mindfull / analyze_codebase.py
IamSamk
Mindfull Gradio Space deploy
27caffe
Raw
History Blame Contribute Delete
41.6 kB
"""
Codebase Call Graph Analyzer
Traces function calls, assets (model checkpoints, audio, images, configs),
URLs, environment variables, and subprocess commands across all Python files.
Usage: python analyze_codebase.py
"""
import ast
import os
import re
import sys
from pathlib import Path
from collections import defaultdict
from typing import Dict, Set, List, Tuple, Optional
# ─── Asset extension groups ───────────────────────────────────────────────────
ASSET_EXTS = {
"model": {".pth", ".pt", ".ckpt", ".safetensors", ".bin", ".onnx", ".mar"},
"audio": {".wav", ".mp3", ".flac", ".ogg", ".m4a"},
"video": {".mp4", ".avi", ".mov", ".mkv"},
"image": {".png", ".jpg", ".jpeg", ".bmp", ".gif", ".webp"},
"config": {".json", ".yaml", ".yml", ".cfg", ".ini", ".toml", ".txt"},
"data": {".csv", ".npy", ".npz", ".pkl", ".h5", ".hdf5"},
}
# ─── Configuration ────────────────────────────────────────────────────────────
ROOT = Path(__file__).parent
# Directories to skip entirely
SKIP_DIRS = {"venv", "__pycache__", ".git", "node_modules", "gfpgan"}
# Starting entry points (traced in order)
ENTRY_POINTS = [
"mindfull_web_api.py",
"mindfull_pipeline.py",
"mindfull_config.py",
"simple_video_gen.py",
"simple_audio_gen.py",
"sadtalker+wav2lip/enhanced_pipeline.py",
"sadtalker+wav2lip/simple_pipeline.py",
"sadtalker+wav2lip/wav2lip/inference.py",
"sadtalker+wav2lip/sadtalker/inference.py",
"python313_compat_patch.py",
"model_manager.py",
"ollama_to_f5tts.py",
"infer_cli.py",
]
# ─── Data Structures ──────────────────────────────────────────────────────────
class FunctionInfo:
def __init__(self, name: str, file: str, lineno: int, is_method: bool = False, class_name: str = None):
self.name = name
self.file = file
self.lineno = lineno
self.is_method = is_method
self.class_name = class_name
self.calls: List[str] = [] # raw call names seen in body
self.docstring: Optional[str] = None
def full_name(self):
if self.class_name:
return f"{self.class_name}.{self.name}"
return self.name
def __repr__(self):
return f"<Func {self.full_name()} @ {self.file}:{self.lineno}>"
class AssetRef:
"""A non-Python file referenced as a string literal in source code."""
def __init__(self, path_str: str, asset_type: str, lineno: int, in_func: Optional[str]):
self.path_str = path_str # raw string from source
self.asset_type = asset_type # model / audio / image / config / …
self.lineno = lineno
self.in_func = in_func # function name, or None for module-level
self.exists = Path(path_str).exists() or Path(ROOT / path_str).exists()
def __repr__(self):
return f"<Asset {self.asset_type}:{self.path_str}>"
class FileInfo:
def __init__(self, path: str):
self.path = path # relative path from ROOT
self.imports: Dict[str, str] = {} # alias -> module string
self.from_imports: Dict[str, Tuple[str, str]] = {} # name -> (module, original_name)
self.functions: List[FunctionInfo] = []
self.classes: List[str] = []
self.top_level_calls: List[str] = [] # calls outside any function
self.third_party: Set[str] = set() # modules clearly not local
self.assets: List[AssetRef] = [] # file-path string literals
self.urls: List[Tuple[int, str, Optional[str]]] = [] # (line, url, func)
self.env_vars: List[Tuple[int, str, Optional[str]]] = [] # (line, var, func)
self.subproc_cmds: List[Tuple[int, str, Optional[str]]] = [] # (line, cmd, func)
# ─── Helpers ──────────────────────────────────────────────────────────────────
def collect_py_files(root: Path) -> List[Path]:
files = []
for p in root.rglob("*.py"):
if any(skip in p.parts for skip in SKIP_DIRS):
continue
files.append(p)
return sorted(files)
def rel(path: Path) -> str:
try:
return str(path.relative_to(ROOT)).replace("\\", "/")
except ValueError:
return str(path).replace("\\", "/")
def module_to_file(module: str, from_file: Path) -> Optional[str]:
"""Try to resolve a module name to a file in the project."""
