9.9 KiB
@elizaos/plugin-contacts
Android address-book overlay app for elizaOS: provides a full-screen UI surface for browsing, searching, creating, and importing contacts, plus a read-only dynamic provider that injects address-book context into the agent planner.
Purpose / role
This plugin adds Android address-book capability to an Eliza agent. It ships two surfaces:
- A dynamic provider (
androidContacts) that reads up to 50 contacts from the device and injects them as planning context — scoped tocontactsandmessagingconversation contexts, gated toADMINrole sessions, cached per-turn. - A full-screen overlay app (
ContactsAppView) and one shipped GUI view declaration (ContactsView) registered via@elizaos/ui.
The plugin is Android-only (elizaos.app.androidOnly: true). The src/register.ts side-effect module skips registration on non-elizaOS runtimes. The /plugin export is the entry point for the elizaOS runtime adapter.
Plugin surface
Registered in appContactsPlugin (src/plugin.ts):
| Kind | Name | Description |
|---|---|---|
| Provider | androidContacts |
Read-only: fetches up to 50 contacts (id, displayName, phones, emails, starred) from @elizaos/capacitor-contacts and emits JSON context. Dynamic; contexts: contacts, messaging; roleGate: ADMIN; cacheScope: turn. |
| View | contacts |
GUI address-book view — ContactsView component, path /contacts. |
No actions, services, evaluators, events, or routes are registered.
Layout
src/
index.ts Public package entry — re-exports plugin, app, register, ui
plugin.ts appContactsPlugin definition (providers + views)
register.ts Side-effect: calls registerContactsApp() when isElizaOS()
ui.ts Re-exports ContactsAppView, contactsApp, registerContactsApp
providers/
contacts.ts androidContacts provider implementation
contacts.test.ts Vitest unit tests for the provider
components/
contacts-app.ts OverlayApp descriptor + registerContactsApp()
contacts-app.test.ts Tests for OverlayApp descriptor
contacts-view-bundle.ts View bundle registration helpers
contacts-contract.test.ts Contract tests for the overlay-app view surface
ContactsAppView.tsx Full-screen overlay UI (list / detail / new modes)
ContactsAppView.helpers.ts Helper utilities for ContactsAppView
ContactsAppView.interact.ts Exports interact(capability, params) for programmatic view actions
ContactsAppView.test.ts Tests for ContactsAppView
ContactsSpatialView.tsx Presentational spatial-primitives view
The ./plugin export (declared in package.json exports map) resolves to dist/plugin.js / dist/plugin.d.ts and is the entry the runtime adapter imports directly.
Commands
Only scripts that exist in this package's package.json:
bun run --cwd plugins/plugin-contacts typecheck # tsgo --noEmit
bun run --cwd plugins/plugin-contacts lint # biome check src/
bun run --cwd plugins/plugin-contacts test # vitest run
bun run --cwd plugins/plugin-contacts build # build:js + build:views + build:types
bun run --cwd plugins/plugin-contacts build:js # tsup (shared config)
bun run --cwd plugins/plugin-contacts build:views # vite build for overlay bundle
bun run --cwd plugins/plugin-contacts build:types # tsc declaration emit
bun run --cwd plugins/plugin-contacts clean # rm -rf dist
Config / env vars
This plugin reads no environment variables and has no settings keys. All address-book access goes through @elizaos/capacitor-contacts Contacts native API, which requires the Android READ_CONTACTS / WRITE_CONTACTS permissions to be granted at the OS level.
The provider limit is a hardcoded constant CONTACTS_PROVIDER_LIMIT = 50 in src/providers/contacts.ts.
How to extend
Add a provider: create src/providers/<name>.ts exporting a Provider object, then add it to the providers array in src/plugin.ts.
Add a view: define a new ViewDeclaration descriptor object in src/plugin.ts views array with a unique id + viewType. Add the corresponding React component to src/components/ContactsAppView.tsx or a new file, then re-export it from src/ui.ts and src/index.ts.
Add an action: the current design intentionally uses no actions (reads are providers; writes happen in the UI layer via the native Contacts API directly). If you add an action, import it in src/plugin.ts and add it to the actions array.
