adversarial-workflow 0.6.0__tar.gz → 0.6.1__tar.gz

{adversarial_workflow-0.6.0 → adversarial_workflow-0.6.1}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: adversarial-workflow
-Version: 0.6.0
+Version: 0.6.1
 Summary: Multi-stage AI code review system preventing phantom work - Author/Evaluator pattern
 Author: Fredrik Matheson
 License: MIT
@@ -35,6 +35,10 @@ Dynamic: license-file
 
 # Adversarial Workflow
 
+[![PyPI version](https://badge.fury.io/py/adversarial-workflow.svg)](https://pypi.org/project/adversarial-workflow/)
+[![Python 3.10+](https://img.shields.io/badge/python-3.10+-blue.svg)](https://www.python.org/downloads/)
+[![License: MIT](https://img.shields.io/badge/License-MIT-yellow.svg)](https://opensource.org/licenses/MIT)
+
 **A multi-stage AI code review system that makes your code better**
 
 Evaluate proposals, sort out ideas, and prevent "phantom work" (AI claiming to implement but not delivering) through adversarial verification using independent review stages. A battle-tested workflow from the [thematic-cuts](https://github.com/movito/thematic-cuts) project that achieved 96.9% test pass rate improvement.
@@ -51,6 +55,31 @@ 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
+
+🔌 **Plugin Architecture** - Define custom evaluators without modifying the package:
+
+```bash
+# Create a custom evaluator
+mkdir -p .adversarial/evaluators
+cat > .adversarial/evaluators/athena.yml << 'EOF'
+name: athena
+description: Knowledge evaluation using Gemini 2.5 Pro
+model: gemini-2.5-pro
+api_key_env: GEMINI_API_KEY
+prompt: |
+  You are Athena, a knowledge evaluation specialist...
+EOF
+
+# Use it immediately
+adversarial athena docs/research-plan.md
+
+# List all available evaluators
+adversarial list-evaluators
+```
+
+See [Custom Evaluators](#custom-evaluators) for full documentation, or check the [CHANGELOG](CHANGELOG.md) for complete release history.
+
 ## Prerequisites
 
 Before installing, ensure you have:
@@ -856,12 +885,13 @@ From the [thematic-cuts](https://github.com/movito/thematic-cuts) project:
 
 ## Documentation
 
-- **Interaction Patterns**: How Author-Reviewer collaboration works
+- **[Custom Evaluators Guide](docs/CUSTOM_EVALUATORS.md)**: Create project-specific evaluators
+- **[Integration Guide](docs/INTEGRATION-GUIDE.md)**: Detailed integration strategies
+- **[CHANGELOG](CHANGELOG.md)**: Release history and version notes
+- **Interaction Patterns**: How Author-Evaluator collaboration works
 - **Token Optimization**: Detailed Aider configuration guide
 - **Workflow Phases**: Step-by-step guide for each phase
 - **Troubleshooting**: Common issues and solutions
-- **Examples**: Real integration scenarios
-- **Terminology**: Official standards for Author/Reviewer concepts
 
 See `docs/` directory for comprehensive guides.
 

{adversarial_workflow-0.6.0 → adversarial_workflow-0.6.1}/README.md

@@ -1,5 +1,9 @@
 # Adversarial Workflow
 
+[![PyPI version](https://badge.fury.io/py/adversarial-workflow.svg)](https://pypi.org/project/adversarial-workflow/)
+[![Python 3.10+](https://img.shields.io/badge/python-3.10+-blue.svg)](https://www.python.org/downloads/)
+[![License: MIT](https://img.shields.io/badge/License-MIT-yellow.svg)](https://opensource.org/licenses/MIT)
+
 **A multi-stage AI code review system that makes your code better**
 
 Evaluate proposals, sort out ideas, and prevent "phantom work" (AI claiming to implement but not delivering) through adversarial verification using independent review stages. A battle-tested workflow from the [thematic-cuts](https://github.com/movito/thematic-cuts) project that achieved 96.9% test pass rate improvement.
@@ -16,6 +20,31 @@ 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
+
+🔌 **Plugin Architecture** - Define custom evaluators without modifying the package:
+
+```bash
+# Create a custom evaluator
+mkdir -p .adversarial/evaluators
+cat > .adversarial/evaluators/athena.yml << 'EOF'
+name: athena
+description: Knowledge evaluation using Gemini 2.5 Pro
+model: gemini-2.5-pro
+api_key_env: GEMINI_API_KEY
+prompt: |
+  You are Athena, a knowledge evaluation specialist...
+EOF
+
+# Use it immediately
+adversarial athena docs/research-plan.md
+
+# List all available evaluators
+adversarial list-evaluators
+```
+
+See [Custom Evaluators](#custom-evaluators) for full documentation, or check the [CHANGELOG](CHANGELOG.md) for complete release history.
+
 ## Prerequisites
 
 Before installing, ensure you have:
@@ -821,12 +850,13 @@ From the [thematic-cuts](https://github.com/movito/thematic-cuts) project:
 
 ## Documentation
 
-- **Interaction Patterns**: How Author-Reviewer collaboration works
+- **[Custom Evaluators Guide](docs/CUSTOM_EVALUATORS.md)**: Create project-specific evaluators
+- **[Integration Guide](docs/INTEGRATION-GUIDE.md)**: Detailed integration strategies
+- **[CHANGELOG](CHANGELOG.md)**: Release history and version notes
+- **Interaction Patterns**: How Author-Evaluator collaboration works
 - **Token Optimization**: Detailed Aider configuration guide
 - **Workflow Phases**: Step-by-step guide for each phase
 - **Troubleshooting**: Common issues and solutions
-- **Examples**: Real integration scenarios
-- **Terminology**: Official standards for Author/Reviewer concepts
 
 See `docs/` directory for comprehensive guides.
 

{adversarial_workflow-0.6.0 → adversarial_workflow-0.6.1}/adversarial_workflow/__init__.py

@@ -12,7 +12,7 @@ Usage:
     adversarial validate "pytest"
 """
 
-__version__ = "0.6.0"
+__version__ = "0.6.1"
 __author__ = "Fredrik Matheson"
 __license__ = "MIT"
 

{adversarial_workflow-0.6.0 → adversarial_workflow-0.6.1}/adversarial_workflow/cli.py

@@ -27,9 +27,9 @@ from pathlib import Path
 from typing import Dict, List, Optional, Tuple
 
 import yaml
-from dotenv import load_dotenv
+from dotenv import load_dotenv, dotenv_values
 
-__version__ = "0.6.0"
+__version__ = "0.6.1"
 
 # ANSI color codes for better output
 RESET = "\033[0m"
@@ -800,26 +800,37 @@ def check() -> int:
     issues: List[Dict] = []
     good_checks: List[str] = []
 
-    # Check for .env file first (before loading environment variables)
+    # Check for .env file (note: already loaded by main() at startup)
     env_file = Path(".env")
     env_loaded = False
-    env_keys_before = set(os.environ.keys())
 
     if env_file.exists():
         try:
+            # Load .env into environment (idempotent - safe to call again after main())
             load_dotenv(env_file)
-            env_keys_after = set(os.environ.keys())
-            new_keys = env_keys_after - env_keys_before
+            # Use dotenv_values() to count variables directly from file
+            # This gives accurate count regardless of what was already in environment
+            env_vars = dotenv_values(env_file)
             env_loaded = True
             good_checks.append(
-                f".env file found and loaded ({len(new_keys)} variables)"
+                f".env file found ({len(env_vars)} variables configured)"
             )
-        except Exception as e:
+        except (FileNotFoundError, PermissionError) as e:
+            # File access errors
             issues.append(
                 {
                     "severity": "WARNING",
-                    "message": f".env file found but could not be loaded: {e}",
-                    "fix": "Check .env file format and permissions",
+                    "message": f".env file found but could not be read: {e}",
+                    "fix": "Check .env file permissions",
+                }
+            )
+        except (OSError, ValueError) as e:
+            # Covers UnicodeDecodeError (ValueError subclass) and other OS errors
+            issues.append(
+                {
+                    "severity": "WARNING",
+                    "message": f".env file found but could not be parsed: {e}",
+                    "fix": "Check .env file encoding (should be UTF-8)",
                 }
             )
     else:
@@ -2868,6 +2879,14 @@ def list_evaluators() -> int:
 def main():
     """Main CLI entry point."""
     import logging
+    import sys
+
+    # Load .env file before any commands run
+    # Wrapped in try/except so CLI remains usable even with malformed .env
+    try:
+        load_dotenv()
+    except Exception as e:
+        print(f"Warning: Could not load .env file: {e}", file=sys.stderr)
 
