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. The default is local. Cloud is optional — an
OAuth-authenticated CLI (agy / grok) or an API key you configure — and
nothing leaves your machine without explicit consent.
대화·파일·폴더·웹페이지가 전부 내 컴퓨터 안의 지식 그래프로 쌓입니다. 기본은 로컬이고, 클라우드는 선택입니다 (OAuth CLI 지원).
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 (
.latticebrainencrypted 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.
- Runs on the model your machine can hold — the loop measures what a model can actually produce and adapts: a model too small to emit a tool call is walked through numbered choices instead, with the same permission gates, snapshots and sanitizers as every other profile.
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/v12.1.0/SCREENSHOT_INDEX.md
Current Release
The current release is 12.1.0 — Fast Path:
12.0.0 opened the house. This release makes remembering the house faster:
folder ingest and vault-watch overlap parse/extract/embed across files,
one /worker/embed call carries the document plus its chunks, rebuild
hashes outside the SQLite write lock, and a Brain with 512+ vectors tries
hnsw+rescore when a worker sidecar is bound. The door is still 422
operations / 41 families over a 20-route worker. Cloud is still
optional and the default is still local.
- Ingest is a pipeline, not a queue of one. Up to four files parse and
embed at once. Unchanged files still skip on fingerprint and never
enter the pool. Writes still go through the single
GraphWriter. - One embed round-trip per note. The document vector and every chunk
share a
textsbody; extract runs beside it, not in front of it; the native incremental index files those vectors instead of hashing them again. - Search stays exact until it does not need to. Env default is still
brute. At 512 vectors with a bound sidecar, search asks HNSW first and falls back to the exact scan with the originalbruteblock if the sidecar cannot answer.
What this release does not close — small-model content quality, the
mock-only api_key path, hash embeddings as the no-download fallback,
watch never deleting on its own, the ad-hoc-signed dmg — is listed in
RELEASE_NOTES_v12.1.0.md.
Expected artifacts for 12.1.0 release must use exact filenames:
dist/ltcai-12.1.0-py3-none-any.whldist/ltcai-12.1.0.tar.gzltcai-12.1.0.tgzdist/ltcai-12.1.0.vsixsrc-tauri/target/release/bundle/dmg/Lattice AI_12.1.0_aarch64.dmg
Do not use wildcard artifact uploads. Package registry publishing remains owner-run.
Release notes: RELEASE.md · Full history: docs/CHANGELOG.md
Architecture At A Glance
One Rust server on localhost is the source of truth: lattice-host answers
every product route, owns every write to the Brain, and supervises a Python
AI worker it reaches over loopback for the things a model does — inference,
embedding, extraction, parsing, rendering, speech-to-text. The React/Vite
frontend and the Tauri desktop shell sit on top of that one door. Local-first
by default — cloud is optional (OAuth CLI or an API key you configure),
and downloads, Brain Network, and update checks are opt-in.
See ARCHITECTURE.md for details and
docs/DEVELOPMENT.md for the developer workflow
(npm start / bin/ltcai.js is the product; npm run dev is the
20-route worker). Each of the two largest crates carries its own domain
map — rust/lattice-agent/ARCHITECTURE.md
and rust/lattice-platform/ARCHITECTURE.md
— and open gaps are tracked in docs/ROADMAP.md.
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.
- SQLite is the live local Brain store. The optional PostgreSQL/pgvector migration tooling is not part of the 12.1.0 worker.
- Docker, model downloads, cloud model calls, Brain Network, and update checks
require explicit user action. Cloud is optional:
cli_oauth(agy/grok) was live-checked at zero billing; theapi_keypath is mock-verified only. - The Telegram bridge was removed in 11.6.0 — it lived in the platform code that became the AI worker. SSO/OIDC login and callback flows were removed with it; the configuration surface remains and password login is native.
- 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.
POST /worker/render/pdfships working out of the box:reportlabis a required dependency as of 11.6.0 (it used to be an undeclared lazy import that raised a 500;ltcai[pdf]remains as an empty alias for older install instructions).- The multimodal image and video analysis functions have no HTTP door as of 11.8.0: their only route wrapped a native image ingest that was never built. The observation code stays in Brain Core under unit test, and its module header says so.