Conventions / gotchas
- Android-only.
isElizaOS()guard insrc/register.tsprevents the overlay app from registering on web/iOS/desktop. The provider will still be instantiated anywhere the plugin is loaded, butContacts.listContactswill throw on non-Android runtimes — the provider catches the error and returnscontactsAvailable: false. - No update or delete. The
@elizaos/capacitor-contactsnative plugin does not expose contact mutation beyond create and import. The detail panel is read-only; the "Edit" path was intentionally omitted. - In-app Call/Text linking. The detail view phone rows do not use a
tel:OS handoff. Each number renders "Call" and "Text" controls that dispatcheliza:navigate:viewwith{ viewId, viewPath, payload }for the in-app Phone and Messages views, pre-seeding the target through the generic navigation payload handoff. Email keeps itsmailto:anchor (there is no in-app email view). Do not reintroducetel:. - Provider roleGate.
roleGate: { minRole: "ADMIN" }means theandroidContactsprovider only fires in admin-role sessions. Do not change this without reviewing the address-book privacy model. - View interact() function.
src/components/ContactsAppView.interact.tsexportsinteract(capability, params)which handleslist-contacts,create-contact, andimport-vcardcapability strings for programmatic view actions. - Spatial view.
ContactsSpatialView.tsxis authored with the spatial-UI vocabulary and is purely presentational (snapshot + action callback) with no Capacitor runtime imports. - Views bundle. The overlay UI is built separately via
vite.config.views.tsintodist/views/bundle.js.bundlePathin the view descriptors points there. The tsup build (build:js) and the vite build (build:views) are independent steps. - Peer deps. React 19 and react-dom 19 are peer dependencies. The host app must provide them.
- See the root
AGENTS.mdfor repo-wide architecture rules, logging conventions, and git workflow.
⛔ NON-NEGOTIABLE — evidence, trajectories & real end-to-end tests
The binding, repo-wide standard is AGENTS.md. Read it. Nothing in this package is done until it is proven done — a reviewer must confirm it works without reading the code, from the artifacts you attach. This applies to every feature, fix, refactor, and chore here. "Tests pass" is not proof; "CI is green" is not proof.
- Record AND read model trajectories. Capture the actual inputs and outputs of the model
from a live LLM — not the deterministic proxy, not a mock: the prompt, the
providers/context, the raw model output, every tool/action call, and the result. Then open
the trajectory and review it by hand. A captured-but-unread trajectory is not evidence
(
packages/scenario-runner/bin/eliza-scenarios run <scenario> --report <out>). - Real, full-featured E2E — no larp. Every feature ships detailed end-to-end tests that drive the real path end to end. Not the happy "front door" only: cover error paths, edge/empty/invalid input, concurrency, roles/permissions, and adversarial input. A test that asserts against a mock/stub/fixture standing in for the thing under test does not count. If the real model/device/chain/connector/account is hard to reach, make it reachable — that is the work, not an excuse to mock. If the existing tests here are shallow or mocked, fixing them is part of your change.
- Screenshots + logs at every phase, plus a complete walkthrough video/run-through of
the entire feature or view, start to finish (
bun run test:e2e:record). - Manually review every artifact the change touches — never just the green check: client
logs (console + network), server logs (
[ClassName] …), the model trajectories in and out, before/after full-page screenshots, and the domain artifacts listed below for this package. - No residuals. No shortcuts. The goal is not "done" — it is everything done. Clear every blocker by the hard path: build the real architecture, stand up the real model/device/service, actually test it. Never leave a TODO, a stub, a stepping-stone, or a "follow-up." When unsure, research thoroughly, weigh the options, and ship the best, highest-effort, production-ready version. Keep going until every possibility is exhausted.
Artifacts → attached inline in the PR (MP4 video, JPG screenshots, logs in <details>); attach each evidence type or
explicitly mark it N/A with a reason — never leave it blank. If develop moved and changed
behavior, re-capture evidence; stale proof is worse than none.
Capture & manually review for this package — platform connector:
- A real (or sandbox-account) round-trip on the platform: inbound message → agent → outbound reply, captured as logs and a screenshot/recording of the actual conversation.
- The raw inbound event/webhook payload and the outbound API request/response, with IDs mapped correctly (
stringToUuid/createUniqueUuid). - Attachments, threads/replies, edits, multi-account, and rate-limit/error paths — not just a single text ping.
- The agent trajectory for the turn the connector drove.