     from adversarial_workflow.evaluators import (
         get_all_evaluators,

{adversarial_workflow-0.6.0 → adversarial_workflow-0.6.1}/adversarial_workflow.egg-info/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: adversarial-workflow
-Version: 0.6.0
+Version: 0.6.1
 Summary: Multi-stage AI code review system preventing phantom work - Author/Evaluator pattern
 Author: Fredrik Matheson
 License: MIT
@@ -35,6 +35,10 @@ Dynamic: license-file
 
 # Adversarial Workflow
 
+[![PyPI version](https://badge.fury.io/py/adversarial-workflow.svg)](https://pypi.org/project/adversarial-workflow/)
+[![Python 3.10+](https://img.shields.io/badge/python-3.10+-blue.svg)](https://www.python.org/downloads/)
+[![License: MIT](https://img.shields.io/badge/License-MIT-yellow.svg)](https://opensource.org/licenses/MIT)
+
 **A multi-stage AI code review system that makes your code better**
 
 Evaluate proposals, sort out ideas, and prevent "phantom work" (AI claiming to implement but not delivering) through adversarial verification using independent review stages. A battle-tested workflow from the [thematic-cuts](https://github.com/movito/thematic-cuts) project that achieved 96.9% test pass rate improvement.
@@ -51,6 +55,31 @@ 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
+
+🔌 **Plugin Architecture** - Define custom evaluators without modifying the package:
+
+```bash
+# Create a custom evaluator
+mkdir -p .adversarial/evaluators
+cat > .adversarial/evaluators/athena.yml << 'EOF'
+name: athena
+description: Knowledge evaluation using Gemini 2.5 Pro
+model: gemini-2.5-pro
+api_key_env: GEMINI_API_KEY
+prompt: |
+  You are Athena, a knowledge evaluation specialist...
+EOF
+
+# Use it immediately
+adversarial athena docs/research-plan.md
+
+# List all available evaluators
+adversarial list-evaluators
+```
+
+See [Custom Evaluators](#custom-evaluators) for full documentation, or check the [CHANGELOG](CHANGELOG.md) for complete release history.
+
 ## Prerequisites
 
 Before installing, ensure you have:
@@ -856,12 +885,13 @@ From the [thematic-cuts](https://github.com/movito/thematic-cuts) project:
 
 ## Documentation
 
-- **Interaction Patterns**: How Author-Reviewer collaboration works
+- **[Custom Evaluators Guide](docs/CUSTOM_EVALUATORS.md)**: Create project-specific evaluators
+- **[Integration Guide](docs/INTEGRATION-GUIDE.md)**: Detailed integration strategies
+- **[CHANGELOG](CHANGELOG.md)**: Release history and version notes
+- **Interaction Patterns**: How Author-Evaluator collaboration works
 - **Token Optimization**: Detailed Aider configuration guide
 - **Workflow Phases**: Step-by-step guide for each phase
 - **Troubleshooting**: Common issues and solutions
-- **Examples**: Real integration scenarios
-- **Terminology**: Official standards for Author/Reviewer concepts
 
 See `docs/` directory for comprehensive guides.
 

{adversarial_workflow-0.6.0 → adversarial_workflow-0.6.1}/adversarial_workflow.egg-info/SOURCES.txt

@@ -38,6 +38,7 @@ adversarial_workflow/utils/validation.py
 tests/test_cli.py
 tests/test_cli_dynamic_commands.py
 tests/test_config.py
+tests/test_env_loading.py
 tests/test_evaluate.py
 tests/test_evaluator_config.py
 tests/test_evaluator_discovery.py

{adversarial_workflow-0.6.0 → adversarial_workflow-0.6.1}/pyproject.toml

@@ -4,7 +4,7 @@ build-backend = "setuptools.build_meta"
 
 [project]
 name = "adversarial-workflow"
-version = "0.6.0"
+version = "0.6.1"
 description = "Multi-stage AI code review system preventing phantom work - Author/Evaluator pattern"
 readme = "README.md"
 authors = [

