Skip to content

hermes.md

Latest commit

 

History

History
169 lines (129 loc) · 8.38 KB

hermes.md

File metadata and controls

169 lines (129 loc) · 8.38 KB
Raw

Hermes Agent

Install the official Inline platform adapter for Hermes Agent.

Requirements

  • Hermes Agent 0.17.x
  • Node.js 20 or newer
  • The Inline CLI for guided bot creation, or an existing Inline bot/user token

The setup wizard can create and configure the bot for you. Manual bot creation is also available in Creating a Bot.

Install

npm install -g @inline-chat/hermes-agent-adapter
inline-hermes install
hermes plugins enable inline-platform
hermes gateway setup

Select Inline in the messaging-platform picker. The default path is: go to Inline → Settings → Bots → Create a new bot, then paste its token. The optional CLI path can install/sign in to the Inline CLI and create the bot from the terminal. Both paths save the token through Hermes' credential helper and configure access.

Coding Agent Setup Prompt

Use this with Codex, Claude Code, or another local coding agent for a simple setup:

Set up the Inline Hermes Agent adapter on this machine.

Constraints:
- Do not read, print, or edit .env files.
- Do not print Inline tokens or other secrets.
- Use an Inline token from INLINE_TOKEN or INLINE_BOT_TOKEN; if neither is present, stop and point me to https://inline.chat/docs/creating-a-bot.

Tasks:
1. Verify Node.js is version 20 or newer and Hermes Agent is installed.
2. Install or upgrade @inline-chat/hermes-agent-adapter globally.
3. Run inline-hermes install and hermes plugins enable inline-platform.
4. Run hermes gateway setup and select Inline. Prefer its guided bot-creation path; if no Inline token is available, let this interactive wizard ask the user to sign in or paste one.
5. Run inline-hermes doctor --json and inline-hermes test-send --dry-run --to chat:123 --text "Inline Hermes dry-run" --json.
6. Report the exact commands run and any remaining manual steps, without revealing secrets.

Manual configuration

The guided setup above is the default. For headless deployments or managed environments, configure the same values manually:

Set the token in the Hermes gateway environment:

export INLINE_TOKEN="<INLINE_BOT_TOKEN>"

Enable the Inline platform in ~/.hermes/config.yaml:

platforms:
  inline:
    enabled: true

Inline uses work-chat defaults: typing/presence while Hermes is working, final answers in chat, and no durable tool-call progress bubbles. If your Hermes config has global tool progress enabled, add this per-platform override:

display:
  platforms:
    inline:
      tool_progress: off
      cleanup_progress: true
      streaming: false
      interim_assistant_messages: false

If a process manager cannot inject the token directly, Hermes config can point to an environment variable without storing the token in the file:

platforms:
  inline:
    enabled: true
    token: ${INLINE_TOKEN}

Verify

inline-hermes doctor --json
hermes inline status
inline-hermes --version
inline-hermes test-send --dry-run --to chat:123 --text "Inline Hermes dry-run" --json

With a real target and a token in the environment or Hermes config:

export INLINE_TOKEN="<INLINE_BOT_TOKEN>"
inline-hermes test-send --to chat:123 --text "Inline Hermes test"

Hermes' generic send command uses the inline:<chat-id> target form:

hermes send --to inline:123 "Hello from Hermes"

Update

After upgrading the npm package, refresh the Hermes plugin copy:

npm install -g @inline-chat/hermes-agent-adapter@latest
inline-hermes install --force
inline-hermes --version
inline-hermes doctor --json

This does not edit ~/.hermes/config.yaml or tokens. It only replaces the installed plugin files and verifies the bundled sidecar hash.

