adversarial-workflow 0.6.2__tar.gz → 0.6.3__tar.gz

{adversarial_workflow-0.6.2 → adversarial_workflow-0.6.3}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: adversarial-workflow
-Version: 0.6.2
+Version: 0.6.3
 Summary: Multi-stage AI code review system preventing phantom work - Author/Evaluator pattern
 Author: Fredrik Matheson
 License: MIT
@@ -55,9 +55,30 @@ Evaluate proposals, sort out ideas, and prevent "phantom work" (AI claiming to i
 - 🎯 **Tool-agnostic**: Use with Claude Code, Cursor, Aider, manual coding, or any workflow
 - ✨ **Interactive onboarding**: Guided setup wizard gets you started in <5 minutes
 
-## What's New in v0.6.0
+## What's New in v0.6.3
 
-🔌 **Plugin Architecture** - Define custom evaluators without modifying the package:
+### Upgrade
+
+```bash
+pip install --upgrade adversarial-workflow
+```
+
+### v0.6.3 - Configurable Timeouts
+
+- **Per-evaluator timeout**: Add `timeout: 300` to evaluator YAML for slow models like Mistral Large
+- **CLI override**: Use `--timeout 400` to override YAML config on-the-fly
+- **Timeout logging**: See which timeout source is used (CLI/YAML/default)
+- **Safety limits**: Maximum 600 seconds to prevent runaway processes
+
+### v0.6.2 - .env Loading & Stability
+
+- **Automatic .env loading**: API keys in `.env` files are now loaded at CLI startup
+- **Custom evaluator support**: Evaluators using `api_key_env: GEMINI_API_KEY` (or other keys) now work with `.env` files
+- **Better diagnostics**: `adversarial check` correctly reports the number of variables loaded from `.env`
+
+### v0.6.0 - Plugin Architecture
+
+🔌 **Custom Evaluators** - Define your own evaluators without modifying the package:
 
 ```bash
 # Create a custom evaluator
@@ -459,6 +480,7 @@ Starting with v0.6.0, you can define project-specific evaluators without modifyi
 | `aliases` | No | Alternative command names |
 | `log_prefix` | No | CLI output prefix |
 | `fallback_model` | No | Fallback model if primary fails |
+| `timeout` | No | Timeout in seconds (default: 180, max: 600) |
 | `version` | No | Evaluator version (default: 1.0.0) |
 
 ### Listing Available Evaluators

{adversarial_workflow-0.6.2 → adversarial_workflow-0.6.3}/README.md

@@ -20,9 +20,30 @@ Evaluate proposals, sort out ideas, and prevent "phantom work" (AI claiming to i
 - 🎯 **Tool-agnostic**: Use with Claude Code, Cursor, Aider, manual coding, or any workflow
 - ✨ **Interactive onboarding**: Guided setup wizard gets you started in <5 minutes
 
-## What's New in v0.6.0
+## What's New in v0.6.3
 
-🔌 **Plugin Architecture** - Define custom evaluators without modifying the package:
+### Upgrade
+
+```bash
+pip install --upgrade adversarial-workflow
+```
+
+### v0.6.3 - Configurable Timeouts
+
+- **Per-evaluator timeout**: Add `timeout: 300` to evaluator YAML for slow models like Mistral Large
+- **CLI override**: Use `--timeout 400` to override YAML config on-the-fly
+- **Timeout logging**: See which timeout source is used (CLI/YAML/default)
+- **Safety limits**: Maximum 600 seconds to prevent runaway processes
+
+### v0.6.2 - .env Loading & Stability
+
+- **Automatic .env loading**: API keys in `.env` files are now loaded at CLI startup
+- **Custom evaluator support**: Evaluators using `api_key_env: GEMINI_API_KEY` (or other keys) now work with `.env` files
+- **Better diagnostics**: `adversarial check` correctly reports the number of variables loaded from `.env`
+
+### v0.6.0 - Plugin Architecture
+
+🔌 **Custom Evaluators** - Define your own evaluators without modifying the package:
 
 ```bash
 # Create a custom evaluator
@@ -424,6 +445,7 @@ Starting with v0.6.0, you can define project-specific evaluators without modifyi
 | `aliases` | No | Alternative command names |
 | `log_prefix` | No | CLI output prefix |
 | `fallback_model` | No | Fallback model if primary fails |
+| `timeout` | No | Timeout in seconds (default: 180, max: 600) |
 | `version` | No | Evaluator version (default: 1.0.0) |
 
 ### Listing Available Evaluators

{adversarial_workflow-0.6.2 → adversarial_workflow-0.6.3}/adversarial_workflow/__main__.py

@@ -1,4 +1,5 @@
 """Allow execution via python -m adversarial_workflow."""
+
 from .cli import main
 
 if __name__ == "__main__":

{adversarial_workflow-0.6.2 → adversarial_workflow-0.6.3}/adversarial_workflow/cli.py

@@ -322,16 +322,20 @@ def init_interactive(project_path: str = ".") -> int:
         f"{GREEN}✅ Setup Complete!{RESET}",
         [
             "Created:",
-            "  ✓ .env (with your API keys - added to .gitignore)"
-            if (anthropic_key or openai_key)
-            else "  ⚠️ .env (skipped - no API keys provided)",
+            (
+                "  ✓ .env (with your API keys - added to .gitignore)"
+                if (anthropic_key or openai_key)
+                else "  ⚠️ .env (skipped - no API keys provided)"
+            ),
             "  ✓ .adversarial/config.yml",
             "  ✓ .adversarial/scripts/ (3 workflow scripts)",
             "  ✓ .aider.conf.yml (aider configuration)",
             "",
-            "Your configuration:"
-            if (anthropic_key or openai_key)
-            else "Configuration (no API keys yet):",
+            (
+                "Your configuration:"
+                if (anthropic_key or openai_key)
+                else "Configuration (no API keys yet):"
+            ),
             f"  Author (implementation): {'Claude 3.5 Sonnet (Anthropic)' if anthropic_key else 'GPT-4o (OpenAI)' if openai_key else 'Not configured'}",
             f"  Evaluator: {'GPT-4o (OpenAI)' if openai_key else 'Claude 3.5 Sonnet (Anthropic)' if anthropic_key else 'Not configured'}",
             f"  Cost per workflow: {'~$0.02-0.10' if (anthropic_key and openai_key) else '~$0.05-0.15' if (anthropic_key or openai_key) else 'N/A'}",
@@ -2284,7 +2288,9 @@ def fetch_agent_template(url: str, template_type: str = "standard") -> Optional[
                 )
                 return None
         else:
-            print(f"{RED}❌ ERROR: {template_type} template not found in package{RESET}")
+            print(
+                f"{RED}❌ ERROR: {template_type} template not found in package{RESET}"
+            )
             return None
 
     elif template_type == "custom" and url:
@@ -3082,8 +3088,8 @@ For more information: https://github.com/movito/adversarial-workflow
             "--timeout",
             "-t",
             type=int,
-            default=180,
-            help="Timeout in seconds (default: 180)",
+            default=None,
+            help="Timeout in seconds (default: from evaluator config or 180, max: 600)",
         )
         # Store config for later execution
         eval_parser.set_defaults(evaluator_config=config)
@@ -3096,10 +3102,34 @@ For more information: https://github.com/movito/adversarial-workflow
 
     # Check for evaluator command first (has evaluator_config attribute)
     if hasattr(args, "evaluator_config"):
+        # Determine timeout: CLI flag > YAML config > default (180s)
+        if args.timeout is not None:
+            timeout = args.timeout
+            source = "CLI override"
+        elif args.evaluator_config.timeout != 180:
+            timeout = args.evaluator_config.timeout
+            source = "evaluator config"
+        else:
+            timeout = args.evaluator_config.timeout  # 180 (default)
+            source = "default"
+
+        # Validate CLI timeout (consistent with YAML validation)
+        if timeout <= 0:
+            print(f"{RED}Error: Timeout must be positive (> 0), got {timeout}{RESET}")
+            return 1
+        if timeout > 600:
+            print(
+                f"{YELLOW}Warning: Timeout {timeout}s exceeds maximum (600s), clamping to 600s{RESET}"
+            )
+            timeout = 600
+
+        # Log actual timeout and source
+        print(f"Using timeout: {timeout}s ({source})")
+
         return run_evaluator(
             args.evaluator_config,
             args.file,
-            timeout=args.timeout,
+            timeout=timeout,
         )
 
     # Execute static commands

{adversarial_workflow-0.6.2 → adversarial_workflow-0.6.3}/adversarial_workflow/evaluators/config.py

@@ -26,6 +26,7 @@ class EvaluatorConfig:
         fallback_model: Fallback model if primary fails
         aliases: Alternative command names
         version: Evaluator version
+        timeout: Timeout in seconds (default: 180, max: 600)
         source: "builtin" or "local" (set internally)
         config_file: Path to YAML file if local (set internally)
     """
@@ -43,6 +44,7 @@ class EvaluatorConfig:
     fallback_model: str | None = None
     aliases: list[str] = field(default_factory=list)
     version: str = "1.0.0"
+    timeout: int = 180  # Timeout in seconds (default: 180, max: 600)
 
     # Metadata (set internally during discovery, not from YAML)
     source: str = "builtin"

{adversarial_workflow-0.6.2 → adversarial_workflow-0.6.3}/adversarial_workflow/evaluators/discovery.py

@@ -122,6 +122,35 @@ def parse_evaluator_yaml(yml_file: Path) -> EvaluatorConfig:
                     f"Field '{field}' must be a string, got {type(value).__name__}: {value!r}"
                 )
 
+    # Validate timeout if present
+    if "timeout" in data:
+        timeout = data["timeout"]
+        # Handle null/empty values
+        if timeout is None or timeout == "":
+            raise EvaluatorParseError("Field 'timeout' cannot be null or empty")
+        # Check for bool before int (bool is subclass of int in Python)
+        # YAML parses 'yes'/'true' as True, 'no'/'false' as False
+        if isinstance(timeout, bool):
+            raise EvaluatorParseError(
+                f"Field 'timeout' must be an integer, got bool: {timeout!r}"
+            )
+        if not isinstance(timeout, int):
+            raise EvaluatorParseError(
+                f"Field 'timeout' must be an integer, got {type(timeout).__name__}: {timeout!r}"
+            )
+        # timeout=0 is invalid (does not disable timeout - use a large value instead)
+        if timeout <= 0:
+            raise EvaluatorParseError(
+                f"Field 'timeout' must be positive (> 0), got {timeout}"
+            )
+        if timeout > 600:
+            logger.warning(
+                "Timeout %ds exceeds maximum (600s), clamping to 600s in %s",
+                timeout,
+                yml_file.name,
+            )
+            data["timeout"] = 600
+
     # Filter to known fields only (log unknown fields)
     known_fields = {
         "name",
@@ -134,6 +163,7 @@ def parse_evaluator_yaml(yml_file: Path) -> EvaluatorConfig:
         "fallback_model",
         "aliases",
         "version",
+        "timeout",
     }
     unknown = set(data.keys()) - known_fields
     if unknown:

{adversarial_workflow-0.6.2 → adversarial_workflow-0.6.3}/adversarial_workflow.egg-info/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: adversarial-workflow
-Version: 0.6.2
+Version: 0.6.3
 Summary: Multi-stage AI code review system preventing phantom work - Author/Evaluator pattern
 Author: Fredrik Matheson
 License: MIT
@@ -55,9 +55,30 @@ Evaluate proposals, sort out ideas, and prevent "phantom work" (AI claiming to i
 - 🎯 **Tool-agnostic**: Use with Claude Code, Cursor, Aider, manual coding, or any workflow
 - ✨ **Interactive onboarding**: Guided setup wizard gets you started in <5 minutes
 
