fix(anthropic): use model-native output limits instead of hardcoded 16K (#3426)

The Anthropic adapter defaulted to max_tokens=16384 when no explicit value
was configured.  This severely limits thinking-enabled models where thinking
tokens count toward max_tokens:

- Claude Opus 4.6 supports 128K output but was capped at 16K
- Claude Sonnet 4.6 supports 64K output but was capped at 16K

With extended thinking (adaptive or budget-based), the model could exhaust
the entire 16K on reasoning, leaving zero tokens for the actual response.
This caused two user-visible errors:
- 'Response truncated (finish_reason=length)' — thinking consumed most tokens
- 'Response only contains think block with no content' — thinking consumed all

Fix: add _ANTHROPIC_OUTPUT_LIMITS lookup table (sourced from Anthropic docs
and Cline's model catalog) and use the model's actual output limit as the
default.  Unknown future models default to 128K (the current maximum).

Also adds context_length clamping: if the user configured a smaller context
window (e.g. custom endpoint), max_tokens is clamped to context_length - 1
to avoid exceeding the window.

Closes #2706

tests/test_anthropic_adapter.py

@@ -926,7 +926,8 @@ class TestBuildAnthropicKwargs:
         )
         assert "thinking" not in kwargs
 
-    def test_default_max_tokens(self):
+    def test_default_max_tokens_uses_model_output_limit(self):
+        """When max_tokens is None, use the model's native output limit."""
         kwargs = build_anthropic_kwargs(
             model="claude-sonnet-4-20250514",
             messages=[{"role": "user", "content": "Hi"}],
@@ -934,7 +935,135 @@ class TestBuildAnthropicKwargs:
             max_tokens=None,
             reasoning_config=None,
         )
-        assert kwargs["max_tokens"] == 16384
+        assert kwargs["max_tokens"] == 64_000  # Sonnet 4 output limit
+
+    def test_default_max_tokens_opus_4_6(self):
+        kwargs = build_anthropic_kwargs(
+            model="claude-opus-4-6",
+            messages=[{"role": "user", "content": "Hi"}],
+            tools=None,
+            max_tokens=None,
+            reasoning_config=None,
+        )
+        assert kwargs["max_tokens"] == 128_000
+
+    def test_default_max_tokens_sonnet_4_6(self):
+        kwargs = build_anthropic_kwargs(
+            model="claude-sonnet-4-6",
+            messages=[{"role": "user", "content": "Hi"}],
+            tools=None,
+            max_tokens=None,
+            reasoning_config=None,
+        )
+        assert kwargs["max_tokens"] == 64_000
+
+    def test_default_max_tokens_date_stamped_model(self):
+        """Date-stamped model IDs should resolve via substring match."""
+        kwargs = build_anthropic_kwargs(
+            model="claude-sonnet-4-5-20250929",
+            messages=[{"role": "user", "content": "Hi"}],
+            tools=None,
+            max_tokens=None,
+            reasoning_config=None,
+        )
+        assert kwargs["max_tokens"] == 64_000
+
+    def test_default_max_tokens_older_model(self):
+        kwargs = build_anthropic_kwargs(
+            model="claude-3-5-sonnet-20241022",
+            messages=[{"role": "user", "content": "Hi"}],
+            tools=None,
+            max_tokens=None,
+            reasoning_config=None,
+        )
+        assert kwargs["max_tokens"] == 8_192
+
+    def test_default_max_tokens_unknown_model_uses_highest(self):
+        """Unknown future models should get the highest known limit."""
+        kwargs = build_anthropic_kwargs(
+            model="claude-ultra-5-20260101",
+            messages=[{"role": "user", "content": "Hi"}],
+            tools=None,
+            max_tokens=None,
+            reasoning_config=None,
+        )
+        assert kwargs["max_tokens"] == 128_000
+
+    def test_explicit_max_tokens_overrides_default(self):
+        """User-specified max_tokens should be respected."""
+        kwargs = build_anthropic_kwargs(
+            model="claude-opus-4-6",
+            messages=[{"role": "user", "content": "Hi"}],
+            tools=None,
+            max_tokens=4096,
+            reasoning_config=None,
+        )
+        assert kwargs["max_tokens"] == 4096
+
+    def test_context_length_clamp(self):
+        """max_tokens should be clamped to context_length if it's smaller."""
+        kwargs = build_anthropic_kwargs(
+            model="claude-opus-4-6",  # 128K output
+            messages=[{"role": "user", "content": "Hi"}],
+            tools=None,
+            max_tokens=None,
+            reasoning_config=None,
+            context_length=50000,
+        )
+        assert kwargs["max_tokens"] == 49999  # context_length - 1
+
+    def test_context_length_no_clamp_when_larger(self):
+        """No clamping when context_length exceeds output limit."""
+        kwargs = build_anthropic_kwargs(
+            model="claude-sonnet-4-6",  # 64K output
+            messages=[{"role": "user", "content": "Hi"}],
+            tools=None,
+            max_tokens=None,
+            reasoning_config=None,
+            context_length=200000,
+        )
+        assert kwargs["max_tokens"] == 64_000
+
+
+# ---------------------------------------------------------------------------
+# Model output limit lookup
+# ---------------------------------------------------------------------------
+
+
+class TestGetAnthropicMaxOutput:
+    def test_opus_4_6(self):
+        from agent.anthropic_adapter import _get_anthropic_max_output
+        assert _get_anthropic_max_output("claude-opus-4-6") == 128_000
+
+    def test_opus_4_6_variant(self):
+        from agent.anthropic_adapter import _get_anthropic_max_output
+        assert _get_anthropic_max_output("claude-opus-4-6:1m:fast") == 128_000
+
+    def test_sonnet_4_6(self):
+        from agent.anthropic_adapter import _get_anthropic_max_output
+        assert _get_anthropic_max_output("claude-sonnet-4-6") == 64_000
+
+    def test_sonnet_4_date_stamped(self):
+        from agent.anthropic_adapter import _get_anthropic_max_output
+        assert _get_anthropic_max_output("claude-sonnet-4-20250514") == 64_000
+
+    def test_claude_3_5_sonnet(self):
+        from agent.anthropic_adapter import _get_anthropic_max_output
+        assert _get_anthropic_max_output("claude-3-5-sonnet-20241022") == 8_192
+
+    def test_claude_3_opus(self):
+        from agent.anthropic_adapter import _get_anthropic_max_output
+        assert _get_anthropic_max_output("claude-3-opus-20240229") == 4_096
+
+    def test_unknown_future_model(self):
+        from agent.anthropic_adapter import _get_anthropic_max_output
+        assert _get_anthropic_max_output("claude-ultra-5-20260101") == 128_000
+
+    def test_longest_prefix_wins(self):
+        """'claude-3-5-sonnet' should match before 'claude-3-5'."""
+        from agent.anthropic_adapter import _get_anthropic_max_output
+        # claude-3-5-sonnet (8192) should win over a hypothetical shorter match
+        assert _get_anthropic_max_output("claude-3-5-sonnet-20241022") == 8_192
 
 
 # ---------------------------------------------------------------------------