parts = module.replace(".", "/")
# Relative to the file's own directory
candidates = [
from_file.parent / (parts + ".py"),
from_file.parent / parts / "__init__.py",
ROOT / (parts + ".py"),
ROOT / parts / "__init__.py",
]
for c in candidates:
if c.exists():
return rel(c)
return None
def get_call_name(node) -> Optional[str]:
"""Extract a simple string name from a Call node."""
if isinstance(node.func, ast.Name):
return node.func.id
if isinstance(node.func, ast.Attribute):
# e.g. self.foo() -> foo, obj.method() -> method
return node.func.attr
return None
# ─── AST Parser ───────────────────────────────────────────────────────────────
def parse_file(path: Path) -> Optional[FileInfo]:
info = FileInfo(rel(path))
try:
src = path.read_text(encoding="utf-8", errors="ignore")
tree = ast.parse(src, filename=str(path))
except SyntaxError as e:
print(f" [SKIP] Syntax error in {rel(path)}: {e}")
return None
# ── Collect imports ──────────────────────────────────────────
for node in ast.walk(tree):
if isinstance(node, ast.Import):
for alias in node.names:
name = alias.asname or alias.name
info.imports[name] = alias.name
elif isinstance(node, ast.ImportFrom):
module = node.module or ""
for alias in node.names:
imported_name = alias.asname or alias.name
info.from_imports[imported_name] = (module, alias.name)
# ── Collect functions and their calls ────────────────────────
current_class = [None]
class Visitor(ast.NodeVisitor):
def __init__(self):
self._func_stack: List[FunctionInfo] = []
def visit_ClassDef(self, node):
prev = current_class[0]
current_class[0] = node.name
info.classes.append(node.name)
self.generic_visit(node)
current_class[0] = prev
def visit_FunctionDef(self, node):
self._visit_func(node)
def visit_AsyncFunctionDef(self, node):
self._visit_func(node)
def _visit_func(self, node):
fi = FunctionInfo(
name=node.name,
file=info.path,
lineno=node.lineno,
is_method=(current_class[0] is not None),
class_name=current_class[0],
)
# Docstring
if (node.body and isinstance(node.body[0], ast.Expr)
and isinstance(node.body[0].value, ast.Constant)
and isinstance(node.body[0].value.value, str)):
fi.docstring = node.body[0].value.value
# Calls inside this function's body
for child in ast.walk(node):
if isinstance(child, ast.Call):
cname = get_call_name(child)
if cname and cname not in fi.calls:
fi.calls.append(cname)
self._func_stack.append(fi)
info.functions.append(fi)
self.generic_visit(node)
self._func_stack.pop()
Visitor().visit(tree)
# Top-level calls (outside functions/classes)
for node in tree.body:
if isinstance(node, ast.Expr) and isinstance(node.value, ast.Call):
cname = get_call_name(node.value)
if cname:
info.top_level_calls.append(cname)
# ── Asset / URL / ENV / subprocess scan ─────────────────────
_extract_assets(tree, info)
return info
def _current_func_at_line(info: FileInfo, lineno: int) -> Optional[str]:
"""Return the innermost function name that contains `lineno`."""
# Functions are ordered by source; find last one whose lineno <= target
best = None
for fn in info.functions:
if fn.lineno <= lineno:
best = fn.full_name()
return best
_URL_RE = re.compile(r'https?://[^\s\'"\\>]+')
_ENV_GETENV = {"getenv", "environ"}
_EXEC_NAMES = {"python", "python3", "python.exe", "ffmpeg", "ffmpeg.exe",
"ffprobe", "git", "pip", "conda", "bash", "sh", "cmd", "powershell"}
def _extract_assets(tree: ast.AST, info: FileInfo) -> None:
"""Walk the AST and pull out asset paths, URLs, env vars, subprocess calls."""
# ── Pass 1: capture cmd-variable assignments (cmd = ["ffmpeg", ...]) ──
# Maps variable name -> (lineno, first-token string)
_cmd_vars: Dict[str, Tuple[int, str]] = {}
for node in ast.walk(tree):
if not isinstance(node, ast.Assign):
continue
val = node.value
if not (isinstance(val, ast.List) and val.elts):
continue
first_elt = val.elts[0]
first_str: Optional[str] = None
if isinstance(first_elt, ast.Constant) and isinstance(first_elt.value, str):
first_str = first_elt.value
elif isinstance(first_elt, ast.Attribute):
# sys.executable β†’ keep as placeholder
first_str = "<python>"
if first_str is None:
continue
lower = first_str.lower().replace("\\\\", "/")
base = lower.split("/")[-1].split(".")[0]
if base in _EXEC_NAMES or lower == "<python>":
# collect all string parts for a readable label
parts = []
for el in val.elts[:8]:
if isinstance(el, ast.Constant) and isinstance(el.value, str):
parts.append(el.value)
elif isinstance(el, ast.Attribute) and el.attr == "executable":
parts.append("<python>")
else:
parts.append("...")
