Add provider for-tools command and max_tokens support

- Add 'cmdforge providers for-tools <tool>...' to list providers used by tools
- Add --warm flag to pre-load local models (Ollama) for faster first inference
- Warm larger models first to claim contiguous GPU memory
- Skip cloud providers (claude, opencode, gemini) during warm-up
- Add max_tokens field to PromptStep for controlling output length
- Pass max_tokens to call_provider with provider-specific flags
- Add comprehensive tests for provider functionality

Co-Authored-By: Claude Opus 4.5 <[email protected]>

tests/test_providers.py

@@ -12,7 +12,7 @@ from cmdforge.providers import (
     load_providers, save_providers, get_provider,
     add_provider, delete_provider,
     call_provider, mock_provider,
-    DEFAULT_PROVIDERS
+    DEFAULT_PROVIDERS, strip_ansi
 )
 
 
@@ -507,3 +507,122 @@ class TestProviderFallback:
 
         assert result.success is True
         assert "[MOCK]" in result.text
+
+
+class TestStripAnsi:
+    """Tests for ANSI escape code stripping."""
+
+    def test_strip_basic_color_codes(self):
+        """Strip basic color codes like \\x1b[0m."""
+        text = "\x1b[31mred text\x1b[0m"
+        assert strip_ansi(text) == "red text"
+
+    def test_strip_cursor_movements(self):
+        """Strip cursor movement codes."""
+        text = "\x1b[2Jhello\x1b[H"
+        assert strip_ansi(text) == "hello"
+
+    def test_strip_private_mode_sequences(self):
+        """Strip private mode sequences like \\x1b[?25l (hide cursor)."""
+        text = "\x1b[?2026h\x1b[?25lspinner\x1b[?25h\x1b[?2026l"
+        assert strip_ansi(text) == "spinner"
+
+    def test_strip_ollama_spinner(self):
+        """Strip Ollama spinner output."""
+        text = "\x1b[?2026h\x1b[?25l⠋ \x1b[?25h\x1b[?2026l\x1b[?2026h\x1b[?25l⠙ \x1b[?25h"
+        result = strip_ansi(text)
+        # Should only have spinner characters left, not escape codes
+        assert "\x1b" not in result
+        assert "⠋" in result or "⠙" in result
+
+    def test_strip_carriage_return(self):
+        """Strip carriage returns."""
+        text = "line1\rline2"
+        assert strip_ansi(text) == "line1line2"
+
+    def test_preserve_normal_text(self):
+        """Normal text should be preserved."""
+        text = "Hello, World! 123 [test] {data}"
+        assert strip_ansi(text) == text
+
+    def test_strip_osc_sequences(self):
+        """Strip OSC (Operating System Command) sequences."""
+        text = "\x1b]0;Window Title\x07content"
+        assert strip_ansi(text) == "content"
+
+
+class TestMaxTokens:
+    """Tests for max_tokens parameter."""
+
+    @pytest.fixture
+    def temp_providers_file(self, tmp_path):
+        providers_file = tmp_path / ".cmdforge" / "providers.yaml"
+        with patch('cmdforge.providers.PROVIDERS_FILE', providers_file):
+            yield providers_file
+
+    @patch('subprocess.run')
+    @patch('shutil.which')
+    def test_max_tokens_appended_for_claude(self, mock_which, mock_run, temp_providers_file):
+        """max_tokens should append --max-tokens for claude providers."""
+        mock_which.return_value = "/usr/bin/claude"
+        mock_run.return_value = MagicMock(returncode=0, stdout="response", stderr="")
+        save_providers([Provider("claude-haiku", "claude -p --model haiku")])
+
+        call_provider("claude-haiku", "Test prompt", max_tokens=4096)
+
+        # Verify the command included --max-tokens
+        call_args = mock_run.call_args
+        cmd = call_args[0][0]
+        assert "--max-tokens 4096" in cmd
+
+    @patch('subprocess.run')
+    @patch('shutil.which')
+    def test_max_tokens_appended_for_gemini(self, mock_which, mock_run, temp_providers_file):
+        """max_tokens should append --max-output-tokens for gemini providers."""
+        mock_which.return_value = "/usr/bin/gemini"
+        mock_run.return_value = MagicMock(returncode=0, stdout="response", stderr="")
+        save_providers([Provider("gemini", "gemini --model gemini-2.5-pro")])
+
+        call_provider("gemini", "Test prompt", max_tokens=8192)
+
+        call_args = mock_run.call_args
+        cmd = call_args[0][0]
+        assert "--max-output-tokens 8192" in cmd
+
+    @patch('subprocess.run')
+    @patch('shutil.which')
+    def test_max_tokens_not_appended_when_none(self, mock_which, mock_run, temp_providers_file):
+        """No flag should be added when max_tokens is None."""
+        mock_which.return_value = "/usr/bin/claude"
+        mock_run.return_value = MagicMock(returncode=0, stdout="response", stderr="")
+        save_providers([Provider("claude", "claude -p")])
+
+        call_provider("claude", "Test prompt", max_tokens=None)
+
+        call_args = mock_run.call_args
+        cmd = call_args[0][0]
+        assert "--max-tokens" not in cmd
+
+    @patch('subprocess.run')
+    @patch('shutil.which')
+    def test_max_tokens_passed_through_fallback(self, mock_which, mock_run, temp_providers_file):
+        """max_tokens should be passed to fallback provider."""
+        mock_which.return_value = "/usr/bin/claude"
+        # First call fails, second succeeds
+        mock_run.side_effect = [
+            MagicMock(returncode=1, stdout="", stderr="Error"),
+            MagicMock(returncode=0, stdout="fallback response", stderr="")
+        ]
+        save_providers([
+            Provider("primary", "claude -p --model opus", fallback="fallback"),
+            Provider("fallback", "claude -p --model haiku")
+        ])
+
+        result = call_provider("primary", "Test prompt", max_tokens=4096)
+
+        assert result.success is True
+        # Both calls should have had max_tokens
+        calls = mock_run.call_args_list
+        assert "--max-tokens 4096" in calls[0][0][0]
+        assert "--max-tokens 4096" in calls[1][0][0]
+