Feature Support

Supported:

  • DMs, group chats, Inline reply threads, and hermes send --to inline:<chat-id>.
  • Token setup through INLINE_TOKEN, INLINE_BOT_TOKEN, platforms.inline.token, or inline.token, including simple ${ENV_NAME} config references.
  • Realtime inbound messages, catch-up, replies, action callbacks, opt-in edit-message streaming, edits, deletes, typing, presence, and long outbound replies.
  • Inline-native clarify, approval, slash confirmation, and model picker buttons.
  • Native Inline / command-menu sync for Hermes slash commands, including /threads and /update.
  • Native Hermes inline tool for bounded current-chat/thread history and search, exact message lookup, sending, editing, deleting bot-owned messages, reactions, pin/unpin/list pins, typing/presence, reply-thread creation, and private or explicitly public top-level thread/chat creation outside the current conversation.
  • Per-turn Inline sender/chat/thread IDs, selective reply/thread/observed context, and parent-thread context, with prompt guidance for sender mentions and current chat/thread Markdown links.
  • Local photo, video, voice, and document uploads; URL-backed inbound media caching when available.
  • User/group allowlists, group sender allowlists, mention controls, allowed chats, free-response chats, default reply-thread routing with /threads controls, thread prompts, and skill bindings.
  • Inline entity summaries for live turns and tool-fetched history, including mentions, text links, thread links, thread-title links, code/pre blocks, bot commands, and group mentions as untrusted Hermes context.
  • Reactions on bot messages and opt-in lifecycle/system events.
  • Hermes-native typing_indicator and gateway_restart_notification toggles.

Unsupported or intentionally limited:

  • Multiple Inline accounts in one Hermes process.
  • Full Inline member, space, and admin tools beyond bounded current-chat/thread message access.
  • Full Inline rich-text span conversion. Rich entities are summarized for the agent instead of converted into Hermes-specific spans.
  • Native animated draft streaming. Inline can stream by editing a preview message, but edit-based streaming is off by default like Discord and Slack.
  • Ephemeral in-channel private replies or realtime voice/calls.
  • Media without a usable Inline CDN/local URL. Hermes still receives a text summary, but may not receive a local file path.

Troubleshooting

  • Hermes plugin 'inline-platform' is not enabled: run hermes plugins enable inline-platform.
  • Missing-token diagnostics: set a valid Inline token in the Hermes gateway environment, or configure platforms.inline.token/inline.token for the adapter.
  • SESSION_REVOKED: the token reached Inline realtime but is expired or revoked. Create or rotate the token and retry.
  • Node errors: install Node.js 20+ or set INLINE_NODE_BIN to the Node executable Hermes should use.
  • Slash commands typed as text always work; /command@botusername is accepted when the suffix matches this bot and ignored when it targets another bot. The native / menu is synced on gateway connect and can be disabled with INLINE_SYNC_COMMANDS=false or platforms.inline.sync_commands: false.
  • Room controls are available through INLINE_ALLOWED_CHATS, INLINE_FREE_RESPONSE_CHATS, and INLINE_STRICT_MENTION. Parent chat ids also match Inline reply threads.
  • In mention-gated groups, a concrete mention of another person at the beginning of a message overrides inferred follow/reply attention unless Hermes is also explicitly mentioned.
  • Top-level DM and group replies use Inline reply threads by default. Use /threads on, /threads off, or /threads auto in Inline to configure a chat, or set INLINE_REPLY_THREADS=false globally.
  • Automatic context uses INLINE_CONTEXT_BACKFILL=selective by default: parent/reply-thread metadata, parent messages, reply windows, first thread turns, mention-gap catch-up, and observed unmentioned group context are included when relevant. Use INLINE_THREAD_CONTEXT_LIMIT, INLINE_REPLY_CONTEXT_LIMIT, INLINE_OBSERVE_UNMENTIONED_MESSAGES=false, or INLINE_OBSERVED_CONTEXT_LIMIT to tune it. Exact older history remains available through the inline tool.
  • Thread-specific prompts and skill bindings are supported through platforms.inline.channel_prompts and platforms.inline.channel_skill_bindings; Inline checks the thread chat id first, then the parent chat id. See the package README for examples.
  • Set display.platforms.inline.tool_progress: off to suppress tool-call progress messages. If you opt into progress, keep display.platforms.inline.cleanup_progress: true so successful runs do not leave progress bubbles behind.

Full package docs: hermes-agent/README.md.