cmd_preview = " ".join(parts)[:120]
for tgt in node.targets:
if isinstance(tgt, ast.Name):
_cmd_vars[tgt.id] = (node.lineno, cmd_preview)
for node in ast.walk(tree):
lineno = getattr(node, "lineno", 0)
func_ctx = None # resolved lazily below
# ── String literals ─────────────────────────────────────
if isinstance(node, ast.Constant) and isinstance(node.value, str):
s = node.value.strip()
# File-path heuristic: contains a known extension
suffix = Path(s).suffix.lower()
for atype, exts in ASSET_EXTS.items():
if suffix in exts and len(s) > 3:
if func_ctx is None:
func_ctx = _current_func_at_line(info, lineno)
info.assets.append(AssetRef(s, atype, lineno, func_ctx))
break
# URL
if _URL_RE.match(s):
if func_ctx is None:
func_ctx = _current_func_at_line(info, lineno)
info.urls.append((lineno, s, func_ctx))
# ── os.environ / os.getenv ───────────────────────────────
elif isinstance(node, ast.Call):
fname = get_call_name(node)
if fname in {"getenv"}:
if node.args and isinstance(node.args[0], ast.Constant):
if func_ctx is None:
func_ctx = _current_func_at_line(info, lineno)
info.env_vars.append((lineno, node.args[0].value, func_ctx))
# subprocess.run / call / Popen with a plain string command
if fname in {"run", "call", "Popen", "check_output", "check_call", "system"}:
if node.args:
first = node.args[0]
cmd_str = None
if isinstance(first, ast.Constant) and isinstance(first.value, str):
cmd_str = first.value[:120]
elif isinstance(first, ast.List) and first.elts:
parts = []
for el in first.elts[:6]:
if isinstance(el, ast.Constant) and isinstance(el.value, str):
parts.append(el.value)
elif isinstance(el, ast.Attribute) and el.attr == "executable":
parts.append("<python>")
else:
parts.append("...")
if parts:
cmd_str = " ".join(parts)[:120]
elif isinstance(first, ast.Name):
# variable β€” look up in our cmd-variable map
if first.id in _cmd_vars:
cmd_str = _cmd_vars[first.id][1]
if cmd_str:
if func_ctx is None:
func_ctx = _current_func_at_line(info, lineno)
info.subproc_cmds.append((lineno, cmd_str, func_ctx))
# ── os.environ["VAR"] subscript ──────────────────────────
elif isinstance(node, ast.Subscript):
if (isinstance(node.value, ast.Attribute)
and node.value.attr == "environ"
and isinstance(node.slice, ast.Constant)
and isinstance(node.slice.value, str)):
if func_ctx is None:
func_ctx = _current_func_at_line(info, lineno)
info.env_vars.append((lineno, node.slice.value, func_ctx))
# ─── Resolver ─────────────────────────────────────────────────────────────────
STDLIB_ROOTS = {
"os", "sys", "re", "json", "time", "subprocess", "pathlib", "shutil",
"tempfile", "typing", "collections", "functools", "itertools", "logging",
"argparse", "hashlib", "math", "random", "copy", "io", "abc", "enum",
"threading", "asyncio", "dataclasses", "datetime", "inspect", "gc",
"traceback", "warnings", "platform", "struct", "socket", "http",
"urllib", "uuid", "base64", "csv", "glob", "fnmatch",
}
THIRD_PARTY_ROOTS = {
"torch", "cv2", "numpy", "np", "PIL", "flask", "fastapi", "requests",
"scipy", "librosa", "soundfile", "sf", "tqdm", "yaml", "omegaconf",
"imageio", "skimage", "einops", "transformers", "diffusers", "gradio",
"pydantic", "uvicorn", "starlette", "aiofiles", "f5_tts", "cached_path",
"vocos", "safetensors",
}
def is_local(module: str) -> bool:
root = module.split(".")[0]
return root not in STDLIB_ROOTS and root not in THIRD_PARTY_ROOTS
def resolve_calls(all_files: Dict[str, FileInfo]) -> Dict[str, List[Tuple[str, str]]]:
"""
Returns call_graph[file_rel][func_name] = [(target_file, target_func), ...]
