KnowledgeOS

A local-first personal AI knowledge operating system.
Dump everything, structure it, search it semantically, traverse it as a graph, and let AI agents operate it — all entirely on your Mac, with no cloud dependency and no data leaving your machine.

Philosophy

Most knowledge tools are either too simple (note apps that can't reason) or too complex (enterprise wikis that require a team to maintain). KnowledgeOS is built on a different premise:

Your knowledge should be a living system, not a filing cabinet.

The design decisions all follow from this:

Local-first, always. Every byte lives under ~/KnowledgeOS/. No sync service, no cloud account required. You own your data unconditionally.
Agent-ready from the start. Claude, ChatGPT, Codex, and local LLMs can search, read, write, and link knowledge through a built-in MCP server — not as a bolt-on feature, but as a first-class access mode.
Structured but flexible. Everything is a typed object (KosObject) with a common base. Pages, sources, assets, claims, projects, and chats are all objects you can search, link, and reason over in the same interface.
Soft everything. No hard deletes. No permanent overwrites. Every agent write is logged with a before/after diff. You can always undo, audit, or restore.
Durable indexes, not canonical truth. Qdrant (vector) and Kùzu (graph) are re-buildable from Postgres + filesystem. Postgres is the operational source of truth. Indexes are caches.

The full product vision is in project-phases/IDEA-DRAFT.md.

Current Status

Phases 1–4, Phase 5 AI Assistant, Phase 6A/6B Chat Import, Phase 7A MCP Read/Search, Phase 7B MCP Write Tools, and Phase 8A Workspace Lite are complete. Phase 8 (multi-pane workspaces), Phase 9 (career memory), and the multilingual search hardening track are in progress or planned.

See PROGRESS.md for the canonical progress tracker.

Phase	Status
Phase 1 — Foundation	✅ Complete
Phase 2 — Sources & Rich Media	✅ Complete
Phase 3 — Search	✅ Complete
Phase 4 — Graph Lite	✅ Complete
Phase 5 — AI Assistant + Inbox/Triage	✅ Complete
Phase 6A — Chat Import Lite	✅ Complete
Phase 6B — Structured Chat Import	✅ Complete
Phase 7A — MCP Read/Search	✅ Complete
Phase 7B — MCP Write Tools	✅ Complete
Phase 8A — Workspace Lite (split pane)	✅ Complete
Hardening — Search Quality / Multilingual	🚧 Partial (ILIKE fallback only)
Phase 8 — Multi-Pane Workspaces	⬜ Planned
Phase 9 — Career & Project Memory	⬜ Planned
PHASE-PHONE-00 — Mobile Architecture & Contract	✅ Complete (docs only)
PHASE-PHONE-01A/B/C — Mobile Auth + Networking + iOS Scaffold	⬜ Planned
PHASE-PHONE-02A/B → 06 — iOS client, capture, AI, edit-lite, offline, device install	⬜ Planned

Test counts (as of 2026-05-15):

Frontend unit/component tests: 27 in apps/web/src/**/__tests__/
API integration + backend unit tests: 147 across tests/api/ and tests/unit/ (131 api + 16 unit)
Worker extractor tests: 19 in tests/worker/
MCP package tests: 35 in services/mcp/tests/
Playwright E2E tests: 10 in tests/e2e/specs/
Total: 238 tests across all local suites

See each phase's plan in project-phases/ for the full subtask spec.

Architecture

Browser (Next.js 14 + React + Tiptap)
  └─► FastAPI  /api/v1
        ├─► Postgres 16          ← operational source of truth
        │     objects, pages, edges, chunks, agent_runs, ingestion_jobs
        ├─► Redis                ← RQ job queue + cache
        ├─► ~/KnowledgeOS/library/  ← binary assets (SHA-256 content-addressed)
        ├─► Qdrant               ← vector index (re-buildable)
        └─► Kùzu                 ← graph index (re-buildable)

Python Workers (RQ)
  └─► ingestion → text extraction → chunking → embedding → graph sync → AI extraction

MCP Server (:8765)
  └─► read / search / write / ingest tools
        ↑
  Claude / ChatGPT / Codex / Cursor / local agents

Canonical source of truth: Postgres (structured data) + local filesystem (binary assets).
Qdrant and Kùzu are indexes. Treat them as caches; they can be rebuilt from Postgres at any time.
All host ports bind to 127.0.0.1 only. LAN access is opt-in.

