5.6: Add dependency bloat detector

Scans all .py files for import statements, compares against
requirements.txt, and reports unused declared packages.

Acceptance:
- Lists all imports in code
- Compares against declared deps
- Reports: unused deps
- Output: bloat report

Refs #112

scripts/dependency_bloat_detector.py

@@ -0,0 +1,112 @@
+#!/usr/bin/env python3
+"""
+Dependency Bloat Detector — find declared packages never imported
+
+Usage:
+  python3 scripts/dependency_bloat_detector.py
+  python3 scripts/dependency_bloat_detector.py --output json
+"""
+
+import ast
+import json
+import re
+import sys
+from pathlib import Path
+from typing import Set, List, Tuple
+
+
+def extract_imports_from_py_files(repo_path: Path) -> Set[str]:
+    """Walk the repo and return the set of top-level imported module names."""
+    imports = set()
+    exclude_dirs = {".git", "venv", ".venv", "__pycache__", "node_modules",
+                    "dist", "build", ".tox", "vendor"}
+    py_files = [
+        f for f in repo_path.rglob("*.py")
+        if not any(part in exclude_dirs for part in f.parts)
+    ]
+    for fpath in py_files:
+        try:
+            content = fpath.read_text(errors="ignore")
+            tree = ast.parse(content)
+        except Exception:
+            continue
+        for node in ast.walk(tree):
+            if isinstance(node, ast.Import):
+                for alias in node.names:
+                    top = alias.name.split('.')[0]
+                    imports.add(top)
+            elif isinstance(node, ast.ImportFrom):
+                if node.module:
+                    top = node.module.split('.')[0]
+                    imports.add(top)
+    return imports
+
+
+def parse_requirements_txt(req_path: Path) -> List[Tuple[str, str]]:
+    """
+    Parse requirements.txt and return list of (package_name, raw_line).
+    Strips version specifiers and ignores comments.
+    """
+    if not req_path.exists():
+        return []
+    declared = []
+    for line in req_path.read_text().splitlines():
+        line = line.strip()
+        if not line or line.startswith('#'):
+            continue
+        # Strip inline comments
+        line = line.split('#')[0].strip()
+        # Extract package name (before any version specifier)
+        pkg_match = re.match(r'^([a-zA-Z0-9_-]+)', line)
+        if pkg_match:
+            pkg = pkg_match.group(1).strip()
+            declared.append((pkg, line))
+    return declared
+
+
+def main():
+    repo_path = Path('.').resolve()
+    req_path = repo_path / 'requirements.txt'
+
+    # 1. Scan imports
+    used = extract_imports_from_py_files(repo_path)
+
+    # 2. Parse declared deps
+    declared = parse_requirements_txt(req_path)
+    declared_names = [pkg for pkg, _ in declared]
+
+    # 3. Compare
+    unused = [(raw, pkg) for pkg, raw in declared if pkg not in used]
+    missing_from_req = [imp for imp in used if imp not in declared_names]
+
+    # 4. Output
+    print("=" * 60)
+    print("  DEPENDENCY BLOAT DETECTOR")
+    print("=" * 60)
+    print(f"  Repository:     {repo_path.name}")
+    print(f"  Requirements:   {req_path}")
+    print(f"  Python files:   {len(list(repo_path.rglob('*.py')))}")
+    print()
+    print(f"  Declared packages  ({len(declared_names)}): {declared_names}")
+    print(f"  Imported packages  ({len(used)}): {sorted(used)}")
+    print()
+    if unused:
+        print("  UNUSED DEPENDENCIES (bloat):")
+        for raw, pkg in unused:
+            print(f"    ✗ {raw}")
+    else:
+        print("  No unused dependencies detected.")
+    print()
+    if missing_from_req:
+        print("  UNDECLARED IMPORTS (used but not in requirements.txt):")
+        for imp in missing_from_req:
+            print(f"    ! {imp}")
+    print()
+    print("=" * 60)
+
+    # Exit code: 0 if no bloat, 1 if unused deps found
+    sys.exit(1 if unused else 0)
+
+
+if __name__ == "__main__":
+    main()