as a flat dict: key = "file::func", value = list of "file::func"
"""
# Build lookup: function_name -> [(file, FunctionInfo)]
name_index: Dict[str, List[Tuple[str, FunctionInfo]]] = defaultdict(list)
for frel, finfo in all_files.items():
for fn in finfo.functions:
name_index[fn.name].append((frel, fn))
call_graph: Dict[str, List[str]] = defaultdict(list)
for frel, finfo in all_files.items():
file_path = ROOT / frel.replace("/", os.sep)
for fn in finfo.functions:
key = f"{frel}::{fn.full_name()}"
for raw_call in fn.calls:
# Check from_imports first
if raw_call in finfo.from_imports:
mod, orig = finfo.from_imports[raw_call]
target_file = module_to_file(mod, file_path)
if target_file and target_file in all_files:
call_graph[key].append(f"{target_file}::{orig}")
continue
elif not is_local(mod):
call_graph[key].append(f"[{mod}]::{orig}")
continue
# Check regular imports (module.func pattern won't hit here, but alias might)
if raw_call in finfo.imports:
mod = finfo.imports[raw_call]
target_file = module_to_file(mod, file_path)
if target_file:
call_graph[key].append(f"{target_file}::<module>")
continue
# Global function name match across project files
if raw_call in name_index:
for tfile, tfn in name_index[raw_call]:
entry = f"{tfile}::{tfn.full_name()}"
if entry not in call_graph[key]:
call_graph[key].append(entry)
continue
# Unknown β€” could be builtin, stdlib, or third-party
root_mod = raw_call.split(".")[0]
if root_mod in THIRD_PARTY_ROOTS:
call_graph[key].append(f"[lib:{root_mod}]::{raw_call}")
return call_graph
# ─── README Generator ─────────────────────────────────────────────────────────
def generate_readme(
all_files: Dict[str, FileInfo],
call_graph: Dict[str, List[str]],
entry_points: List[str],
) -> str:
lines = []
lines += [
"# Codebase Call Graph & Function Map",
"",
"Auto-generated by `analyze_codebase.py`.",
"",
"**Legend**",
"- `β†’` = calls",
"- `[lib:X]` = third-party library",
"- `[stdlib:X]` = Python standard library",
"- Line numbers link to source definitions",
"",
"---",
"",
]
# ── Section 1: File summary ──────────────────────────────────
lines += ["## 1. Files in Project", ""]
lines += ["| File | Classes | Functions | Imports |",
"|------|---------|-----------|---------|"]
for frel in sorted(all_files):
fi = all_files[frel]
classes = ", ".join(fi.classes) if fi.classes else "β€”"
func_count = len(fi.functions)
imp_count = len(fi.imports) + len(fi.from_imports)
lines.append(f"| `{frel}` | {classes} | {func_count} | {imp_count} |")
lines += ["", "---", ""]
# ── Section 2: Entry point trace ────────────────────────────
lines += ["## 2. Entry Points", ""]
for ep in entry_points:
ep_norm = ep.replace("\\", "/")
if ep_norm in all_files:
lines.append(f"- **`{ep_norm}`** βœ“")
else:
lines.append(f"- ~~`{ep_norm}`~~ *(not found)*")
lines += ["", "---", ""]
# ── Section 3: Per-file function breakdown ───────────────────
lines += ["## 3. Function Map (`filename β†’ function β†’ calls`)", ""]
for frel in sorted(all_files):
fi = all_files[frel]
if not fi.functions:
continue
lines += [f"### `{frel}`", ""]
# Imports summary for this file
if fi.from_imports or fi.imports:
lines.append("**Imports:**")
seen_mods: Set[str] = set()
for name, (mod, orig) in fi.from_imports.items():
if mod not in seen_mods:
seen_mods.add(mod)
resolved = module_to_file(mod, ROOT / frel.replace("/", os.sep))
tag = f"β†’ `{resolved}`" if resolved else ("*(stdlib)*" if not is_local(mod) else "*(external)*")
lines.append(f"- `from {mod} import ...` {tag}")
for alias, mod in fi.imports.items():
if mod not in seen_mods:
seen_mods.add(mod)
resolved = module_to_file(mod, ROOT / frel.replace("/", os.sep))
tag = f"β†’ `{resolved}`" if resolved else ("*(stdlib)*" if not is_local(mod) else "*(external)*")
lines.append(f"- `import {mod}` {tag}")
lines.append("")
# Functions
lines.append("**Functions:**")
lines.append("")
lines.append("| # | Function | Line | Calls |")
lines.append("|---|----------|------|-------|")
for i, fn in enumerate(fi.functions, 1):
