Companions & Channels

Channels

With channels, you operate NomiFun’s companions from external chat apps instead of sitting at the desktop. Enable a connector, paste its credentials, authorize a chat user with a one-time pairing code, and from then on messages to your bot route to a companion — its replies come back to the same chat.

The default master-agent mode has the companion itself answer remote messages: the session inherits the companion’s personality and memory and is wired to the Desktop Gateway tools — so the thing you’re chatting with on your phone isn’t an isolated bot, it’s the agent that runs your whole desktop. “Move my daily-report cron to 9am and tell me what’s running on the desktop right now” is a single Lark message.

Where to find it

Open the Nomi page (sidebar “Companions & Channels”, route /nomi), pick a companion, and open its Remote tab. It lists the connectors available to that companion — each with a status pill (stopped / connected), the bot username once connected, the current authorized-user count, and a per-channel default companion / default model picker.

Companion remote-connection / channel binding

The 12 channels in the default build

Telegram, Lark / Feishu, DingTalk, WeChat, WeCom (Enterprise WeChat), Slack, Discord, Matrix, Mattermost, Twitch, Nostr, and QQ Bot — 12 built-in channels compiled into the default Desktop backend. Each platform still requires you to create its bot or app, provide credentials, and follow that platform’s network and review requirements.

Steps

  1. Find the entry point. In Nomi → Remote, select the connector you want to wire up. Each connector is its own row of configuration and they’re independent.

  2. Paste credentials and enable. Per platform:

    • Telegram: create a bot with @BotFather, save the token (123456:ABC-DEF…), paste it in, click Test to verify (the backend calls getMe and echoes the bot username), then Enable.
    • Lark / DingTalk: create a custom app in the respective developer console, fill in the App ID / App Secret (DingTalk uses Client ID / Secret with Stream mode enabled), then Enable.
    • WeChat: just click Enable and scan the QR code with the WeChat app; the plugin flips to connected.

  3. Pair an authorized user. Pairing has two entry points:

    • A platform user messages the bot for the first time, and the bot replies with a 6-digit pairing code (10-minute TTL);
    • You paste or confirm that code under Remote → Pending pairings on the desktop and click Approve.
    • Approved users appear in Authorized users with a last active timestamp and can be revoked anytime; revoking also clears that user’s active session, so their next message re-pairs from scratch.

  4. Choose the handling companion and default model. In the connector panel, set the companion and default model for that channel — the configuration panel shown above is the entry point. Each IM platform can bind one companion; switching a binding resets that channel’s active session so the next message is handled by the new companion. A channel row with no binding falls back to the default companion.

What you can do from IM

  • Plain text both ways, with streaming edited replies — the companion’s incremental updates are edited into the in-flight bot message (except WeChat, which falls back to sending a new message).
  • Action buttons — confirm prompts, retries, etc., rendered as each platform’s native inline keyboard (Telegram) / interactive card (Lark) or equivalent.
  • @-only in groups — a group chat can be set to respond only when the bot is @-mentioned.

Not yet doable from IM: creating a team (use the desktop / WebUI), file uploads beyond the platform plugin’s native ability, or per-user workspace selection.

Notes & boundaries

  • One companion per platform. Bots bind to a companion per channel row, and the structure guarantees one bot is never bound to a second companion; the same platform can host multiple bots (e.g. a separate Lark custom app per companion).
  • No binding → default companion. A channel row with no explicit binding falls back through the platform preference to the default companion; if a bound companion is later deleted, it auto-falls back to the default and resets the session.
  • Memory has shared / private scopes. Shared memories are family-wide; private memories are visible only to their owning companion. Skills, by contrast, are per-companion isolated, each with its own skill library.
  • Pairing-code safety. 6-digit random code, 10-minute TTL, with a periodic sweep that expires stale codes.

FAQ

Which companion handles a channel with no binding? It falls back to the default companion, still carrying personality, memory, and the gateway tools.

Can I turn an IM conversation directly into a requirement / issue? Inbound-to-requirement isn’t supported yet — it’s a roadmap item (coming soon). Completion-style notifications, by contrast, are outbound webhooks (Lark signed card / Slack / HTTP).

Why can’t WeChat edit replies in place? The WeChat API doesn’t support message editing, so replies are delivered as new messages in the same chat.

  • Companions — multi-companion management, shared/private memory scopes, and per-companion skill libraries.
  • AutoWork & Requirements — outbound completion-notification webhooks and requirement logging.

Full docs → GitHub