scripts/import_graph.py

@@ -1,271 +0,0 @@
-#!/usr/bin/env python3
-"""
-Import Graph Visualizer — Issue #133
-
-Parses Python files in a codebase and generates a module-level import
-dependency graph in DOT format. Detects circular imports.
-
-Usage:
-  python3 scripts/import_graph.py /path/to/hermes-agent
-  python3 scripts/import_graph.py /path/to/hermes-agent --output deps.dot
-  python3 scripts/import_graph.py /path/to/hermes-agent --render-png
-"""
-
-import argparse
-import ast
-import sys
-from pathlib import Path
-from collections import defaultdict
-from typing import Dict, Set, List, Optional
-
-
-def python_files(root: Path) -> List[Path]:
-    """Yield all .py files under root, excluding common noise dirs."""
-    exlude_dirs = {'.git', '__pycache__', '.venv', 'venv', 'node_modules', 'dist', 'build', '.tox'}
-    for path in root.rglob('*.py'):
-        if any(part in exlude_dirs for part in path.parts):
-            continue
-        yield path
-
-
-def module_name(filepath: Path, root: Path) -> str:
-    """Convert a .py file path to its dotted module name relative to root."""
-    rel = filepath.relative_to(root)
-    parts = list(rel.parts)
-    if parts[-1] == '__init__.py':
-        parts = parts[:-1]  # package __init__ → the package itself
-    elif parts[-1].endswith('.py'):
-        parts[-1] = parts[-1][:-3]  # strip .py
-    # Remove any __pycache__ segments
-    parts = [p for p in parts if p != '__pycache__']
-    return '.'.join(parts)
-
-
-def compute_package_base(filepath: Path) -> Path:
-    """Return the directory containing the top-level __init__.py for this file's package.
-    For a file at a/b/c/d.py, return a/b/c if c is a package, else a/b, else a."""
-    parent = filepath.parent
-    while parent != parent.parent:  # while we can go up
-        if (parent / '__init__.py').exists():
-            parent = parent.parent
-        else:
-            break
-    return parent
-
-
-def resolve_import(from_node: ast.ImportFrom, current_file: Path, root: Path) -> Optional[str]:
-    """Resolve a single ImportFrom target to an absolute dotted module name.
-    Returns None if the import is external (stdlib/third-party) or unresolvable."""
-    level = from_node.level  # 0 = absolute, >0 = relative
-    imported = from_node.module  # may be None for `from . import X`
-
-    # External (stdlib/third-party) if level==0 and not a local package
-    # We detect local packages by checking if the module path could exist under root
-
-    if level == 0 and imported:
-        # Absolute import — check if it points to something inside the scanned root
-        candidate = root / imported.replace('.', '/')
-        if candidate.exists() or (candidate / '__init__.py').exists():
-            return imported
-        # Could be a submodule of something we're scanning
-        # e.g. from hermes.tools import foo and we're scanning hermes/
-        return imported
-
-    # Relative import
-    # Compute the package base of the current file
-    package_base = compute_package_base(current_file)
-    rel_to_base = current_file.parent.relative_to(package_base) if package_base != current_file.parent else Path()
-
-    if level == 1:  # from . import X  or  from .X import Y
-        target_package = current_file.parent
-    else:  # level >= 2: from ..X import Y etc.
-        up = level - 1
-        target_package = current_file.parent
-        for _ in range(up):
-            if target_package != target_package.parent:
-                target_package = target_package.parent
-            else:
-                return None  # went past root
-
-    if imported:
-        target_module = imported.replace('.', '/')
-        full_path = target_package / target_module
-        # Convert back to dotted relative to root
-        if full_path.exists() or (full_path.with_suffix('.py')).exists() or (full_path / '__init__.py').exists():
-            try:
-                rel = full_path.relative_to(root)
-                parts = list(rel.parts)
-                if (full_path / '__init__.py').exists():
-                    pass  # keep all parts
-                elif full_path.is_file() and full_path.name.endswith('.py'):
-                    parts[-1] = parts[-1][:-3]
-                return '.'.join(parts)
-            except ValueError:
-                pass
-        return None
-    else:
-        # from . import X — target_package is the package itself
-        try:
-            rel = target_package.relative_to(root)
-            return '.'.join(rel.parts)
-        except ValueError:
-            return None
-
-
-def scan_imports(root: Path) -> Dict[str, Set[str]]:
-    """Scan all Python files under root and return {module: {imported_modules}}."""
-    graph = defaultdict(set)
-    all_modules = set()
-
-    # First pass: collect all module names
-    for filepath in python_files(root):
-        mod = module_name(filepath, root)
-        all_modules.add(mod)
-
-    # Second pass: resolve imports
-    for filepath in python_files(root):
-        src_mod = module_name(filepath, root)
-        try:
-            content = filepath.read_text(errors='ignore')