key = f"{frel}::{fn.full_name()}"
raw_calls = fn.calls
resolved_calls = call_graph.get(key, [])
# Build call cell
if not raw_calls:
call_cell = "*(none)*"
else:
# Show resolved targets when available, fallback to raw name
resolved_map = {}
for rc in resolved_calls:
parts = rc.split("::")
func_part = parts[-1]
resolved_map[func_part] = rc
call_parts = []
for c in raw_calls:
if c in resolved_map:
target = resolved_map[c]
if target.startswith("["):
call_parts.append(f"`{target}`")
else:
tfile, tfunc = target.rsplit("::", 1)
call_parts.append(f"`{tfunc}` *({tfile})*")
else:
call_parts.append(f"`{c}`")
call_cell = ", ".join(call_parts[:8])
if len(call_parts) > 8:
call_cell += f" *+{len(call_parts)-8} more*"
prefix = f"{fn.class_name}." if fn.class_name else ""
lines.append(f"| {i} | `{prefix}{fn.name}` | L{fn.lineno} | {call_cell} |")
lines += ["", ""]
# ── Section 4: Full call graph (deduplicated) ────────────────
lines += ["---", "", "## 4. Full Call Graph (all edges)", ""]
lines += ["```"]
for key in sorted(call_graph):
frel, fname = key.split("::", 1)
short_f = frel.split("/")[-1]
targets = call_graph[key]
if targets:
for t in targets:
if "::" in t:
tfile, tfunc = t.rsplit("::", 1)
tshort = tfile.split("/")[-1] if not tfile.startswith("[") else tfile
lines.append(f"{short_f}::{fname} β†’ {tshort}::{tfunc}")
lines += ["```", ""]
# ── Section 5: Assets, URLs, Env vars, Subprocess ──────────
lines += ["---", "", "## 5. Non-Python Assets Referenced", ""]
lines += ["> All string literals that look like file paths, grouped by type.", ""]
# Collect all assets across files
asset_rows: Dict[str, List[Tuple[str, int, str, bool]]] = defaultdict(list)
for frel, fi in all_files.items():
for a in fi.assets:
asset_rows[a.asset_type].append((frel, a.lineno, a.path_str, a.exists))
for atype in ["model", "audio", "video", "image", "config", "data"]:
rows = asset_rows.get(atype, [])
if not rows:
continue
lines.append(f"### {atype.capitalize()} Files")
lines.append("")
lines.append("| File | Line | Path | On Disk |")
lines.append("|------|------|------|---------|")
seen = set()
for frel, ln, path_str, exists in sorted(rows, key=lambda x: (x[0], x[1])):
key = (frel, path_str)
if key in seen:
continue
seen.add(key)
short_f = frel.split("/")[-1]
disk = "βœ“" if exists else "βœ— missing"
safe = path_str.replace("|", "&#124;")[:80]
lines.append(f"| `{short_f}` | L{ln} | `{safe}` | {disk} |")
lines.append("")
# URLs
lines += ["### URLs Referenced", ""]
lines += ["| File | Line | URL | In Function |"]
lines += ["|------|------|-----|-------------|"]
url_seen = set()
for frel, fi in sorted(all_files.items()):
for ln, url, func in fi.urls:
key = (frel, url)
if key in url_seen:
continue
url_seen.add(key)
short_f = frel.split("/")[-1]
func_str = f"`{func}`" if func else "*(module level)*"
lines.append(f"| `{short_f}` | L{ln} | `{url[:80]}` | {func_str} |")
if not url_seen:
lines.append("| β€” | β€” | *(none found)* | β€” |")
lines.append("")
# Env vars
lines += ["### Environment Variables Read", ""]
lines += ["| File | Line | Variable | In Function |"]
lines += ["|------|------|----------|-------------|"]
env_seen = set()
for frel, fi in sorted(all_files.items()):
for ln, var, func in fi.env_vars:
key = (frel, var)
if key in env_seen:
continue
env_seen.add(key)
short_f = frel.split("/")[-1]
func_str = f"`{func}`" if func else "*(module level)*"
lines.append(f"| `{short_f}` | L{ln} | `{var}` | {func_str} |")
if not env_seen:
lines.append("| β€” | β€” | *(none found)* | β€” |")
lines.append("")
# Subprocess commands
lines += ["### Subprocess Commands Launched", ""]
lines += ["| File | Line | Command (truncated) | In Function |"]
lines += ["|------|------|---------------------|-------------|"]
cmd_seen = set()
for frel, fi in sorted(all_files.items()):
for ln, cmd, func in fi.subproc_cmds:
key = (frel, cmd[:60])
if key in cmd_seen:
continue
cmd_seen.add(key)
short_f = frel.split("/")[-1]
func_str = f"`{func}`" if func else "*(module level)*"