-## What's New in v0.6.0
+## What's New in v0.6.3
 
-🔌 **Plugin Architecture** - Define custom evaluators without modifying the package:
+### Upgrade
+
+```bash
+pip install --upgrade adversarial-workflow
+```
+
+### v0.6.3 - Configurable Timeouts
+
+- **Per-evaluator timeout**: Add `timeout: 300` to evaluator YAML for slow models like Mistral Large
+- **CLI override**: Use `--timeout 400` to override YAML config on-the-fly
+- **Timeout logging**: See which timeout source is used (CLI/YAML/default)
+- **Safety limits**: Maximum 600 seconds to prevent runaway processes
+
+### v0.6.2 - .env Loading & Stability
+
+- **Automatic .env loading**: API keys in `.env` files are now loaded at CLI startup
+- **Custom evaluator support**: Evaluators using `api_key_env: GEMINI_API_KEY` (or other keys) now work with `.env` files
+- **Better diagnostics**: `adversarial check` correctly reports the number of variables loaded from `.env`
+
+### v0.6.0 - Plugin Architecture
+
+🔌 **Custom Evaluators** - Define your own evaluators without modifying the package:
 
 ```bash
 # Create a custom evaluator
@@ -459,6 +480,7 @@ Starting with v0.6.0, you can define project-specific evaluators without modifyi
 | `aliases` | No | Alternative command names |
 | `log_prefix` | No | CLI output prefix |
 | `fallback_model` | No | Fallback model if primary fails |
+| `timeout` | No | Timeout in seconds (default: 180, max: 600) |
 | `version` | No | Evaluator version (default: 1.0.0) |
 
 ### Listing Available Evaluators

{adversarial_workflow-0.6.2 → adversarial_workflow-0.6.3}/adversarial_workflow.egg-info/SOURCES.txt

@@ -48,4 +48,5 @@ tests/test_list_evaluators.py
 tests/test_python_version.py
 tests/test_scripts_project.py
 tests/test_split_command.py
+tests/test_timeout_integration.py
 tests/test_utils_validation.py

{adversarial_workflow-0.6.2 → adversarial_workflow-0.6.3}/pyproject.toml

@@ -5,7 +5,7 @@ build-backend = "setuptools.build_meta"
 [project]
 name = "adversarial-workflow"
 
-version = "0.6.2"
+version = "0.6.3"
 
 description = "Multi-stage AI code review system preventing phantom work - Author/Evaluator pattern"
 readme = "README.md"

{adversarial_workflow-0.6.2 → adversarial_workflow-0.6.3}/tests/test_cli_dynamic_commands.py

@@ -351,7 +351,9 @@ class TestBackwardsCompatibility:
 class TestGracefulDegradation:
     """Test graceful degradation on errors."""
 
-    def test_help_works_without_local_evaluators_dir(self, tmp_path, monkeypatch, run_cli):
+    def test_help_works_without_local_evaluators_dir(
+        self, tmp_path, monkeypatch, run_cli
+    ):
         """CLI help works even without .adversarial/evaluators/ directory."""
         adv_dir = tmp_path / ".adversarial"
         adv_dir.mkdir(parents=True)
@@ -420,7 +422,9 @@ class TestReviewCommandBackwardsCompatibility:
         # Review should NOT have --timeout flag (that's for evaluators)
         assert "--timeout" not in result.stdout
 
-    def test_review_command_not_overridden_by_evaluator(self, tmp_path, monkeypatch, run_cli):
+    def test_review_command_not_overridden_by_evaluator(
+        self, tmp_path, monkeypatch, run_cli
+    ):
         """Review command cannot be overridden by local evaluator."""
         adv_dir = tmp_path / ".adversarial"
         adv_dir.mkdir(parents=True)
@@ -488,7 +492,9 @@ aliases:
         assert "--path" in result_init.stdout
         assert "--interactive" in result_init.stdout
 
-    def test_evaluator_with_conflicting_name_and_alias(self, tmp_path, monkeypatch, run_cli):
+    def test_evaluator_with_conflicting_name_and_alias(
+        self, tmp_path, monkeypatch, run_cli
+    ):
         """Evaluator with conflicting name doesn't crash when alias is processed."""
         adv_dir = tmp_path / ".adversarial"