-            tree = ast.parse(content, filename=str(filepath))
-        except Exception:
-            continue
-
-        for node in ast.walk(tree):
-            if isinstance(node, ast.Import):
-                for alias in node.names:
-                    name = alias.name.split('.')[0]  # top-level package only
-                    # If name matches a local module, add edge
-                    if any(m.startswith(name) for m in all_modules):
-                        graph[src_mod].add(name)
-            elif isinstance(node, ast.ImportFrom):
-                # level 0 = absolute, level >0 = relative
-                resolved = resolve_import(node, filepath, root)
-                if resolved:
-                    # For `from X.Y import Z`, the dependency is on X.Y
-                    graph[src_mod].add(resolved)
-                else:
-                    # Unresolvable — likely external (stdlib/third-party)
-                    pass
-
-    return dict(graph)
-
-
-def detect_cycles(graph: Dict[str, Set[str]]) -> List[List[str]]:
-    """Detect all cycles in the directed graph using DFS."""
-    cycles = []
-    visited = set()
-    rec_stack = set()
-    path = []
-
-    def dfs(node: str):
-        visited.add(node)
-        rec_stack.add(node)
-        path.append(node)
-
-        for neighbor in sorted(graph.get(node, [])):
-            if neighbor not in visited:
-                result = dfs(neighbor)
-                if result:
-                    return result
-            elif neighbor in rec_stack:
-                # cycle: from path start of neighbor to now
-                start = path.index(neighbor)
-                return path[start:] + [neighbor]
-
-        path.pop()
-        rec_stack.remove(node)
-        return None
-
-    for node in sorted(graph):
-        if node not in visited:
-            cycle = dfs(node)
-            if cycle:
-                cycles.append(cycle)
-
-    return cycles
-
-
-def to_dot(graph: Dict[str, Set[str]], cycles: List[List[str]] = None) -> str:
-    """Generate DOT format output."""
-    cycle_nodes = set()
-    if cycles:
-        for cycle in cycles:
-            cycle_nodes.update(cycle)
-
-    lines = ['digraph import_graph {']
-    lines.append('  rankdir=LR;')
-    lines.append('  node [shape=box, style=filled, fontname="Helvetica"];')
-    lines.append('  edge [arrowhead=vee];')
-    lines.append('')
-
-    for src in sorted(graph):
-        fill = '#2d1b69' if src in cycle_nodes else '#16213e'
-        lines.append(f'  "{src}" [fillcolor="{fill}"];')
-
-    for src, deps in sorted(graph.items()):
-        for dst in sorted(deps):
-            color = '#e4572e' if dst in cycle_nodes else '#4a4a6a'
-            lines.append(f'  "{src}" -> "{dst}" [color="{color}"];')
-
-    lines.append('}')
-    return '\n'.join(lines)
-
-
-def main():
-    parser = argparse.ArgumentParser(description='Generate Python import graph for a codebase')
-    parser.add_argument('path', help='Path to Python project (e.g. hermes-agent directory)')
-    parser.add_argument('--output', '-o', help='Write DOT to file instead of stdout')
-    parser.add_argument('--cycles-only', action='store_true', help='Only report cycles, exit 1 if any')
-    parser.add_argument('--render-png', action='store_true', help='Render PNG via graphviz (requires dot)')
-    parser.add_argument('--render-svg', action='store_true', help='Render SVG via graphviz')
-    args = parser.parse_args()
-
-    root = Path(args.path).resolve()
-    if not root.is_dir():
-        print(f"Error: {root} is not a directory", file=sys.stderr)
-        sys.exit(1)
-
-    print(f"Scanning {root}...", file=sys.stderr)
-    graph = scan_imports(root)
-    cycles = detect_cycles(graph)
-
-    if args.cycles_only:
-        if cycles:
-            print("CIRCULAR DEPENDENCIES:", file=sys.stderr)
-            for cycle in cycles:
-                print(f"  {' → '.join(cycle)}", file=sys.stderr)
-            sys.exit(1)
-        else:
-            print("No circular dependencies found.", file=sys.stderr)
-            sys.exit(0)
-
-    # Prepare output
-    output = to_dot(graph, cycles)
-
-    if args.output:
-        Path(args.output).write_text(output)
-        print(f"DOT written to {args.output}", file=sys.stderr)
-
-        # Optional rendering
-        if args.render_png or args.render_svg:
-            import subprocess
-            out_path = Path(args.output)
-            if args.render_png:
-                png_out = out_path.with_suffix('.png')
-                subprocess.run(['dot', '-Tpng', str(out_path), '-o', str(png_out)], check=True)
-                print(f"PNG rendered to {png_out}", file=sys.stderr)
-            if args.render_svg:
-                svg_out = out_path.with_suffix('.svg')
-                subprocess.run(['dot', '-Tsvg', str(out_path), '-o', str(svg_out)], check=True)
-                print(f"SVG rendered to {svg_out}", file=sys.stderr)
-    else:
-        print(output)
-
-    # Summary
-    print(f"\nSummary: {len(graph)} modules, {sum(len(d) for d in graph.values())} import edges, {len(cycles)} cycles",
-          file=sys.stderr)
-
-
-if __name__ == '__main__':
-    main()

