feat: local models — managed llama.cpp runtime with one-click desktop setup

Run models locally as a first-class provider. The CLI grows a managed
llama.cpp runtime (engine install, model download, server supervision);
the desktop app grows the full setup and management story on top of it.
GUI surfaces ship behind the desktop --local launch flag (hermes desktop
--local, or the flag on the packaged app); backend routes and the CLI
are always live.

Runtime (hermes_cli/local_runtime/):
- curated GGUF catalog with per-machine variant selection: hardware
  probe (VRAM/RAM/UMA), fit planning with spill accounting, quant choice
  by context window
- derived recommendation: quality-ranked picks gated by a predicted
  decode-speed floor, bandwidth-aware on unified memory; the decision
  table is pinned as a test (pick AND reason per memory class), and the
  Recommended badge explains its pick in a tooltip fed by the resolver's
  actual branch
- engine install + model download with resumable split parts, cumulative
  plan-level progress, and staged-model integrity (a split GGUF counts
  only when every part is present)
- server supervision: spawn/adopt/stop, router mode with per-model load
  progress relayed over SSE, abandoned-request cleanup

Desktop:
- Settings -> Providers -> Local models: one-click quickstart (install
  engine, download the recommended model, boot) plus per-model download/
  activate/eject, fit-ranked catalog with context pills
- model pickers (composer dropdown + Cmd+K) show staged local models,
  in-flight downloads as live progress rows, and load-into-memory bars
- local-setup campaign tip for eligible hardware; System resources
  statusbar widget (GPU/VRAM/RAM); in-chat load progress during sends
- friendly dead-server errors, and failed agent builds retry on the next
  send instead of wedging the session

Co-developed with NVIDIA field feedback on RTX 5090 and DGX Spark.

hermes_cli/runtime_provider.py

@@ -1228,6 +1228,53 @@ def _resolve_named_custom_runtime(
     # `provider: ollama` with a LAN/WireGuard `base_url` doesn't silently
     # fall through to OpenRouter.
     requested_norm = (requested_provider or "").strip().lower()
+
+    # Managed llama.cpp runtime: a llamacpp-flavored alias with no explicit
+    # base_url resolves to the supervised server (or a detected external
+    # one) before the generic custom fallthrough. Explicit base_url always
+    # wins — a user pointing at a specific server means that server.
+    if requested_norm in ("llamacpp", "llama.cpp", "llama-cpp") and not explicit_base_url:
+        try:
+            from hermes_cli.local_runtime.endpoint import resolve_llamacpp_endpoint
+
+            endpoint = resolve_llamacpp_endpoint()
+        except Exception:  # noqa: BLE001 — resolution is best-effort
+            endpoint = None
+        if endpoint:
+            return {
+                "provider": "custom",
+                "api_mode": "chat_completions",
+                "base_url": endpoint["base_url"],
+                "api_key": (explicit_api_key or "").strip()
+                or endpoint["api_key"] or "no-key-required",
+                "source": "local-runtime",
+                "requested_provider": requested_provider,
+            }
+        # No server to serve this model. Say so and stop — falling through
+        # to the generic custom path sends the request to whatever provider
+        # picks it up (OpenRouter with a placeholder key), and the user's
+        # "local server is off" surfaces as that provider's baffling
+        # "401 Invalid API key". The switch's own state picks the message:
+        # the user who turned the server off gets pointed at the switch,
+        # anyone else at the setup pane.
+        try:
+            from hermes_cli.config import load_config as _load_cfg
+
+            _lr_enabled = bool((_load_cfg().get("local_runtime") or {}).get("enabled"))
+        except Exception:  # noqa: BLE001
+            _lr_enabled = False
+        if _lr_enabled:
+            raise ValueError(
+                "The local model server isn't running. It may still be "
+                "starting — try again in a moment, or check Settings → "
+                "Providers → Local models."
+            )
+        raise ValueError(
+            "The local model server is turned off. Turn it back on in "
+            "Settings → Providers → Local models, or switch to another "
+            "model."
+        )
+
     if requested_norm and requested_norm != "custom":
         try:
             from hermes_cli.auth import resolve_provider as _resolve_provider