fix(agent): suppress intermediate retry status messages in chat thread

hermes_cli/model_switch.py

@@ -1094,6 +1094,22 @@ def _record_builtin_endpoint(slug: str) -> None:
         curated["lmstudio"] = live
 
     # --- 1. Check Hermes-mapped providers ---
+    # Pre-scan custom_providers for base_urls so we can suppress built-in
+    # providers that would duplicate a user-defined endpoint (e.g. OpenAI
+    # native vs. user-defined OpenAI pointing at api.openai.com/v1).
+    _custom_base_urls: set[str] = set()
+    if custom_providers and isinstance(custom_providers, list):
+        for entry in custom_providers:
+            if isinstance(entry, dict):
+                bu = (entry.get("base_url") or "").strip().rstrip("/")
+                if bu:
+                    _custom_base_urls.add(bu.lower())
+
+    def _custom_provider_has_base_url(base_url: str) -> bool:
+        """Return True if a custom provider already claims this base_url."""
+        target = base_url.strip().rstrip("/").lower()
+        return target in _custom_base_urls
+
     for hermes_id, mdev_id in PROVIDER_TO_MODELS_DEV.items():
         # Skip aliases that map to the same models.dev provider (e.g.
         # kimi-coding and kimi-coding-cn both → kimi-for-coding).
@@ -1146,6 +1162,16 @@ def _record_builtin_endpoint(slug: str) -> None:
         pinfo = _mdev_pinfo(mdev_id)
         display_name = pinfo.name if pinfo else mdev_id
 
+        # Suppress built-in `openai` when a custom provider already claims
+        # the same base_url (e.g. api.openai.com/v1). The custom provider
+        # typically has a richer model list and is the intended entry.
+        if slug == "openai" and _custom_provider_has_base_url(
+            "https://api.openai.com/v1"
+        ):
+            seen_slugs.add(slug.lower())
+            seen_mdev_ids.add(mdev_id)
+            continue
+
         results.append({
             "slug": slug,
             "name": display_name,

hermes_cli/runtime_provider.py

@@ -521,10 +521,18 @@ def _resolve_named_custom_runtime(
             pool_result["model"] = model_name
         return pool_result
 
+    configured_api_key = str(custom_provider.get("api_key", "") or "").strip()
+    configured_key_env = str(custom_provider.get("key_env", "") or "").strip()
+    configured_api_key_env = ""
+    if configured_api_key.startswith("env:"):
+        configured_api_key_env = configured_api_key.split(":", 1)[1].strip()
+        configured_api_key = ""
+
     api_key_candidates = [
         (explicit_api_key or "").strip(),
-        str(custom_provider.get("api_key", "") or "").strip(),
-        os.getenv(str(custom_provider.get("key_env", "") or "").strip(), "").strip(),
+        configured_api_key,
+        os.getenv(configured_key_env, "").strip(),
+        os.getenv(configured_api_key_env, "").strip(),
         os.getenv("OPENAI_API_KEY", "").strip(),
         os.getenv("OPENROUTER_API_KEY", "").strip(),
     ]

run_agent.py

@@ -7100,10 +7100,17 @@ def _call():
                             result["partial_tool_names"] = []
                             deltas_were_sent["yes"] = False
                             first_delta_fired["done"] = False
-                            self._emit_status(
-                                f"⚠️ Connection dropped mid tool-call "
-                                f"({type(e).__name__}). Reconnecting… "
-                                f"(attempt {_stream_attempt + 2}/{_max_stream_retries + 1})"
+                            # Intermediate retry messages are logged (not emitted)
+                            # to avoid cluttering the chat thread with transient
+                            # warnings that remain even after successful retries.
+                            # See: https://github.com/NousResearch/hermes-agent/issues/5151
+                            logger.info(
+                                "⚠️ Connection dropped mid tool-call "
+                                "(%s). Reconnecting… "
+                                "(attempt %s/%s)",
+                                type(e).__name__,
+                                _stream_attempt + 2,
+                                _max_stream_retries + 1,
                             )
                             self._touch_activity(
                                 f"stream retry {_stream_attempt + 2}/{_max_stream_retries + 1} "
@@ -7166,10 +7173,17 @@ def _call():
                                     type(e).__name__,
                                     e,
                                 )
-                                self._emit_status(
-                                    f"⚠️ Connection to provider dropped "
-                                    f"({type(e).__name__}). Reconnecting… "
-                                    f"(attempt {_stream_attempt + 2}/{_max_stream_retries + 1})"
+                                # Intermediate retry messages are logged (not emitted)
+                                # to avoid cluttering the chat thread with transient
+                                # warnings that remain even after successful retries.
+                                # See: https://github.com/NousResearch/hermes-agent/issues/5151
+                                logger.info(
+                                    "⚠️ Connection to provider dropped "
+                                    "(%s). Reconnecting… "
+                                    "(attempt %s/%s)",
+                                    type(e).__name__,
+                                    _stream_attempt + 2,
+                                    _max_stream_retries + 1,
                                 )
                                 self._touch_activity(
                                     f"stream retry {_stream_attempt + 2}/{_max_stream_retries + 1} "