safe_cmd = cmd.replace("|", "&#124;")[:80]
lines.append(f"| `{short_f}` | L{ln} | `{safe_cmd}` | {func_str} |")
if not cmd_seen:
lines.append("| β€” | β€” | *(none found)* | β€” |")
lines += ["", "---", ""]
# ── Section 6: Third-party libraries used ────────────────────
lines += ["## 6. Third-Party & External Libraries Referenced", ""]
lib_usages: Dict[str, Set[str]] = defaultdict(set)
for key, targets in call_graph.items():
frel = key.split("::")[0]
for t in targets:
if t.startswith("[lib:"):
lib_name = t.split("[lib:")[1].split("]")[0]
lib_usages[lib_name].add(frel)
# Also pull from imports directly
for frel, fi in all_files.items():
for name, (mod, orig) in fi.from_imports.items():
root = mod.split(".")[0]
if root in THIRD_PARTY_ROOTS:
lib_usages[root].add(frel)
for alias, mod in fi.imports.items():
root = mod.split(".")[0]
if root in THIRD_PARTY_ROOTS:
lib_usages[root].add(frel)
lines.append("| Library | Used in |")
lines.append("|---------|---------|")
for lib in sorted(lib_usages):
files_str = ", ".join(f"`{f.split('/')[-1]}`" for f in sorted(lib_usages[lib]))
lines.append(f"| `{lib}` | {files_str} |")
lines += ["", "---", "", "*Generated by `analyze_codebase.py` β€” re-run any time to refresh.*", ""]
return "\n".join(lines)
# ─── Unused File Detection ────────────────────────────────────────────────────
# Dirs to skip when walking disk for non-Python assets
ASSET_SKIP_DIRS = {"venv", "__pycache__", ".git", "node_modules", "outputs", "temp",
"temp_enhanced_pipeline", "datasets-1", "gfpgan"}
# These Python file patterns are entry points by convention β€” never "unused"
ALWAYS_KEEP_PATTERNS = {
"analyze_codebase.py", # this script
"__init__.py",
"__main__.py",
"setup.py",
"conftest.py",
}
def build_import_graph(all_files: Dict[str, FileInfo]) -> Dict[str, Set[str]]:
"""
Returns imported_by[file] = set of files that import it.
Also returns the full set of reachable files when starting from entry points.
"""
# imported_by: file -> who imports it
imported_by: Dict[str, Set[str]] = defaultdict(set)
for frel, fi in all_files.items():
file_path = ROOT / frel.replace("/", os.sep)
for alias, mod in fi.imports.items():
target = module_to_file(mod, file_path)
if target and target in all_files:
imported_by[target].add(frel)
for name, (mod, orig) in fi.from_imports.items():
target = module_to_file(mod, file_path)
if target and target in all_files:
imported_by[target].add(frel)
return imported_by
def reachable_from_entries(
entry_points: List[str],
imported_by: Dict[str, Set[str]],
all_files: Dict[str, FileInfo],
) -> Set[str]:
"""
BFS from entry points following imports to find all reachable files.
We need forward edges (what does a file import), so invert imported_by.
"""
# Build forward: importer -> set of files it imports
imports_map: Dict[str, Set[str]] = defaultdict(set)
for target, importers in imported_by.items():
for importer in importers:
imports_map[importer].add(target)
visited: Set[str] = set()
queue: List[str] = []
for ep in entry_points:
ep_norm = ep.replace("\\", "/")
if ep_norm in all_files:
queue.append(ep_norm)
while queue:
current = queue.pop()
if current in visited:
continue
visited.add(current)
for dep in imports_map.get(current, set()):
if dep not in visited:
queue.append(dep)
return visited
def categorise_python_file(frel: str, fi: FileInfo) -> str:
"""Return a human label for why a Python file exists."""
name = frel.split("/")[-1]
if name in ALWAYS_KEEP_PATTERNS:
return "framework/convention"
if name.startswith("test_"):
return "test file"
# Has a __main__ guard
src_path = ROOT / frel.replace("/", os.sep)
try:
src = src_path.read_text(encoding="utf-8", errors="ignore")
if '__name__ == "__main__"' in src or "__name__ == '__main__'" in src:
return "standalone script (has __main__)"
except Exception:
pass
if not fi.functions and not fi.classes:
return "config/constants only"
return "unreachable module"
def collect_disk_assets(root: Path) -> Dict[str, Path]:
"""
Walk disk and return all non-Python files with known asset extensions.
