 11.5.1 tree — a Rust↔Python duplication map, an arc-level dead-code and test audit, and a live front-door parity sweep across 192 endpoints (gateway vs. direct worker). The duplication map's verdict was that the boundary itself is sound: zero cross-boundary deletions, because every twin is either the live Python-direct surface or a copy pinned by a parity golden. The real duplication was inside each language, and that is what this release removes.

• About 1,100 lines gone. Six moved-module shims, a multimodal streaming seam that was never wired, roughly 27 zero-caller symbols whose only callers were their own tests, an interface method (metadata_for) no consumer ever called, a dead measurement script, and the legacy Electron shell — superseded by Tauri, yet still shipping inside the npm tarball. Every deletion came with per-assertion test surgery so the 100% line and branch floor never moved, and a new guard forbids the sys.modules[__name__] shim pattern from coming back. The 183 coverage work-package test files were all verified load-bearing (ten of them are the sole owners of 79 branch arcs); seven coverage-redundant feature tests were deliberately kept for their regression value.
• One home per thing. The byte-identical embedder pair is single-sourced on the golden-pinned copy — silent vector drift between two write paths was the failure mode. The workspace selector's four verbatim re-derivations now use the canonical rule, an intended behaviour change: chat, agent, upload, computer-use and admin 403 on disagreement where the copies silently preferred the header. sha256, SSE frame builders, the data-dir default, the mode services and the module-importability probe each collapsed to one implementation, and seven byte-identical Rust copies were consolidated (including a wholesale duplicate of clock.rs).
• The front door, fixed with live before/after proof. Proxied redirects now pass through with Set-Cookie and Location intact — the invite gate was a hard dead-end, SSO login silently never authenticated, and twelve deep links lost their fragment. Absolute Location values naming the internal worker are rewritten to the gateway origin. The native /rust/* and /host/status|jobs lanes are posture-gated fail-closed: they used to serve the whole graph unauthenticated while the worker required auth. X-Forwarded-For/Proto/ Host cross the hop and are honoured only from a loopback or listed trusted-proxy peer, which fixes --no-spawn CSRF refusals and the invite links, notifications and SSO URLs that named the internal worker port. The supervisor injects CORS origins alongside the CSRF ones, stale ws:// CSP entries are gone, and the Tauri shell no longer navigates to a dead origin when the gateway fails to bind.
• What the audits surfaced. POST /api/search/graph was allow-listed but unreachable until now; GET /api/ingestion/multimodal was documented but unwired. Two golden families were added: a recent_chat family pinning the build_recent_chat_context the live /chat path actually calls — it caught a real divergence, where Python's limit=0 tail slice keeps everything and Rust returned empty (Python is the reference; Rust was fixed) — plus document_targets and agent_profiles helper goldens, 97 rows. The corpus stands at 251 golden files.
• The boundaries, stated plainly. The native lanes are an open-posture, single-local-owner surface (a closed or unknown posture answers 401). The recent-chat context seam is still owned by the ad-hoc prepends in /chat rather than the assembler — wiring it changes the shape of a live prompt, which is a minor-release item, not something to ship quietly here. workspace_scope_from_request itself remains lenient for its two remaining callers. And the proxy hop deliberately carries no request timeout: a blanket timeout would kill long-lived SSE.

The floor holds: 7,022 Python tests (11 skipped) at 100.00% statement and branch coverage — 40,307 statements, 10,970 branches, fail_under=100 with the pragma budget untouched — 1,761 frontend tests across 101 files at 100% on all four metrics, and 760 Rust workspace tests, with the OpenAPI schema regenerated at 421 paths and the whole suite re-verified in a fresh-resolve Python 3.11 environment.

Release notes: RELEASE.md · Full history: docs/CHANGELOG.md

Expected artifacts for 11.5.2 release must use exact filenames:

• dist/ltcai-11.5.2-py3-none-any.whl
• dist/ltcai-11.5.2.tar.gz
• ltcai-11.5.2.tgz
• dist/ltcai-11.5.2.vsix
• src-tauri/target/release/bundle/dmg/Lattice AI_11.5.2_aarch64.dmg

Do not use wildcard artifact uploads. Package registry publishing remains owner-run.

Architecture At A Glance

FastAPI on localhost is the source of truth; the React/Vite frontend and the Tauri desktop shell sit on top; the independent lattice_brain package owns the graph, memory, ingestion, and portability. Local-first by default — cloud calls, downloads, Telegram, and update checks are opt-in.

See ARCHITECTURE.md for details and docs/DEVELOPMENT.md for the developer workflow (npm install && npm run dev, validation via npm run lint, npm run test:unit, npm run test:visual).

Known Limitations

• External package registries are owner-published and can lag behind GitHub.
• PostgreSQL/pgvector is optional scale/migration tooling. SQLite remains the live local Brain store in 10.3.0.
• Docker, model downloads, cloud model calls, Telegram, Brain Network, and update checks require explicit user action.
• Conversation does not fabricate answers when no model is loaded. Agent and workflow simulation without a loaded LLM is deterministic and LLM-free (it does not call a model) — labeled as such, never presented as autonomous model success.
• Some backend-generated messages (for example the Postgres DSN notice) are produced server-side in English and are shown as-is; server-side i18n is not part of 10.3.0.

Release History

Per-release details: RELEASE_NOTES.md

Documentation

• docs/WHY_LATTICE.md — product philosophy
• docs/TRUST_MODEL.md — local-first trust model
• PRIVACY.md — privacy and external communication policy
• FEATURE_STATUS.md — feature status and limitations
• SECURITY.md — security posture
• RELEASE.md — release guide and notes

License

MIT. See LICENSE.