fix(copilot-acp): tighten deprecation detection + sharpen GitHub Models 413 hint

Follow-up improvements on top of @konsisumer's cherry-picked fix for #10648:

1. Deprecation patterns required BOTH a product fingerprint ('gh-copilot') and
   a deprecation marker. The previous list included 'copilot-cli' and bare
   'deprecation', which would false-positive on stderr from the NEW
   @github/copilot CLI — whose repo is literally github.com/github/copilot-cli
   and which legitimately surfaces those substrings in its own messages.

2. Replace the deprecation hint. The user in #10648 installed
   'gh extension install github/gh-copilot' (the deprecated extension)
   thinking that's what ACP mode uses, when ACP actually spawns the new
   'copilot' binary from '@github/copilot'. The hint now points users at the
   correct install command ('npm install -g @github/copilot') with the new
   CLI's repo URL, and demotes provider-switching to a fallback alternative.

3. Change _URL_TO_PROVIDER value for models.inference.ai.azure.com from the
   'github-models' alias to the canonical 'copilot' provider id, matching the
   convention used by every other entry in the table.

4. Sharpen the 413 hint message. The free tier's ~8K cap is below the
   system-prompt floor, so this endpoint is fundamentally incompatible with
   an agentic loop — not a 'use a different URL' problem.

Tests:
- New parametrized false-positive coverage for the new CLI's stderr shape.
- Updated assertion to require canonical 'copilot' provider mapping.
- All 14 deprecation/URL tests pass.

agent/copilot_acp_client.py

@@ -30,15 +30,29 @@ _DEFAULT_TIMEOUT_SECONDS = 900.0
 _TOOL_CALL_BLOCK_RE = re.compile(r"<tool_call>\s*(\{.*?\})\s*</tool_call>", re.DOTALL)
 _TOOL_CALL_JSON_RE = re.compile(r"\{\s*\"id\"\s*:\s*\"[^\"]+\"\s*,\s*\"type\"\s*:\s*\"function\"\s*,\s*\"function\"\s*:\s*\{.*?\}\s*\}", re.DOTALL)
 
-# Patterns that indicate the gh-copilot CLI extension has been deprecated.
-_DEPRECATION_PATTERNS = (
+# Stderr fingerprint of the deprecated `gh copilot` CLI extension
+# (https://github.blog/changelog/2025-09-25-upcoming-deprecation-of-gh-copilot-cli-extension).
+# We require BOTH the literal product name ("gh-copilot") AND a deprecation
+# marker, so generic stderr from the NEW `@github/copilot` CLI — whose repo
+# is github.com/github/copilot-cli and which legitimately mentions "copilot-cli"
+# in its own banners and error messages — doesn't get misclassified as the
+# deprecated extension.
+_DEPRECATION_REQUIRED = ("gh-copilot",)
+_DEPRECATION_MARKERS = (
     "has been deprecated",
     "no commands will be executed",
-    "deprecation",
-    "copilot-cli",
 )
 
 
+def _is_gh_copilot_deprecation_message(stderr_text: str) -> bool:
+    """True iff stderr looks like the deprecated gh-copilot extension's banner."""
+
+    lower = stderr_text.lower()
+    if not any(req in lower for req in _DEPRECATION_REQUIRED):
+        return False
+    return any(marker in lower for marker in _DEPRECATION_MARKERS)
+
+
 def _resolve_command() -> str:
     return (
         os.getenv("HERMES_COPILOT_ACP_COMMAND", "").strip()
@@ -514,16 +528,19 @@ class CopilotACPClient:
 
             stderr_text = "\n".join(stderr_tail).strip()
             if proc.poll() is not None and stderr_text:
-                stderr_lower = stderr_text.lower()
-                if any(pat in stderr_lower for pat in _DEPRECATION_PATTERNS):
+                if _is_gh_copilot_deprecation_message(stderr_text):
                     raise RuntimeError(
-                        "The gh-copilot CLI extension has been deprecated by GitHub and "
-                        "can no longer be used for ACP mode.\n\n"
-                        "Alternatives:\n"
-                        "  1. Use the GitHub Copilot provider instead of ACP mode:\n"
-                        "     hermes setup  →  select 'GitHub Copilot' (uses Copilot Chat API)\n"
-                        "  2. Set HERMES_COPILOT_ACP_COMMAND to point to a compatible ACP server\n"
-                        "  3. Use a different provider (e.g. OpenAI, Anthropic, Nous)\n\n"
+                        "Hermes ACP mode requires the NEW GitHub Copilot CLI "
+                        "(github.com/github/copilot-cli), but the binary it just "
+                        "spawned is the deprecated `gh copilot` extension.\n\n"
+                        "Install the new CLI:\n"
+                        "  npm install -g @github/copilot\n"
+                        "  # then verify with: copilot --help\n\n"
+                        "If `copilot` already resolves to the new CLI but you still see this,\n"
+                        "point Hermes at it explicitly:\n"
+                        "  export HERMES_COPILOT_ACP_COMMAND=/path/to/new/copilot\n\n"
+                        "Alternative: use the `copilot` provider (no ACP, hits the Copilot API\n"
+                        "directly with a Copilot subscription token) via `hermes setup`.\n\n"
                         f"Original error:\n{stderr_text}"
                     )
                 raise RuntimeError(f"Copilot ACP process exited early: {stderr_text}")

agent/model_metadata.py

@@ -358,7 +358,12 @@ _URL_TO_PROVIDER: Dict[str, str] = {
     "api.deepseek.com": "deepseek",
     "api.githubcopilot.com": "copilot",
     "models.github.ai": "copilot",
-    "models.inference.ai.azure.com": "github-models",
+    # GitHub Models free tier (Azure-hosted prototyping endpoint) — same
+    # canonical provider as the Copilot API.  Hard per-request token cap
+    # (often 8K) makes it unusable for Hermes' system prompt, but mapping
+    # it here lets us recognize the endpoint and emit a targeted hint
+    # instead of falling through the unknown-custom-endpoint path.
+    "models.inference.ai.azure.com": "copilot",
     "api.fireworks.ai": "fireworks",
     "opencode.ai": "opencode-go",
     "api.x.ai": "xai",