Key = relative path string (forward slashes), Value = absolute Path.
"""
all_exts: Set[str] = set()
for exts in ASSET_EXTS.values():
all_exts.update(exts)
found: Dict[str, Path] = {}
for p in root.rglob("*"):
if not p.is_file():
continue
if any(skip in p.parts for skip in ASSET_SKIP_DIRS):
continue
if p.suffix.lower() in all_exts:
found[rel(p)] = p
return found
def normalise_ref(raw: str) -> str:
"""Strip drive letters, normalise slashes, lowercase for fuzzy matching."""
# e.g. C:\\Users\\...\\foo.wav -> foo.wav (just the basename for fuzzy match)
p = Path(raw.strip())
return str(p).replace("\\", "/").lower()
def find_unused(
all_files: Dict[str, FileInfo],
entry_points: List[str],
) -> str:
"""Generate the UNUSED_FILES.md content."""
lines = [
"# Unused & Unreferenced Files",
"",
"Auto-generated by `analyze_codebase.py`.",
"",
"> **Unused Python** = never imported by any other file in the project and not a declared entry point. ",
"> **Unreferenced Asset** = file exists on disk but no string literal in any `.py` file points to it. ",
"> Items marked ⚠️ *might still be used* at runtime via dynamic paths β€” verify before deleting.",
"",
"---",
"",
]
# ── 1. Python files ──────────────────────────────────────────────────────
imported_by = build_import_graph(all_files)
reachable = reachable_from_entries(entry_points, imported_by, all_files)
# Normalise entry point list for quick lookup
ep_set = {ep.replace("\\", "/") for ep in entry_points}
unused_py: List[Tuple[str, str, Set[str]]] = [] # (frel, reason, imported_by_set)
for frel, fi in sorted(all_files.items()):
name = frel.split("/")[-1]
if frel in ep_set:
continue
if name in ALWAYS_KEEP_PATTERNS:
continue
if frel in reachable:
continue
# Not reachable from any entry point
reason = categorise_python_file(frel, fi)
imported_by_set = imported_by.get(frel, set())
unused_py.append((frel, reason, imported_by_set))
lines += [
f"## 1. Python Files Never Imported ({len(unused_py)} files)",
"",
"These files are not reachable from any entry point via import chains.",
"",
"| File | Reason | Imported by |",
"|------|--------|-------------|",
]
for frel, reason, iby in unused_py:
iby_str = ", ".join(f"`{f.split('/')[-1]}`" for f in sorted(iby)) if iby else "*(nothing)*"
lines.append(f"| `{frel}` | {reason} | {iby_str} |")
lines += ["", "---", ""]
# ── 2. Non-Python assets on disk vs referenced ──────────────────────────
disk_assets = collect_disk_assets(ROOT)
# Build a set of all referenced path strings (raw + basename variants)
ref_strings: Set[str] = set()
for fi in all_files.values():
for a in fi.assets:
s = a.path_str.replace("\\", "/").lower()
ref_strings.add(s)
# Also add just the basename so partial paths match
ref_strings.add(Path(s).name)
unreferenced: Dict[str, List[Tuple[str, str]]] = defaultdict(list) # type -> [(frel, disk_path)]
for disk_rel, disk_abs in sorted(disk_assets.items()):
# Try full relative path match first
norm_rel = disk_rel.lower()
basename = Path(disk_rel).name.lower()
if norm_rel in ref_strings or basename in ref_strings:
continue
# Try matching any suffix of the disk path against references
parts = disk_rel.replace("\\", "/").lower().split("/")
matched = any(
"/".join(parts[i:]) in ref_strings
for i in range(len(parts))
)
if matched:
continue
suffix = disk_abs.suffix.lower()
atype = "other"
for t, exts in ASSET_EXTS.items():
if suffix in exts:
atype = t
break
unreferenced[atype].append((disk_rel, str(disk_abs)))
total_unref = sum(len(v) for v in unreferenced.values())
lines += [
f"## 2. Non-Python Assets on Disk But Never Referenced ({total_unref} files)",
"",
"These files exist in the project folder but no Python source file contains a string that matches their path.",
"",
]
for atype in ["model", "audio", "video", "image", "config", "data", "other"]:
items = unreferenced.get(atype, [])
if not items:
continue
lines += [
f"### {atype.capitalize()} Files ({len(items)})",
"",
"| Path | Size |",
"|------|------|",
]
for disk_rel, disk_abs in items:
try:
size = Path(disk_abs).stat().st_size
size_str = f"{size:,} B" if size < 1024 else (
f"{size//1024:,} KB" if size < 1_048_576 else f"{size//1_048_576:,} MB"
)
except Exception:
size_str = "?"