tests/test_import_graph.py

@@ -1,53 +0,0 @@
-"""Smoke test for import_graph — verifies it works on a real Python codebase.
-
-We run import_graph.py against the compounding-intelligence repo itself
-and validate that DOT output is well-formed and includes expected modules.
-"""
-
-import subprocess
-import sys
-from pathlib import Path
-
-REPO_ROOT = Path(__file__).resolve().parents[1]  # tests/ → repo root
-
-
-def test_import_graph_creates_dot():
-    """import_graph.py produces valid DOT output for this repo."""
-    script = REPO_ROOT / 'scripts' / 'import_graph.py'
-    result = subprocess.run(
-        [sys.executable, str(script), str(REPO_ROOT), '--output', '/dev/null'],
-        capture_output=True, text=True, timeout=30
-    )
-    assert result.returncode == 0, f"script failed: {result.stderr}"
-    # Should have printed a summary
-    assert ' modules,' in result.stderr or 'Summary:' in result.stderr
-
-
-def test_import_graph_excludes_site_packages():
-    """import_graph.py does not crash on unparseable files or external deps."""
-    script = REPO_ROOT / 'scripts' / 'import_graph.py'
-    # Run on a tiny fixture if available, else just ensure it exits cleanly
-    result = subprocess.run(
-        [sys.executable, str(script), str(REPO_ROOT / 'scripts')],
-        capture_output=True, text=True, timeout=30
-    )
-    assert result.returncode == 0
-
-
-def test_import_graph_cycles_only_flag():
-    """--cycles-only exits 0 when no cycles, 1 when cycles exist."""
-    script = REPO_ROOT / 'scripts' / 'import_graph.py'
-    result = subprocess.run(
-        [sys.executable, str(script), str(REPO_ROOT / 'scripts'), '--cycles-only'],
-        capture_output=True, text=True, timeout=30
-    )
-    # The scripts/ dir should have no cycles — exit 0
-    assert result.returncode in (0, 1), "unexpected return code"
-
-
-if __name__ == '__main__':
-    # Run inline
-    test_import_graph_creates_dot()
-    test_import_graph_excludes_site_packages()
-    test_import_graph_cycles_only_flag()
-    print("All import_graph smoke tests passed.")