7.9 KiB
@elizaos/plugin-wifi
Android-only overlay app that lets an Eliza agent scan, inspect, and connect to nearby Wi-Fi networks.
Purpose / role
Adds a Wi-Fi management surface to the elizaOS mobile agent on Android. It registers a wifiNetworks provider that injects nearby network context into the agent's planner, and a full-screen overlay UI (WifiAppView) that the user can open from the app catalog. The plugin is opt-in: it is only registered in the overlay app catalog when isElizaOS() returns true (i.e., running inside the elizaOS Android host). On all other platforms (iOS, desktop, web) the side-effect entry leaves the overlay app catalog unchanged.
Plugin surface
The /plugin export (src/plugin.ts) registers:
| Kind | Name | Description |
|---|---|---|
| Provider | wifiNetworks |
Dynamic, read-only nearby Wi-Fi networks (ssid, bssid, rssi, frequency, secured). Context gate: system. Cache scope: turn. Calls @elizaos/capacitor-wifi WiFi.listAvailableNetworks. |
No actions, evaluators, routes, or events are registered.
The overlay UI surface (registered via src/register.ts side-effect):
| Export | Description |
|---|---|
wifiApp |
OverlayApp descriptor (name, displayName, category: "system", androidOnly: true). |
registerWifiApp() |
Registers wifiApp with @elizaos/ui's overlay app registry. Called automatically on elizaOS Android. |
WifiAppView |
React component. Full-screen overlay: shows connected network, scans for nearby networks, connects/disconnects with optional password entry. |
Layout
src/
index.ts Public barrel — re-exports everything below
plugin.ts appWifiPlugin: Plugin — registers wifiNetworks provider
register.ts Side-effect entry — calls registerWifiApp() if isElizaOS()
ui.ts UI barrel — re-exports WifiAppView + wifi-app helpers
providers/
networks.ts wifiNetworksProvider — calls WiFi.listAvailableNetworks, limit 25
components/
wifi-app.ts wifiApp OverlayApp descriptor + registerWifiApp()
WifiAppView.tsx Full-screen React overlay UI (scan, connect, disconnect)
assets/
hero.png App catalog hero image
Commands
bun run --cwd plugins/plugin-wifi typecheck # tsgo type-check only (no emit)
bun run --cwd plugins/plugin-wifi lint # biome check src/
bun run --cwd plugins/plugin-wifi test # vitest run
bun run --cwd plugins/plugin-wifi build # tsup + tsc declarations → dist/
bun run --cwd plugins/plugin-wifi clean # rm -rf dist
Config / env vars
No env vars or settings keys. The plugin reads no process environment at runtime. @elizaos/capacitor-wifi talks directly to the Android WifiManager via Capacitor; Android ACCESS_FINE_LOCATION permission must be granted at the OS level for scans to succeed.
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. Re-export it from src/index.ts.
Add an action: Create src/actions/<name>.ts exporting an Action object. Add an actions array to appWifiPlugin in src/plugin.ts and push the new action into it. Re-export from src/index.ts.
Add a service: Create src/services/<name>.ts extending Service. Register it in appWifiPlugin.services. Ensure it is exported from src/index.ts.
Conventions / gotchas
- Android-only.
WifiAppViewandregisterWifiApp()are safe to import on non-Android platforms but@elizaos/capacitor-wifimethods will reject or return empty results everywhere except Android. Theregister.tsentry guards registration behindisElizaOS(). - No server routes.
WifiAppViewowns all its data by calling the Capacitor plugin directly; there is no backend API involved. - Scan limit.
wifiNetworksProvidercaps at 25 networks;WifiAppViewcaps its own scan at 50. Keep these consistent if raising the limit. - Location permission. Android requires
ACCESS_FINE_LOCATIONforWifiManager.startScan. If the permission is denied, scans succeed silently with an empty list or throw; the provider maps errors towifiNetworksErrorinvalues. - Provider context gate.
wifiNetworksProviderusescontextGate: { anyOf: ["system"] }— it only fires in system-context conversations, not every agent turn. elizaos.appmetadata.package.jsoncarries anelizaos.appblock (displayName: "WiFi",category: "system",androidOnly: true,heroImage: "assets/hero.png") used by the app catalog tooling.- Root AGENTS.md. Repo-wide architecture rules, logger conventions, ESM requirements, and naming rules live in the root
AGENTS.md. This file covers only plugin-wifi specifics.
⛔ 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 — CLI / tooling:
- The real command/flow invocation transcript (args in, stdout/stderr, exit code) and the artifacts it generated (files, scaffolds, manifests, screenshots/recordings).
- Failure paths: bad args, missing deps, partial state, permission/network errors.
- A recording/log of the actual run end to end — not a unit test of one helper.
- Any model interaction captured as a live trajectory and reviewed.