From 77aa026ca67f65a8e781df1b82c0786d511bfd21 Mon Sep 17 00:00:00 2001
From: githubespresso407 <githubespresso407@users.noreply.github.com>
Date: Sat, 18 Jul 2026 15:09:43 -0500
Subject: [PATCH] Resolve kimi-k3 context length to 1M on canonical Kimi Coding
 endpoints

Kimi Coding serves K3 under the bare slug 'k3', but users can also
configure or select the public-facing aliases 'kimi-k3' and
'kimi-k3-cot'. The endpoint-scoped 1M context window was only keyed
on the bare 'k3' slug, so selecting 'kimi-k3' fell through to the
generic 'kimi' catch-all (262k).

Extend the guard in _endpoint_scoped_context_length to also recognize
'kimi-k3' and 'kimi-k3-cot', while keeping the endpoint check that
limits the 1M value to https://api.kimi.com/coding (legacy Moonshot
endpoints still fall back to 262k). Update the existing test to cover
all three aliases.

Fixes: context window limited to 262k when using kimi-k3 via kimi-coding.
---
 agent/model_metadata.py            | 10 ++++++++--
 tests/agent/test_model_metadata.py | 14 ++++++++------
 2 files changed, 16 insertions(+), 8 deletions(-)

diff --git a/agent/model_metadata.py b/agent/model_metadata.py
index 50e78a8868fd..821a7ebe01c1 100644
--- a/agent/model_metadata.py
+++ b/agent/model_metadata.py
@@ -543,7 +543,13 @@ def _is_known_provider_base_url(base_url: str) -> bool:
 
 
 def _endpoint_scoped_context_length(model: str, base_url: str) -> Optional[int]:
-    """Return metadata confirmed only for one provider endpoint."""
+    """Return metadata confirmed only for the Kimi Coding endpoint.
+
+    Kimi Coding serves K3 under the bare slug ``k3``, but users may also
+    configure or select the public-facing aliases ``kimi-k3`` and
+    ``kimi-k3-cot``. Only canonical ``https://api.kimi.com/coding`` endpoints
+    (legacy Moonshot keys do not serve K3) get the 1 Mi context window.
+    """
     normalized = _normalize_base_url(base_url)
     try:
         parsed = urlparse(normalized)
@@ -559,7 +565,7 @@ def _endpoint_scoped_context_length(model: str, base_url: str) -> Optional[int]:
         and parsed.path.rstrip("/") in {"/coding", "/coding/v1"}
         and not parsed.query
         and not parsed.fragment
-        and model.strip().lower() == "k3"
+        and model.strip().lower() in {"k3", "kimi-k3", "kimi-k3-cot"}
     ):
         return 1_048_576
     return None
diff --git a/tests/agent/test_model_metadata.py b/tests/agent/test_model_metadata.py
index 166e0c1f97fa..ddd49834d579 100644
--- a/tests/agent/test_model_metadata.py
+++ b/tests/agent/test_model_metadata.py
@@ -199,14 +199,16 @@ class TestDefaultContextLengths:
             )
 
             for base_url in accepted_urls:
-                assert get_model_context_length(
-                    "k3", provider="kimi-coding", base_url=base_url
-                ) == 1_048_576
+                for model in ("k3", "kimi-k3", "kimi-k3-cot"):
+                    assert get_model_context_length(
+                        model, provider="kimi-coding", base_url=base_url
+                    ) == 1_048_576
 
             for base_url in rejected_urls:
-                assert get_model_context_length(
-                    "k3", provider="kimi-coding", base_url=base_url
-                ) != 1_048_576
+                for model in ("k3", "kimi-k3", "kimi-k3-cot"):
+                    assert get_model_context_length(
+                        model, provider="kimi-coding", base_url=base_url
+                    ) != 1_048_576
 
     def test_grok_substring_matching(self):
         # Longest-first substring matching must resolve the real xAI model