- The Python coverage gate is a line floor of 90 since 11.8.0 (the 100%-lines-and-branches gate was removed). The enforced claim is the floor, not whatever a given run measures.
- Small-model content quality is gated honestly. With the
guidedprofile even a 0.5B model writes the requested file and reaches DONE; a weak summary still fails the critic and the run ends FAILED/NEEDS_REVIEW rather than claiming success. - Vector search env still defaults to
brute— exact and byte-compatible on a small Brain. At 512+ vectors with a bound worker sidecar it trieshnsw+rescorefirst and falls back to the exact scan if the sidecar cannot answer.LATTICEAI_VECTOR_INDEX=hnswis still the explicit opt-in. - Watch never deletes on its own: a vanished file is reported, and cleanup runs only through the confirmed folder-prune flow.
- The macOS dmg is ad-hoc signed — effectively unsigned — so first launch needs the usual Gatekeeper step.
- Pointer-control tools execute in the worker and are installed with
pip install "ltcai[pointer]". Remaining honest gaps (open_keyspending-only, no Self-Model refiner,delete_nodeleavingPART_OF, review events silent without an owner, KG-api ingest text-only, two store cycles per review mutation) are listed with their reasons in RELEASE_NOTES_v12.1.0.md and prioritized in docs/ROADMAP.md.
Release History
Public history starts at 11.0.0. 11.6.0 rebuilt the product server in Rust and
reduced the Python package to an AI worker, so a 10.x or 9.x install is a
different program; SECURITY.md supports only 11.x, and this table states the
same boundary. Earlier notes stay in the tree as RELEASE_NOTES_v*.md files.
| Version | Theme |
|---|---|
| 12.1.0 | Fast Path |
| 12.0.0 | Open House |
| 11.9.0 | Working Order |
| 11.8.0 | Travel Light |
| 11.7.0 | Clean Sweep |
| 11.6.0 | One Door |
| 11.5.2 | Tight Ship |
| 11.5.1 | Rust Full Loop |
| 11.5.0 | Rust Complete |
| 11.4.0 | Rust Foundation |
| 11.3.0 | Time Remembers |
| 11.2.0 | All Systems On |
| 11.1.0 | Product Intelligence |
| 11.0.1 | Both Branches |
| 11.0.0 | Full Measure |
Per-release details: RELEASE_NOTES.md
1 maintainer
Download files
Download the file for your platform. If you're not sure which to choose, learn more about installing packages.
Source Distribution
Built Distribution
Filter files by name, interpreter, ABI, and platform.
If you're not sure about the file name format, learn more about wheel file names.
Copy a direct link to the current filters
File details
Details for the file ltcai-12.1.0.tar.gz.
File metadata
- Download URL: ltcai-12.1.0.tar.gz
- Upload date:
- Size: 2.3 MB
- Tags: Source
- Uploaded using Trusted Publishing? No
- Uploaded via:
twine/7.0.0 CPython/3.11.16
File hashes
| Algorithm | Hash digest | |
|---|---|---|
| SHA256 |
e5318b5e0a6f84bd5e6c74a84840757cbe137025cb7b40480108f0b0701e3c42
|
|
| MD5 |
6729cb31097e5905f4a0c14c229bad2b
|
|
| BLAKE2b-256 |
8427cfbf8f1f72cd5d6539663eb641137894a8f1dae7898765fc58b7eb7cbe49
|
File details
Details for the file ltcai-12.1.0-py3-none-any.whl.
File metadata
- Download URL: ltcai-12.1.0-py3-none-any.whl
- Upload date:
- Size: 2.0 MB
- Tags: Python 3
- Uploaded using Trusted Publishing? No
- Uploaded via:
twine/7.0.0 CPython/3.11.16
File hashes
| Algorithm | Hash digest | |
|---|---|---|
| SHA256 |
bc7ac108fc057f0e8604f17b7767384ad762152871e6a1bcafa4cf0e13ae5cef
|
|
| MD5 |
80ae0165fd77bbebbf62734d8f5ee821
|
|
| BLAKE2b-256 |
1bc1bea1023be398d037d4956e0adbd8cd8d970124fe4b2c224f3900c5d3b129
|