         adv_dir.mkdir(parents=True)
@@ -518,3 +524,52 @@ aliases:
         assert result.returncode == 0
         # 'init' should still be the static command
         assert "init" in result.stdout
+
+
+class TestTimeoutConfiguration:
+    """Test timeout configuration from YAML and CLI."""
+
+    def test_evaluator_config_timeout_in_yaml(self, tmp_path, monkeypatch, run_cli):
+        """Evaluator YAML timeout appears in help text."""
+        adv_dir = tmp_path / ".adversarial"
+        adv_dir.mkdir(parents=True)
+        (adv_dir / "config.yml").write_text("log_directory: .adversarial/logs/")
+
+        eval_dir = adv_dir / "evaluators"
+        eval_dir.mkdir(parents=True)
+        (eval_dir / "slow-model.yml").write_text(
+            """
+name: slow-model
+description: Slow model evaluator
+model: mistral/mistral-large-latest
+api_key_env: MISTRAL_API_KEY
+prompt: Evaluate this
+output_suffix: SLOW-EVAL
+timeout: 300
+"""
+        )
+
+        monkeypatch.chdir(tmp_path)
+
+        result = run_cli(["slow-model", "--help"], cwd=tmp_path)
+        assert result.returncode == 0
+        # Help should mention timeout flag with updated text
+        assert "--timeout" in result.stdout or "-t" in result.stdout
+        # Help text mentions evaluator config (may wrap across lines)
+        assert "evaluator config" in result.stdout
+        assert "max: 600" in result.stdout
+
+    def test_timeout_help_text_updated(self, tmp_path, monkeypatch, run_cli):
+        """Timeout help text shows it can come from config."""
+        adv_dir = tmp_path / ".adversarial"
+        adv_dir.mkdir(parents=True)
+        (adv_dir / "config.yml").write_text("log_directory: .adversarial/logs/")
+
+        monkeypatch.chdir(tmp_path)
+
+        result = run_cli(["evaluate", "--help"], cwd=tmp_path)
+        assert result.returncode == 0
+        # New help text mentioning evaluator config (may wrap across lines)
+        assert "evaluator config" in result.stdout
+        # Max 600 mentioned
+        assert "max: 600" in result.stdout

{adversarial_workflow-0.6.2 → adversarial_workflow-0.6.3}/tests/test_config.py

@@ -59,13 +59,16 @@ custom_setting: test_value
 
     def test_load_config_with_env_overrides(self):
         """Test that environment variables override config file values."""
-        with patch("os.path.exists", return_value=False), patch.dict(
-            os.environ,
-            {
-                "ADVERSARIAL_EVALUATOR_MODEL": "gpt-4-turbo",
-                "ADVERSARIAL_TEST_COMMAND": "cargo test",
-                "ADVERSARIAL_LOG_DIR": "custom_logs/",
-            },
+        with (
+            patch("os.path.exists", return_value=False),
+            patch.dict(
+                os.environ,
+                {
+                    "ADVERSARIAL_EVALUATOR_MODEL": "gpt-4-turbo",
+                    "ADVERSARIAL_TEST_COMMAND": "cargo test",
+                    "ADVERSARIAL_LOG_DIR": "custom_logs/",
+                },
+            ),
         ):
             config = load_config("nonexistent.yml")
 
@@ -127,13 +130,16 @@ test_command: pytest
 
     def test_load_config_partial_env_overrides(self):
         """Test that only set environment variables override config."""
-        with patch("os.path.exists", return_value=False), patch.dict(
-            os.environ,
-            {
-                "ADVERSARIAL_EVALUATOR_MODEL": "gpt-4",
-                # Only set one env var, others should remain default
-            },
-            clear=True,
+        with (
+            patch("os.path.exists", return_value=False),
+            patch.dict(
+                os.environ,
+                {
+                    "ADVERSARIAL_EVALUATOR_MODEL": "gpt-4",