Stack

Layer	Technology
Frontend	Next.js 14 + React 18 + Tailwind CSS
Editor	Tiptap / ProseMirror (stores content as JSON)
Backend API	FastAPI + SQLAlchemy 2.0 async
Background workers	Python + Redis + RQ
MCP server	Python MCP server over stdio (Phase 7A)
Operational DB	Postgres 16
Vector DB	Qdrant (local Docker)
Graph DB	Kùzu (embedded in worker process)
Queue / cache	Redis 7
Asset storage	Local filesystem, content-addressed by SHA-256
AI	OpenAI API (Phase 5+); Ollama local LLMs (later)
Deployment	Docker Compose on Mac
iPhone client (planned)	SwiftUI native iOS + XcodeGen (`apps/ios/`, Phase PHONE-01C onward)

Quick Start

Prerequisites

macOS, Docker Desktop, Node.js 20+, pnpm 10+, Python 3.12+, uv

Setup

# 1. Clone and enter the repo
git clone <repo-url> && cd agentic-knowledge-management

# 2. Create local directories + copy .env
bash scripts/setup.sh

# 3. Edit infra/.env as needed — all settings have working defaults
#    (SESSION_SECRET is optional; see docs/SECURITY.md for the session model)

# 4. Start all services
docker compose -f infra/docker-compose.yml up -d

# 5. Open the app
open http://localhost:3000

Register an account on first visit. All data stays local.

Simpler, non-engineer walkthrough: quickstart.md.

Docker troubleshooting

Symptom	What to do
`bind: address already in use` on 8001	Free `127.0.0.1:8001` (often a local `uvicorn` on that port). Stop it, then `docker compose -f infra/docker-compose.yml up -d` again.
Mount error / path contains `YOUR_USERNAME`	Set `LIBRARY_ROOT` in `infra/.env` to your real library path (under `~/KnowledgeOS/`). Run `bash scripts/setup.sh` (it patches a stale placeholder). If a bad volume already exists: `docker compose -f infra/docker-compose.yml down`, `docker volume rm infra_library-data`, then `up -d`.
Internal Server Error from the web UI / Next Module not found	Stale `node_modules` volume for `kos-web`: `docker compose -f infra/docker-compose.yml rm -sf web`, `docker volume rm infra_kos-web-node-modules`, `docker compose -f infra/docker-compose.yml up -d --build web`. If you use `pnpm dev` on the host with the API in Docker, add `apps/web/.env.local` from `apps/web/.env.local.example`. Confirm `docker ps` shows `kos-web` on port 3000.

Development

Frontend (apps/web)

pnpm dev          # Next.js dev server on :3000 with hot reload
pnpm typecheck    # tsc --noEmit (strict mode)
pnpm lint         # ESLint via next lint
pnpm build        # production build

Backend API (services/api)

cd services/api
uv run uvicorn app.main:app --reload --port 8000   # dev server
uv run ruff check .                                 # lint
uv run ruff format .                                # format
uv run python -c "from app.models import *; print('OK')"  # sanity check

Integration Tests

Tests run against a live knowledgeos_test Postgres database (auto-created and torn down per test). Requires Postgres running (via Docker or local).

cd tests
uv run pytest api/ unit/ -v                    # API integration + backend unit tests
uv run pytest worker/ -v                       # worker extractor tests
uv run pytest api/test_pages.py -v             # single file
uv run pytest api/test_pages.py::test_create_page  # single test

# MCP package tests (separate project)
cd ../services/mcp && uv run pytest tests/ -v   # MCP tool + config tests

End-to-end Tests

Playwright E2E tests live in tests/e2e and run against the local Compose stack. Use a sandbox library root so tests never write to the real ~/KnowledgeOS/library:

LIBRARY_ROOT=$PWD/tests/e2e/.tmp/library docker compose -f infra/docker-compose.yml up -d
pnpm test:e2e
pnpm --dir tests/e2e report

Testing Matrix

Layer	Command
Frontend unit/component	`make test-unit` or `pnpm test:web`
API + unit pytest	`make test-api`
Worker extractors	`make test-worker`
MCP package	`make test-mcp`
Playwright E2E	`make test-e2e` or `pnpm test:e2e`
Everything	`make test-all` or `pnpm test:all`

make test-e2e expects the Compose stack to be running with a sandbox library root:

LIBRARY_ROOT=$PWD/tests/e2e/.tmp/library docker compose -f infra/docker-compose.yml up -d --build
make test-e2e

Alembic Migrations

cd services/api
uv run alembic revision --autogenerate -m "description"
uv run alembic upgrade head

Infrastructure

docker compose -f infra/docker-compose.yml up -d      # start (background)
docker compose -f infra/docker-compose.yml up         # start (foreground logs)
docker compose -f infra/docker-compose.yml down       # stop
docker compose -f infra/docker-compose.yml down -v    # stop + wipe volumes (destructive)

Compose defaults include bind-mounted sources, Alembic before uvicorn, and optional demo seed. With seeding enabled, sign in as demo@example.com / demo-demo-demo unless overridden in infra/.env.

Local Ports

Docker Compose publishes services on loopback-only host ports:

Service	Host	In-container
Web	`127.0.0.1:3000`	`3000`
API	`127.0.0.1:8001`	`8000`
Postgres	`127.0.0.1:5433`	`5432`
Redis	`127.0.0.1:6379`	`6379`
Qdrant HTTP	`127.0.0.1:6333`	`6333`
Qdrant gRPC	`127.0.0.1:6334`	`6334`

Use http://127.0.0.1:8001 for host-side clients talking to the dockerized API. Use http://api:8000 only from inside the Docker network. If you run the API natively with uvicorn --port 8000, host-side clients should use http://127.0.0.1:8000.

Backup & Restore

Create a local backup with:

bash scripts/backup.sh

Backups are written to ~/KnowledgeOS/backups/<timestamp>/ and include a custom-format pg_dump, a compressed library tarball, and a best-effort Qdrant snapshot metadata file. Postgres plus ~/KnowledgeOS/library/ are canonical; Qdrant is a rebuildable search index.

To restore manually, stop the app, restore postgres.dump into a clean Postgres database with pg_restore, unpack library.tar.gz back under ~/KnowledgeOS/, then rebuild/reindex derived search data as needed. Do not rely on Qdrant snapshots as the only backup of user data.

File Layout

.
├── apps/web/              Next.js frontend
│   └── src/
│       ├── app/           App Router: (auth)/ and (app)/ route groups
│       ├── components/    UI components (editor, assets, auth, layout)
│       ├── lib/           API client, SWR hooks
│       └── types/         Shared TypeScript types
│
├── apps/ios/              SwiftUI iPhone client (planned — Phase PHONE-01C scaffolds it)
│   ├── project.yml        XcodeGen source of truth
│   └── KnowledgeOS/       App / Core / Features / Resources
│
├── services/api/          FastAPI backend
│   └── app/
│       ├── api/v1/        Route handlers (auth, objects, pages, assets, health)
│       ├── core/          Deps, security, library, storage utilities
│       ├── db/            Session, base model
│       ├── models/        SQLAlchemy ORM models
│       ├── schemas/        Pydantic request/response schemas
│       └── services/      Business logic layer
│
├── services/worker/       RQ background worker (Phase 2+) — run as `rq worker kos-ingest`
├── services/mcp/          MCP server, stdio transport (Phase 7A)
│
├── packages/
│   ├── shared-types/      Reserved (currently empty placeholder)
│   ├── schemas/           Reserved (currently empty placeholder)
│   └── prompts/           Reserved (currently empty placeholder)
│
├── infra/                 Docker Compose, Dockerfiles, .env.example
├── scripts/               setup.sh, run_tests.sh, backup.sh
├── tests/api/             Integration tests (pytest + httpx ASGI)
├── tests/unit/            Pure unit tests (parsers, URL safety)
├── docs/                  Architecture, data model, API, MCP, ingestion, security
└── project-phases/        Phase-by-phase implementation plans

Data Model

The `KosObject` pattern

Every entity in the system — pages, assets, notes, bookmarks, collections — is a row in the objects table with a kind discriminator. Specialized tables (pages, assets, etc.) extend it by foreign key.

objects (id, user_id, kind, title, tags[], metadata{}, is_pinned, is_archived, deleted_at)
  ├── pages     (content_json, content_text, word_count, version)
  ├── assets    (sha256, storage_path, content_type, size_bytes, status)
  └── edges     (source_id, target_id, kind: link/child/citation/related/...)