{adversarial_workflow-0.6.0 → adversarial_workflow-0.6.1}/tests/test_cli.py

@@ -26,7 +26,7 @@ class TestCLISmoke:
             text=True,
         )
         assert result.returncode == 0
-        assert "0.6.0" in result.stdout or "0.6.0" in result.stderr
+        assert "0.6.1" in result.stdout or "0.6.1" in result.stderr
 
     def test_help_flag(self):
         """Test that --help returns help text."""

adversarial_workflow-0.6.1/tests/test_env_loading.py

@@ -0,0 +1,214 @@
+"""Tests for .env file loading at CLI startup."""
+
+import os
+import subprocess
+import sys
+
+
+class TestEnvFileLoading:
+    """Tests for automatic .env loading."""
+
+    def test_env_var_available_via_cli_check(self, tmp_path):
+        """Verify .env file is loaded when CLI commands run."""
+        # Create .env with OPENAI_API_KEY
+        (tmp_path / ".env").write_text("OPENAI_API_KEY=sk-test-env-loading\n")
+
+        # Run check command which validates OPENAI_API_KEY
+        # Remove OPENAI_API_KEY from environment to ensure it comes from .env
+        env = {k: v for k, v in os.environ.items() if k != "OPENAI_API_KEY"}
+        env["PATH"] = os.environ.get("PATH", "")
+
+        result = subprocess.run(
+            [sys.executable, "-m", "adversarial_workflow.cli", "check"],
+            capture_output=True,
+            text=True,
+            cwd=tmp_path,
+            env=env,
+        )
+
+        # The check command should see the API key from .env
+        # It will show as valid (green checkmark) in the output
+        combined_output = result.stdout + result.stderr
+        assert "OPENAI_API_KEY" in combined_output, (
+            f"Expected OPENAI_API_KEY check. stdout: {result.stdout}, stderr: {result.stderr}"
+        )
+
+    def test_env_loaded_before_evaluator_commands(self, tmp_path, monkeypatch):
+        """API keys in .env are available to evaluator commands."""
+        # Create .env with test key
+        (tmp_path / ".env").write_text("TEST_API_KEY=secret-test-value\n")
+
+        # Create minimal evaluator config
+        eval_dir = tmp_path / ".adversarial" / "evaluators"
+        eval_dir.mkdir(parents=True)
+        (eval_dir / "test.yml").write_text("""name: test
+description: Test evaluator
+model: gpt-4o-mini
+api_key_env: TEST_API_KEY
+prompt: Test prompt
+output_suffix: TEST
+""")
+
+        monkeypatch.chdir(tmp_path)
+        # Ensure key is NOT in current environment
+        monkeypatch.delenv("TEST_API_KEY", raising=False)
+
+        # list-evaluators should work (loads .env, discovers evaluator)
+        result = subprocess.run(
+            [sys.executable, "-m", "adversarial_workflow.cli", "list-evaluators"],
+            capture_output=True,
+            text=True,
+            cwd=tmp_path,
+        )
+
+        assert result.returncode == 0
+        assert "test" in result.stdout
+
+    def test_env_loaded_for_builtin_commands(self, tmp_path, monkeypatch):
+        """.env is loaded even for built-in commands."""
+        # Create .env with OpenAI key
+        (tmp_path / ".env").write_text("OPENAI_API_KEY=sk-test-key\n")
+
+        monkeypatch.chdir(tmp_path)
+        monkeypatch.delenv("OPENAI_API_KEY", raising=False)
+
+        # check command should find the key from .env
+        result = subprocess.run(
+            [sys.executable, "-m", "adversarial_workflow.cli", "check"],
+            capture_output=True,
+            text=True,
+            cwd=tmp_path,
+        )
+
+        # Should mention OpenAI (found from .env)
+        # The check may fail for other reasons but should see the key
+        assert "OPENAI" in result.stdout or "openai" in result.stdout.lower()
+
+    def test_missing_env_file_no_error(self, tmp_path, monkeypatch):
+        """CLI works fine when no .env file exists."""
+        monkeypatch.chdir(tmp_path)
+
+        result = subprocess.run(
+            [sys.executable, "-m", "adversarial_workflow.cli", "--help"],
+            capture_output=True,
+            text=True,
+            cwd=tmp_path,
+        )
+
+        assert result.returncode == 0
+        assert "adversarial" in result.stdout.lower()
+
+
+class TestCheckEnvCount:
+    """Tests for check() command .env variable counting (ADV-0022).
