# Yumi

A secure AI workspace — frontier models, persistent memory, personal & organizational
knowledge banks, workflow execution, and integrations across email, calendar, files, cloud
storage, and business apps. One **AI colleague** that remembers you, understands your context,
drafts content, performs tasks, and executes workflows — across desktop, web, and mobile,
sharing the same memory and sessions. Data is protected through encryption, hardware-backed
security, and (planned) secure enclaves, so you can confidently work with sensitive client and
business information while keeping full context.

> **Status (2026-06-21): consolidation workspace.** This directory gathers the existing
> baobab / MMD "Cowork" stack, the LLM router, and the Hindsight memory system into one place,
> **verbatim**. Rebranding to Yumi, full SSO, and deeper unification are later phases.
> See [docs/roadmap.md](docs/roadmap.md).

## What's here

| Path | Role |
|---|---|
| `apps/open-cowork-mmd` | Electron desktop app (macOS / Windows), local SQLite cache |
| `apps/mmd-cowork-mobile` | Web app + mobile PWA — **hosts the shared `/api/conversations` API** |
| `services/mmd-cowork-core` | `@mmd/cowork-core` — shared engine (conversation store, auth, agent tools, connector wrappers) |
| `services/mmd-cowork-config` | `@mmd/cowork-config` — shared static config (MCP servers, skills, surfaces, credential refs) |
| `services/mmd-cowork-m365` | Microsoft 365 manifests + tenant packaging |
| `services/mmd-cowork-office` | Outlook / Teams add-in runtime flavor |
| `services/mmd-llm-gateway` | LiteLLM router — the single model endpoint for every surface (`llm.baobab-ts.com`) |
| `services/hindsight` | Memory system (MIT, from `vectorize-io/hindsight`) — durable agent memory + knowledge banks |

Each top-level subdir is **its own repository** with its own history (embedded git credentials
were scrubbed on copy). Provenance: [PROVENANCE.md](PROVENANCE.md).
Architecture: [docs/architecture.md](docs/architecture.md).

## Quickstart

- **LLM router** — `cd services/mmd-llm-gateway && litellm --config config/config.yaml --port 4000`
  (needs `ZAI_API_KEY`; aliases `glm-5.2`, `claude-sonnet-4-6`, `claude-haiku-4-5`, `claude-opus-4-7` → Z.ai GLM-5.2)
- **Memory** — Hindsight via Docker in `services/hindsight/src/docker/`
- **Surfaces** — require SSO + external connectors; see [config/sso/README.md](config/sso/README.md)

Full local-run path and external dependencies: [docs/architecture.md](docs/architecture.md).

## Vision & roadmap

Subscription AI colleague; eventually GPU-as-a-service with enclave TEE, voice, and
SideQuest / Telepathy integration. Phases and the (intentionally deferred) MMD→Yumi rebrand
are in [docs/roadmap.md](docs/roadmap.md).

## Open items for Ben

- **Sharing with Shad:** channel not yet confirmed — push to Gitea vs email/Teams, and whether
  Shad is a push collaborator or reviewer. (This README is the shareable brief.)
- **SSO:** Google / Apple / Microsoft provider app registrations need Ben's interactive console
  access (an agent cannot create them). Runbook: [config/sso/README.md](config/sso/README.md).