Future object types (sources, claims, projects, chats, concepts, tasks, workspaces) extend the same base. See docs/DATA_MODEL.md for the full schema.

Asset storage

Binary files are stored content-addressed under LIBRARY_ROOT:

~/KnowledgeOS/library/assets/<sha256[:2]>/<sha256>/original.<ext>

Write-once semantics: if the path already exists, the upload is a no-op (deduplication by hash).

Soft deletes

All user-owned records use deleted_at (nullable timestamp). Hard deletes are never performed. Trash is filtered out by deleted_at IS NULL in all standard queries.

Product Modules (Full Vision)

The complete product is built across 9 phases. Phases 1–6B, Phase 7A, and Phase 8A are done; remaining phases are planned.

Module	Description	Phase	Status
Wiki pages	Rich Tiptap editor, auto-save, word count, typed links	1	✅
Asset library	Upload, dedup, preview, gallery, full-screen modal	1	✅
Rich media sources	PDF text+thumbnail, YouTube oEmbed+transcript, web article, CSV preview	2	✅
Keyword + semantic search	Postgres FTS, Qdrant vector, hybrid reranking, Cmd+K modal	3	✅
Graph Lite	Typed edges, backlinks, related, ObjectPicker, LinkToModal, GraphPanel	4	✅
AI assistant + Inbox/Triage	Summarize, extract claims/tasks, suggest links, KB Q&A, triage inbox	5	✅
Chat import (raw)	ChatGPT/Claude/Markdown/plain text import → searchable chats	6A	✅
Chat structured import	AI summary, decisions, claims/tasks extracted with turn refs	6B	✅
MCP read/search server	stdio MCP with `search_objects`, `hybrid_search`, `get_*`, `get_related_objects`	7A	✅
Workspace Lite (split pane)	Open any object in a side pane from search/backlinks/related	8A	✅
MCP write tools	`create_page`, `update_page`, `create_edge`, `archive_object`, `ingest_url`, `ingest_file`	7B	✅
Multi-pane workspaces	Persistent layouts, drag-across-pane, workspace-scoped AI	8	⬜
Career/project memory	Project schema, evidence-linked resume bullets, STAR stories	9	⬜
iPhone client — contract	Mobile spec docs: MOBILE_APP, MOBILE_API_CONTRACT, MOBILE_NETWORKING	PHONE-00	✅
iPhone client — MVP	Bearer auth, SwiftUI scaffold, read/search/capture/AI on iOS Simulator + device	PHONE-01A/B/C → 03C	⬜ Planned

AI Features

Career memory: project records, AI-generated resume bullets & STAR stories — see AGENT_GUIDE.md.

Roadmap

Phase	Goal	Status
1 — Foundation	Docker, Postgres, auth, page CRUD, Tiptap editor, asset upload	Done
2 — Sources & Rich Media	PDF/YouTube/web/CSV ingestion, RQ worker, citation edges	Done
3 — Search	Chunking, Postgres FTS, Qdrant vectors, hybrid search, Cmd+K UI	Done
4 — Graph Lite	Typed edge API, backlinks, related objects from Postgres edges	Done
5 — AI Assistant + Inbox/Triage	AI sidebar, summarize/extract/suggest, KB Q&A, triage inbox	Done
6A — Chat Import Lite	Raw upload/paste of ChatGPT/Claude/Markdown/text exports	Done
6B — Structured Chat Import	AI summaries, extracted claims/tasks, turn-grounded graph links	Done
7A — MCP Read/Search	stdio MCP server, read-only tools, internal-token auth	Done
8A — Workspace Lite	Frontend split pane (no schema change)	Done
Hardening — Search Quality	Multilingual ILIKE fallback ✅; snippet sanitization, JP fixtures, debug UI, index-status endpoint pending	Partial
7B — MCP Write Tools	`create_page`, `update_page`, `create_edge`, `archive_object`, `ingest_url`, `ingest_file`	Done
8 — Multi-Pane Workspaces	Persistent layout engine, saved workspaces, workspace-scoped AI	Planned
9 — Career Memory	Project schema UI, resume bullet generator, STAR story generator	Planned
PHONE-00 — Mobile Architecture & Contract	MOBILE_APP, MOBILE_API_CONTRACT, MOBILE_NETWORKING docs	Done
PHONE-01A — Mobile Backend Auth & API	Bearer tokens, `/auth/mobile-login`, `/mobile/bootstrap`, fix `get_current_user` stub	Planned
PHONE-01B — Mac ↔ iPhone Networking	`infra/docker-compose.mobile.yml`, `scripts/mobile_network_check.sh`, ATS strategy	Planned
PHONE-01C — iOS App Scaffold	SwiftUI app + XcodeGen `project.yml`, Connect screen, health check	Planned
PHONE-02A/B — API Client & Simulator QA	Typed API client, Keychain auth store, simulator MCP smoke tests	Planned
PHONE-03A/B/C — Read & Search / Capture / AI MVP	Hybrid search, page/source/chat/project readers, capture, KB Q&A on iPhone	Planned
PHONE-04 — Edit-Lite	Title/tags/plain-body edits with Tiptap JSON safety	Planned
PHONE-05 — Offline Cache & Queue	Recent-object cache + outgoing note queue	Planned
PHONE-06 — Device Install & Private Release	Physical iPhone install via Xcode, TestFlight checklist	Planned