+
+    These are CLI integration tests using subprocess to verify end-to-end behavior.
+    They test that check() correctly reports .env variable count even after
+    main() has already loaded the .env file at startup.
+    """
+
+    def test_check_reports_correct_env_count(self, tmp_path):
+        """check() reports correct .env variable count even after main() loads it.
+
+        This is the primary regression test for ADV-0022. Before the fix,
+        check() would report "0 variables" because main() already loaded them.
+        """
+        # Create .env with 3 variables
+        (tmp_path / ".env").write_text(
+            "OPENAI_API_KEY=sk-test\n"
+            "ANTHROPIC_API_KEY=ant-test\n"
+            "CUSTOM_KEY=custom-value\n"
+        )
+
+        # Remove keys from environment to isolate test
+        env = {k: v for k, v in os.environ.items()
+               if k not in ("OPENAI_API_KEY", "ANTHROPIC_API_KEY", "CUSTOM_KEY")}
+        env["PATH"] = os.environ.get("PATH", "")
+
+        result = subprocess.run(
+            [sys.executable, "-m", "adversarial_workflow.cli", "check"],
+            capture_output=True,
+            text=True,
+            cwd=tmp_path,
+            env=env,
+        )
+
+        # Should report "3 variables configured", not "0 variables"
+        assert "3 variables" in result.stdout, (
+            f"Expected '3 variables' in output. Got: {result.stdout}"
+        )
+
+    def test_check_handles_empty_env_file(self, tmp_path):
+        """check() handles empty .env file gracefully."""
+        (tmp_path / ".env").write_text("")
+
+        env = {k: v for k, v in os.environ.items()}
+        env["PATH"] = os.environ.get("PATH", "")
+
+        result = subprocess.run(
+            [sys.executable, "-m", "adversarial_workflow.cli", "check"],
+            capture_output=True,
+            text=True,
+            cwd=tmp_path,
+            env=env,
+        )
+
+        # Should report "0 variables configured"
+        assert "0 variables" in result.stdout, (
+            f"Expected '0 variables' in output. Got: {result.stdout}"
+        )
+
+    def test_check_handles_comments_in_env(self, tmp_path):
+        """check() correctly counts variables, ignoring comments and empty lines."""
+        (tmp_path / ".env").write_text(
+            "# This is a comment\n"
+            "KEY1=value1\n"
+            "\n"  # Empty line
+            "# Another comment\n"
+            "KEY2=value2\n"
+        )
+
+        env = {k: v for k, v in os.environ.items() if k not in ("KEY1", "KEY2")}
+        env["PATH"] = os.environ.get("PATH", "")
+
+        result = subprocess.run(
+            [sys.executable, "-m", "adversarial_workflow.cli", "check"],
+            capture_output=True,
+            text=True,
+            cwd=tmp_path,
+            env=env,
+        )
+
+        # Should report 2 variables (comments and empty lines ignored)
+        assert "2 variables" in result.stdout, (
+            f"Expected '2 variables' in output. Got: {result.stdout}"
+        )
+
+    def test_check_handles_unusual_env_entries(self, tmp_path):
+        """check() handles unusual .env entries without crashing.
+
+        dotenv_values() treats 'KEY' without = as key with None value.
+        This test verifies the CLI doesn't crash on such inputs.
+        """
+        (tmp_path / ".env").write_text(
+            "VALID_KEY=value\n"
+            "ALSO_VALID=another\n"
+            "KEY_WITHOUT_VALUE\n"  # dotenv treats this as KEY_WITHOUT_VALUE=None
+        )
+
+        env = {k: v for k, v in os.environ.items()
+               if k not in ("VALID_KEY", "ALSO_VALID", "KEY_WITHOUT_VALUE")}
+        env["PATH"] = os.environ.get("PATH", "")
+
+        result = subprocess.run(
+            [sys.executable, "-m", "adversarial_workflow.cli", "check"],
+            capture_output=True,
+            text=True,
+            cwd=tmp_path,
+            env=env,
+        )
+
+        # Should not crash - dotenv_values() returns 3 entries (including key with None value)
+        assert "3 variables" in result.stdout, (
+            f"Expected '3 variables' in output. Got: {result.stdout}"
+        )