feat(config): surface gateway streaming block in DEFAULT_CONFIG (#37285)

The gateway reads top-level streaming.* with StreamingConfig defaults when the
block is absent, so streaming was invisible — a user with no streaming block
sees responses arrive as single messages and has no way to discover the toggle
short of reading source. This materializes the block in config.yaml so it's
discoverable, with values byte-identical to the dataclass defaults (no behavior
change).

- DEFAULT_CONFIG gains a root-level streaming block (enabled, transport,
  edit_interval, buffer_threshold, cursor, fresh_final_after_seconds), each
  documented inline. Values match gateway/config.py StreamingConfig() exactly.
- _KNOWN_ROOT_KEYS gains 'streaming' so the validator accepts the root key.
- No _config_version bump: load_config deep-merges DEFAULT_CONFIG over user
  YAML, so existing installs pick up the default automatically; no value
  migration needed.

Does NOT touch the setup wizard — streaming stays opt-in, just discoverable.

hermes_cli/config.py

@@ -2037,6 +2037,45 @@ DEFAULT_CONFIG = {
         "trust_recent_files_seconds": 600,
     },
 
+    # Real-time token streaming to messaging platforms (Telegram, Discord,
+    # Slack, etc.). Read at the top level by the gateway; absent this block the
+    # gateway falls back to these same defaults, so adding it here only makes
+    # the feature discoverable in config.yaml — it does not change behavior.
+    #
+    # Disabled by default: streaming costs extra edit/draft API calls per
+    # response. Set ``enabled: true`` and restart the gateway to turn it on.
+    "streaming": {
+        # Master switch. When false, each response is delivered as a single
+        # final message (no progressive updates).
+        "enabled": False,
+        # Transport selection:
+        #   "auto"  — prefer native draft streaming where the platform
+        #             supports it (Telegram DMs via sendMessageDraft,
+        #             Bot API 9.5+) and fall back to edit-based elsewhere.
+        #             Safe global default: platforms without draft support
+        #             (Discord, Slack, Matrix, Telegram groups) transparently
+        #             use the edit path, so "auto" only upgrades chats that
+        #             can render the smoother native preview.
+        #   "draft" — explicitly request native drafts; falls back to edit
+        #             when the platform/chat doesn't support them.
+        #   "edit"  — progressive editMessageText only (legacy behavior).
+        #   "off"   — disable streaming entirely (same as enabled: false).
+        "transport": "auto",
+        # Minimum seconds between progressive edits — tuned for Telegram's
+        # ~1 edit/s flood envelope.
+        "edit_interval": 0.8,
+        # Flush the buffer to the platform once this many characters have
+        # accumulated, so short replies feel near-instant.
+        "buffer_threshold": 24,
+        # Cursor glyph appended to the in-progress message while streaming.
+        "cursor": " \u2589",
+        # When >0, the final edit for a long-running streamed response is
+        # delivered as a fresh message if the preview has been visible at
+        # least this many seconds, so the platform timestamp reflects
+        # completion time. Telegram only; other platforms ignore it.
+        "fresh_final_after_seconds": 60.0,
+    },
+
     # Session storage — controls automatic cleanup of ~/.hermes/state.db.
     # state.db accumulates every session, message, tool call, and FTS5 index
     # entry forever.  Without auto-pruning, a heavy user (gateway + cron)
@@ -3756,7 +3795,7 @@ _KNOWN_ROOT_KEYS = {
     "fallback_providers", "credential_pool_strategies", "toolsets",
     "agent", "terminal", "display", "compression", "delegation",
     "auxiliary", "custom_providers", "context", "memory", "gateway",
-    "sessions",
+    "sessions", "streaming",
 }
 
 # Valid fields inside a custom_providers list entry