Each phase has a detailed spec in project-phases/.

MCP & Agent Access

Phases 7A + 7B are shipped. KnowledgeOS exposes a local stdio MCP server in services/mcp/ that external agents (Claude Desktop, Claude Code, Cursor, Codex) can spawn as a subprocess. The server talks to FastAPI over 127.0.0.1 using a shared MCP_INTERNAL_TOKEN.

Enable it by setting MCP_ENABLED=true and a random MCP_INTERNAL_TOKEN in infra/.env, then point your agent client at uv run --project services/mcp kos-mcp. For the dockerized API, keep MCP_API_BASE_URL=http://127.0.0.1:8001; for a native API run on port 8000, override it to http://127.0.0.1:8000.

To enable write tools, also set MCP_ALLOW_WRITE_TOOLS=true.

Read/search tools (Phase 7A — live): search_objects, hybrid_search, get_object, get_page, get_source, get_related_objects, answer_from_kb

Write tools (Phase 7B — live, requires MCP_ALLOW_WRITE_TOOLS=true):

create_page — create a new page
update_page — update title/content/tags (supports expected_version for optimistic locking)
create_edge — link two objects with a typed relationship
archive_object — soft-archive (reversible via restore endpoint)
ingest_url — ingest a URL as a new source (web/youtube)
ingest_file — ingest a local file from LIBRARY_ROOT

Safety invariants (always enforced):

Every write tool call creates an agent_runs audit row
Mutating tools (update_page, archive_object) create object_revisions rows with before/after snapshots
Rate-limited per agent identity: 60 writes/minute, 600 writes/hour
Soft-delete only — archive_object is reversible, no hard deletes through MCP
No arbitrary shell execution through any MCP tool
No file access outside ~/KnowledgeOS/
API keys and session secrets are never returned in tool outputs

See docs/MCP_TOOLS.md for the full tool spec and docs/SECURITY.md for the auth and audit design.

iPhone App (planned)

A native SwiftUI iPhone client lives alongside apps/web/ at apps/ios/ (scaffold lands in Phase PHONE-01C). It is a thin client — reads, searches, captures, and asks grounded questions through the same /api/v1 surface the browser uses. No second backend; no direct DB or filesystem access from the phone.

Phase PHONE-00 (docs-only) is complete. The contract is fixed; implementation phases are planned:

docs/MOBILE_APP.md — product spec, MVP scope, screen map, endpoints intentionally NOT exposed on mobile
docs/MOBILE_API_CONTRACT.md — bearer auth contract (proposed for Phase 01A), verified read-side endpoint table, error envelope, pagination
docs/MOBILE_NETWORKING.md — Simulator / LAN / Tailscale profiles, infra/docker-compose.mobile.yml design, ATS strategy

Implementation phases live under project-phases/PHASE-PHONE-*.md (PHONE-01A backend auth → PHONE-06 device install). The big picture is in project-phases/IDEA-iPHONE-APP.md.

How to open it in Xcode (once Phase PHONE-01C lands):

cd apps/ios
xcodegen generate                 # produces KnowledgeOS.xcodeproj
open KnowledgeOS.xcodeproj

TestFlight distribution is Phase PHONE-06.

Environment Variables

Copy infra/.env.example to infra/.env before starting Docker.

