feat: centralized provider router + fix Codex vision bypass + vision error handling

Three interconnected fixes for auxiliary client infrastructure:

1. CENTRALIZED PROVIDER ROUTER (auxiliary_client.py)
   Add resolve_provider_client(provider, model, async_mode) — a single
   entry point for creating properly configured clients. Given a provider
   name and optional model, it handles auth lookup (env vars, OAuth
   tokens, auth.json), base URL resolution, provider-specific headers,
   and API format differences (Chat Completions vs Responses API for
   Codex). All auxiliary consumers should route through this instead of
   ad-hoc env var lookups.

   Refactored get_text_auxiliary_client, get_async_text_auxiliary_client,
   and get_vision_auxiliary_client to use the router internally.

2. FIX CODEX VISION BYPASS (vision_tools.py)
   vision_tools.py was constructing a raw AsyncOpenAI client from the
   sync vision client's api_key/base_url, completely bypassing the Codex
   Responses API adapter. When the vision provider resolved to Codex,
   the raw client would hit chatgpt.com/backend-api/codex with
   chat.completions.create() which only supports the Responses API.

   Fix: Added get_async_vision_auxiliary_client() which properly wraps
   Codex into AsyncCodexAuxiliaryClient. vision_tools.py now uses this
   instead of manual client construction.

3. FIX COMPRESSION FALLBACK + VISION ERROR HANDLING
   - context_compressor.py: Removed _get_fallback_client() which blindly
     looked for OPENAI_API_KEY + OPENAI_BASE_URL (fails for Codex OAuth,
     API-key providers, users without OPENAI_BASE_URL set). Replaced
     with fallback loop through resolve_provider_client() for each
     known provider, with same-provider dedup.

   - vision_tools.py: Added error detection for vision capability
     failures. Returns clear message to the model when the configured
     model doesn't support vision, instead of a generic error.

Addresses #886

tools/vision_tools.py

@@ -37,27 +37,15 @@ from pathlib import Path
 from typing import Any, Awaitable, Dict, Optional
 from urllib.parse import urlparse
 import httpx
-from openai import AsyncOpenAI
-from agent.auxiliary_client import get_vision_auxiliary_client
+from agent.auxiliary_client import get_async_vision_auxiliary_client
 from tools.debug_helpers import DebugSession
 
 logger = logging.getLogger(__name__)
 
-# Resolve vision auxiliary client at module level; build an async wrapper.
-_aux_sync_client, DEFAULT_VISION_MODEL = get_vision_auxiliary_client()
-_aux_async_client: AsyncOpenAI | None = None
-if _aux_sync_client is not None:
-    _async_kwargs = {
-        "api_key": _aux_sync_client.api_key,
-        "base_url": str(_aux_sync_client.base_url),
-    }
-    if "openrouter" in str(_aux_sync_client.base_url).lower():
-        _async_kwargs["default_headers"] = {
-            "HTTP-Referer": "https://github.com/NousResearch/hermes-agent",
-            "X-OpenRouter-Title": "Hermes Agent",
-                "X-OpenRouter-Categories": "productivity,cli-agent",
-        }
-    _aux_async_client = AsyncOpenAI(**_async_kwargs)
+# Resolve vision auxiliary client at module level.
+# Uses get_async_vision_auxiliary_client() which properly handles Codex
+# routing (Responses API adapter) instead of raw AsyncOpenAI construction.
+_aux_async_client, DEFAULT_VISION_MODEL = get_async_vision_auxiliary_client()
 
 _debug = DebugSession("vision_tools", env_var="VISION_TOOLS_DEBUG")
 
@@ -359,10 +347,28 @@ async def vision_analyze_tool(
         error_msg = f"Error analyzing image: {str(e)}"
         logger.error("%s", error_msg, exc_info=True)
         
+        # Detect vision capability errors — give the model a clear message
+        # so it can inform the user instead of a cryptic API error.
+        err_str = str(e).lower()
+        if any(hint in err_str for hint in (
+            "does not support", "not support image", "invalid_request",
+            "content_policy", "image_url", "multimodal",
+            "unrecognized request argument", "image input",
+        )):
+            analysis = (
+                f"{model} does not support vision or our request was not "
+                f"accepted by the server. Error: {e}"
+            )
+        else:
+            analysis = (
+                "There was a problem with the request and the image could not "
+                f"be analyzed. Error: {e}"
+            )
+        
         # Prepare error response
         result = {
             "success": False,
-            "analysis": "There was a problem with the request and the image could not be analyzed."
+            "analysis": analysis,
         }
         
         debug_call_data["error"] = error_msg