+                    # Only set one env var, others should remain default
+                },
+                clear=True,
+            ),
         ):
             config = load_config("nonexistent.yml")
 

{adversarial_workflow-0.6.2 → adversarial_workflow-0.6.3}/tests/test_evaluate.py

@@ -272,14 +272,18 @@ class TestEvaluate:
         large_content = "# Test task\n" + "Line content\n" * 600
         task_file.write_text(large_content)
 
-        with patch("shutil.which", return_value="/usr/bin/aider"), patch(
-            "adversarial_workflow.cli.load_config",
-            return_value={"log_directory": ".adversarial/logs/"},
-        ), patch("os.path.exists", return_value=True), patch(
-            "adversarial_workflow.cli.validate_evaluation_output",
-            return_value=(True, "APPROVED", "OK"),
-        ), patch(
-            "adversarial_workflow.cli.verify_token_count"
+        with (
+            patch("shutil.which", return_value="/usr/bin/aider"),
+            patch(
+                "adversarial_workflow.cli.load_config",
+                return_value={"log_directory": ".adversarial/logs/"},
+            ),
+            patch("os.path.exists", return_value=True),
+            patch(
+                "adversarial_workflow.cli.validate_evaluation_output",
+                return_value=(True, "APPROVED", "OK"),
+            ),
+            patch("adversarial_workflow.cli.verify_token_count"),
         ):
             result = evaluate(str(task_file))
 
@@ -293,11 +297,14 @@ class TestEvaluate:
         very_large_content = "# Test task\n" + "Line content\n" * 800
         task_file.write_text(very_large_content)
 
-        with patch("shutil.which", return_value="/usr/bin/aider"), patch(
-            "adversarial_workflow.cli.load_config",
-            return_value={"log_directory": ".adversarial/logs/"},
-        ), patch("os.path.exists", return_value=True), patch(
-            "builtins.input", return_value="n"
+        with (
+            patch("shutil.which", return_value="/usr/bin/aider"),
+            patch(
+                "adversarial_workflow.cli.load_config",
+                return_value={"log_directory": ".adversarial/logs/"},
+            ),
+            patch("os.path.exists", return_value=True),
+            patch("builtins.input", return_value="n"),
         ):  # User says no
             result = evaluate(str(task_file))
             assert result == 0  # Cancelled, not error
@@ -450,14 +457,18 @@ class TestEvaluateIntegration:
 
     def test_evaluate_with_sample_task(self, sample_task_file, mock_aider_command):
         """Test evaluate with sample task file from fixture."""
-        with patch("shutil.which", return_value="/usr/bin/aider"), patch(
-            "adversarial_workflow.cli.load_config",
-            return_value={"log_directory": ".adversarial/logs/"},
-        ), patch("os.path.exists", return_value=True), patch(
-            "adversarial_workflow.cli.validate_evaluation_output",
-            return_value=(True, "APPROVED", "OK"),
-        ), patch(
-            "adversarial_workflow.cli.verify_token_count"
+        with (
+            patch("shutil.which", return_value="/usr/bin/aider"),
+            patch(
+                "adversarial_workflow.cli.load_config",
+                return_value={"log_directory": ".adversarial/logs/"},
+            ),
+            patch("os.path.exists", return_value=True),
+            patch(
+                "adversarial_workflow.cli.validate_evaluation_output",
+                return_value=(True, "APPROVED", "OK"),
+            ),
+            patch("adversarial_workflow.cli.verify_token_count"),
         ):
             result = evaluate(str(sample_task_file))
             assert isinstance(result, int)