Variable	Default	Required
`SESSION_SECRET`	(not set)	No — reserved for future signed URLs
`POSTGRES_DB`	`knowledgeos`	No
`POSTGRES_USER`	`kos`	No
`POSTGRES_PASSWORD`	`kospass`	No
`LIBRARY_ROOT`	`~/KnowledgeOS/library`	No
`OPENAI_API_KEY`	(not set)	Phase 5+

Keyboard Shortcuts

Press ? anywhere in the app (outside a text field) to open the keyboard shortcut overlay.

Key	Action
`⌘K`	Open search
`⌘\`	Collapse / expand sidebar
`?`	Show shortcuts
`j / ↓`	Move highlight down (lists)
`k / ↑`	Move highlight up (lists)
`Enter`	Open highlighted item (lists)
`Backspace`	Soft-delete highlighted item (lists)

See docs/SHORTCUTS.md for the full reference.

For AI Agents

If you are a coding agent (Claude, Codex, Cursor) working in this repo:

Read CLAUDE.md first — it is the operating contract for agents.
Read docs/AGENT_GUIDE.md for data model rules and agent write constraints.
Postgres is the source of truth. Do not mutate DB directly — go through the API or typed service functions.
Every write must create an agent_runs audit row.
Qdrant and Kùzu are indexes. They are re-buildable; never treat them as canonical.
Soft-delete only. deleted_at is the pattern. Never call DELETE on user data.

Documentation

File	Contents
`docs/ARCHITECTURE.md`	System diagram, service boundaries, sequence flows
`docs/DATA_MODEL.md`	Full Postgres schema, object types, edge types
`docs/API.md`	REST API reference
`docs/MCP_TOOLS.md`	MCP tool spec, resources, prompts
`docs/INGESTION.md`	Ingestion pipeline stages, job types, media handling
`docs/SECURITY.md`	Auth, sessions, audit log, MCP safety model
`docs/AGENT_GUIDE.md`	Rules for AI agents writing to this system
`project-phases/IDEA-DRAFT.md`	Original full product spec
`project-phases/PHASE-1-FOUNDATION.md`	Phase 1 subtask spec
`docs/MOBILE_APP.md`	iPhone app product spec — MVP scope, screen map, endpoint exclusion list
`docs/MOBILE_API_CONTRACT.md`	iPhone bearer-auth contract + verified read-side endpoint table
`docs/MOBILE_NETWORKING.md`	Simulator / LAN / Tailscale profiles + ATS strategy
`project-phases/IDEA-iPHONE-APP.md`	Canonical iPhone-track concept and phase plan

Name		Name	Last commit message	Last commit date
Latest commit History 404 Commits
.github/workflows		.github/workflows
.tmp/mobile-qa		.tmp/mobile-qa
apps		apps
docs		docs
infra		infra
packages		packages
project-phases		project-phases
scripts		scripts
services		services
tests		tests
.dockerignore		.dockerignore
.editorconfig		.editorconfig
.gitignore		.gitignore
AGENTS.md		AGENTS.md
CLAUDE.md		CLAUDE.md
Makefile		Makefile
PROGRESS.md		PROGRESS.md
README.md		README.md
package.json		package.json
pnpm-lock.yaml		pnpm-lock.yaml
pnpm-workspace.yaml		pnpm-workspace.yaml
pyproject.toml		pyproject.toml
quickstart.md		quickstart.md
uv.lock		uv.lock
uv.toml		uv.toml

Folders and files

Latest commit

History

Repository files navigation

KnowledgeOS

Philosophy

Current Status

Architecture

Stack

Quick Start

Prerequisites

Setup

Docker troubleshooting

Development

Frontend (apps/web)

Backend API (services/api)

Integration Tests

End-to-end Tests

Testing Matrix

Alembic Migrations

Infrastructure

Local Ports

Backup & Restore

File Layout

Data Model

The KosObject pattern

Asset storage

Soft deletes

Product Modules (Full Vision)

AI Features

Roadmap

MCP & Agent Access

iPhone App (planned)

Environment Variables

Keyboard Shortcuts

For AI Agents

Documentation

About

Resources

Security policy

Uh oh!

Stars

Watchers

Forks

Releases

Packages 0

Uh oh!

Contributors

Uh oh!

Languages

The `KosObject` pattern

Packages