lines.append(f"| `{disk_rel}` | {size_str} |")
lines.append("")
lines += ["---", ""]
# ── 3. Referenced assets that are MISSING from disk ─────────────────────
missing_assets: List[Tuple[str, str, int, str]] = [] # (frel, path_str, line, atype)
for frel, fi in sorted(all_files.items()):
for a in fi.assets:
if not a.exists:
# Skip obvious format strings and very short tokens
if "%" in a.path_str or "{" in a.path_str or len(a.path_str) < 5:
continue
# Skip strings that are clearly descriptions, not paths
if " " in a.path_str and not any(c in a.path_str for c in "\\/"):
continue
missing_assets.append((frel, a.path_str, a.lineno, a.asset_type))
lines += [
f"## 3. Assets Referenced in Code But Missing on Disk ({len(missing_assets)} refs)",
"",
"These are paths the code expects to exist but were not found. May indicate missing model downloads.",
"",
"| Source File | Line | Expected Path | Type |",
"|-------------|------|---------------|------|",
]
seen_missing: Set[Tuple[str, str]] = set()
for frel, path_str, lineno, atype in missing_assets:
key = (frel, path_str)
if key in seen_missing:
continue
seen_missing.add(key)
short_f = frel.split("/")[-1]
safe = path_str.replace("|", "&#124;")[:80]
lines.append(f"| `{short_f}` | L{lineno} | `{safe}` | {atype} |")
lines += [
"",
"---",
"",
"*Generated by `analyze_codebase.py` β€” re-run any time to refresh.*",
"",
]
return "\n".join(lines)
# ─── Main ─────────────────────────────────────────────────────────────────────
def main():
print(f"Scanning: {ROOT}")
py_files = collect_py_files(ROOT)
print(f"Found {len(py_files)} Python files\n")
all_files: Dict[str, FileInfo] = {}
for p in py_files:
print(f" Parsing {rel(p)}")
fi = parse_file(p)
if fi:
all_files[fi.path] = fi
print(f"\nBuilding call graph...")
call_graph = resolve_calls(all_files)
total_edges = sum(len(v) for v in call_graph.values())
print(f" {len(call_graph)} functions with outgoing calls, {total_edges} total edges")
print(f"\nGenerating CODEBASE_MAP.md...")
readme = generate_readme(all_files, call_graph, ENTRY_POINTS)
out_path = ROOT / "CODEBASE_MAP.md"
out_path.write_text(readme, encoding="utf-8")
total_assets = sum(len(fi.assets) for fi in all_files.values())
total_urls = sum(len(fi.urls) for fi in all_files.values())
total_envs = sum(len(fi.env_vars) for fi in all_files.values())
total_cmds = sum(len(fi.subproc_cmds) for fi in all_files.values())
print(f" Done -> {out_path}")
print(f" Files analyzed : {len(all_files)}")
print(f" Total functions : {sum(len(fi.functions) for fi in all_files.values())}")
print(f" Call graph edges: {total_edges}")
print(f" Asset refs : {total_assets}")
print(f" URLs : {total_urls} | Env vars: {total_envs} | Subprocess: {total_cmds}")
print(f"\nGenerating UNUSED_FILES.md...")
unused_report = find_unused(all_files, ENTRY_POINTS)
unused_path = ROOT / "UNUSED_FILES.md"
unused_path.write_text(unused_report, encoding="utf-8")
print(f" Done -> {unused_path}")
# Quick summary stats for unused report
imported_by = build_import_graph(all_files)
reachable = reachable_from_entries(ENTRY_POINTS, imported_by, all_files)
ep_set = {ep.replace("\\", "/") for ep in ENTRY_POINTS}
unused_count = sum(
1 for frel, fi in all_files.items()
if frel not in ep_set
and frel.split("/")[-1] not in ALWAYS_KEEP_PATTERNS
and frel not in reachable
)
disk_asset_count = len(collect_disk_assets(ROOT))
print(f" Unused Python files : {unused_count}")
print(f" Disk assets scanned : {disk_asset_count}")
if __name__ == "__main__":
main()