From df22d29522ced894ab79ff66e4496c2c93be65c4 Mon Sep 17 00:00:00 2001
From: teknium1 <127238744+teknium1@users.noreply.github.com>
Date: Sat, 16 May 2026 23:38:45 -0700
Subject: [PATCH] =?UTF-8?q?fix(copilot):=20GitHub=20Models=20413=20hint=20?=
 =?UTF-8?q?=E2=80=94=20port=20to=20extracted=20conversation=5Floop?=
MIME-Version: 1.0
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: 8bit

Original commits 4ded3ede3 (@konsisumer) + 374dc81c2 (Teknium) added a
413 hint to run_agent.py's agent loop. Final-state version (the sharpened
374dc81c2 wording) ported to agent/conversation_loop.py, where the
payload_too_large branch now lives.

The deprecation detection + _URL_TO_PROVIDER changes from both commits
landed in agent/copilot_acp_client.py and agent/model_metadata.py via
the prior merge.

Closes #10648

Co-authored-by: konsisumer <der@konsi.org>
Co-authored-by: Teknium <127238744+teknium1@users.noreply.github.com>
---
 agent/conversation_loop.py | 33 +++++++++++++++++++++++++++++++++
 1 file changed, 33 insertions(+)

diff --git a/agent/conversation_loop.py b/agent/conversation_loop.py
index e121c4b2a7..8096b75429 100644
--- a/agent/conversation_loop.py
+++ b/agent/conversation_loop.py
@@ -2333,6 +2333,39 @@ def run_conversation(
                     classified.reason == FailoverReason.payload_too_large
                 )
 
+                # Actionable hint for GitHub Models (Azure) 413 errors.
+                # The free tier enforces a hard 8K token cap per request,
+                # which Hermes' system prompt + tool schemas alone exceed.
+                # Compression can't help — the floor is the system prompt
+                # itself, not the conversation — so surface a clear "not
+                # compatible" message instead of looping into three futile
+                # compression attempts.
+                if (
+                    status_code == 413
+                    and isinstance(agent.base_url, str)
+                    and "models.inference.ai.azure.com" in agent.base_url
+                ):
+                    agent._vprint(
+                        f"{agent.log_prefix}   💡 GitHub Models free tier (models.inference.ai.azure.com) caps every",
+                        force=True,
+                    )
+                    agent._vprint(
+                        f"{agent.log_prefix}      request at ~8K tokens. Hermes' system prompt + tool schemas baseline",
+                        force=True,
+                    )
+                    agent._vprint(
+                        f"{agent.log_prefix}      exceeds that floor, so this endpoint cannot run an agentic loop.",
+                        force=True,
+                    )
+                    agent._vprint(
+                        f"{agent.log_prefix}      Use the `copilot` provider with a Copilot subscription token (`hermes",
+                        force=True,
+                    )
+                    agent._vprint(
+                        f"{agent.log_prefix}      setup` → GitHub Copilot), or pick any other provider.",
+                        force=True,
+                    )
+
                 if is_payload_too_large:
                     compression_attempts += 1
                     if compression_attempts > max_compression_attempts: