contextl 1.2.9 → 1.2.11

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "contextl",
-  "version": "1.2.9",
+  "version": "1.2.11",
   "description": "contextl — finds the most relevant files in your codebase for any change request. MCP server for AI coding agents.",
   "keywords": [
     "mcp",

package/python/import_parser.py

@@ -22,7 +22,7 @@ from scanner import ScannedFile, ScanResult
 
 # Matches any ES6 import statement and captures the module path
 IMPORT_PATTERN = re.compile(
-    r"""import\s+(?:type\s+)?          # import or import type
+    r"""(?:import|export)\s+(?:type\s+)?          # import or import type or export
         (?:
             \{[^}]*\}                  # named imports: { Foo, Bar }
             |[\w*]+                    # default or namespace: Foo or *
@@ -38,11 +38,12 @@ IMPORT_PATTERN = re.compile(
 DYNAMIC_IMPORT_PATTERN = re.compile(r"""import\s*\(\s*['"](.*?)['"]\s*\)""")
 
 # Python
-PYTHON_FROM_IMPORT_PATTERN = re.compile(r"^from\s+([.\w]+)\s+import", re.MULTILINE)
-PYTHON_IMPORT_PATTERN = re.compile(r"^import\s+([.\w]+)", re.MULTILINE)
+PYTHON_FROM_IMPORT_PATTERN = re.compile(r"^[ \t]*from\s+([.\w]+)\s+import\s+(?:\(([^)]+)\)|([^#\n]+))", re.MULTILINE)
+PYTHON_IMPORT_PATTERN = re.compile(r"^[ \t]*import\s+([.\w]+)", re.MULTILINE)
 
 # Java
-JAVA_IMPORT_PATTERN = re.compile(r"^import\s+(?:static\s+)?([\w.]+);", re.MULTILINE)
+JAVA_IMPORT_PATTERN = re.compile(r"^import\s+(?:static\s+)?([\w.*]+);", re.MULTILINE)
+JAVA_PACKAGE_PATTERN = re.compile(r"^package\s+([\w.]+);", re.MULTILINE)
 
 
 @dataclass
@@ -83,7 +84,15 @@ def _extract_raw_imports(source_code: str, extension: str) -> list[str]:
             paths.append(match.group(1))
     elif extension == ".py":
         for match in PYTHON_FROM_IMPORT_PATTERN.finditer(source_code):
-            paths.append(match.group(1))
+            base = match.group(1)
+            paths.append(base)
+            symbols_str = match.group(2) or match.group(3)
+            if symbols_str:
+                symbols_str = symbols_str.replace('(', '').replace(')', '').replace('\\', '').strip()
+                for sym in symbols_str.split(','):
+                    sym = sym.split(' as ')[0].strip()
+                    if sym and sym != '*':
+                        paths.append(f"{base}.{sym}")
         for match in PYTHON_IMPORT_PATTERN.finditer(source_code):
             paths.append(match.group(1))
     elif extension == ".java":
@@ -134,12 +143,26 @@ def _resolve_alias(import_path: str, alias_map: dict[str, str]) -> str | None:
 def _resolve_relative(import_path: str, source_file: str) -> str:
     """
     Resolve a relative import path against the source file's directory.
-    Example: source="frontend/app/page.tsx", import="../lib/api" → "frontend/lib/api"
+    Returns a repo-relative path string (NOT absolute).
+
+    Example: source="apps/web/src/routes.tsx", import="./App.tsx"
+             → "apps/web/src/App.tsx"
     """
-    source_dir = Path(source_file).parent
-    resolved = (source_dir / import_path).resolve()
-    # Make it relative again (we'll strip the absolute prefix below)
-    return str(resolved)
+    # Use pure PurePosixPath arithmetic — never call .resolve() which
+    # anchors to the OS cwd and produces a wrong absolute path.
+    from pathlib import PurePosixPath
+    source_dir = PurePosixPath(source_file).parent
+    # Normalise away any .. and . components without touching the filesystem
+    resolved = (source_dir / import_path)
+    # PurePosixPath doesn't have .resolve(), so manually normalise ".."
+    parts = []
+    for part in resolved.parts:
+        if part == "..":
+            if parts:
+                parts.pop()
+        elif part != ".":
+            parts.append(part)
+    return str(PurePosixPath(*parts)) if parts else "."
 
 
 def _find_file_in_repo(
@@ -177,6 +200,16 @@ def _find_file_in_repo(
         for path in file_index:
             if path.endswith(suffix):
                 return path
+        
+        # Fallback for static method/field imports: import static com.example.MyClass.myMethod;
+        # In this case, candidate_str is 'com/example/MyClass/myMethod'
+        if "/" in candidate_str:
+            parent_candidate = candidate_str.rsplit("/", 1)[0]
+            parent_suffix = parent_candidate + ".java"
+            for path in file_index:
+                if path.endswith(parent_suffix):
+                    return path
+                    
         return None
 
     if extension == ".py":
@@ -201,50 +234,137 @@ def _find_file_in_repo(
         if with_ext in file_index:
             return with_ext
 
-    # Node.js: Try as a directory index file
+    # Node.js / Deno: Try as a directory index/mod file
     for ext in [".tsx", ".ts", ".jsx", ".js"]:
         index_path = candidate_str + "/index" + ext
         if index_path in file_index:
             return index_path
+        mod_path = candidate_str + "/mod" + ext
+        if mod_path in file_index:
+            return mod_path
+
+    # Fuzzy fallback for dynamic paths (like P4A recipes)
+    if extension == ".py":
+        candidate_suffix = "/" + candidate_str + ".py"
+        matches = [p for p in file_index if p.endswith(candidate_suffix)]
+        if len(matches) == 1:
+            return matches[0]
 
     return None
 
 
+def _load_tsconfig_paths(tsconfig_path: Path) -> dict:
+    """
+    Load compilerOptions.paths from a tsconfig.json, following `extends` chains.
+    Returns merged paths dict (child overrides parent).
+    """
+    import json
+    visited = set()
+    merged_paths = {}
+
+    def _load(path: Path):
+        if path in visited:
+            return
+        visited.add(path)
+        try:
+            text = path.read_text(encoding="utf-8")
+            pattern = r'(".*?(?<!\\)")|(/\*.*?\*/|//[^\r\n]*)'
+            text = re.sub(pattern, lambda m: m.group(1) if m.group(1) else '', text, flags=re.S)
+            data = json.loads(text)
+        except Exception:
+            return
+
+        # Follow extends first (parent paths are base, child overrides)
+        extends = data.get("extends")
+        if extends:
+            parent = (path.parent / extends).resolve()
+            # handle both with and without .json
+            if not parent.suffix:
+                parent = parent.with_suffix(".json")
+            _load(parent)
+
+        # Then apply this tsconfig's own paths (child wins)
+        paths = data.get("compilerOptions", {}).get("paths", {})
+        for alias, targets in paths.items():
+            if targets:
+                merged_paths[alias] = targets[0]
+
+    _load(tsconfig_path)
+    return merged_paths
+
+
 def _detect_alias_map(scan_result: ScanResult) -> dict[str, str]:
     """
-    Auto-detect the @/ alias by finding tsconfig.json or next.config files
-    and inferring the project root. Falls back to a heuristic.
-    
-    Returns a map like {"@/": "frontend/"} or {"@/": ""}
+    Auto-detect TypeScript path aliases by reading every tsconfig.json in the repo.
+    Each tsconfig's aliases are scoped relative to its own directory so that
+    multiple packages with conflicting @/* aliases don't clobber each other.
+
+    Returns a flat map:  alias_prefix -> resolved_repo_relative_prefix
+    e.g. {
+        '@/'            : 'packages/ui/src/',
+        '@components/'  : 'apps/web/src/components/',
+        '@core/'        : 'apps/web/src/core/',
+    }
     """
     root = Path(scan_result.root)
-    alias_map = {}
+    alias_map: dict[str, str] = {}
 
-    # Look for tsconfig.json to find paths config
-    for tsconfig in root.rglob("tsconfig.json"):
+    # Walk all tsconfig.json files (skip node_modules / build dirs)
+    ignore = {"node_modules", "dist", "build", ".next", "coverage", "out"}
+    tsconfigs = []
+    for tc in root.rglob("tsconfig.json"):
+        if any(p in ignore for p in tc.parts):
+            continue
+        tsconfigs.append(tc)
+
+    for tsconfig in tsconfigs:
+        tsconfig_dir = tsconfig.parent
+        paths = _load_tsconfig_paths(tsconfig)
+
+        for alias, target in paths.items():
+            # Normalise: strip trailing /* or *
+            clean_alias = alias.rstrip("/*").rstrip("*")
+            clean_target = target.rstrip("/*").rstrip("*").lstrip("./")
+
+            # Resolve target relative to the tsconfig's own directory
+            resolved = (tsconfig_dir / clean_target).resolve()
+            try:
+                repo_rel = str(resolved.relative_to(root))
+            except ValueError:
+                continue  # outside repo — skip
+
+            alias_key = clean_alias + "/"
+            target_val = repo_rel.replace("\\", "/") + "/"
+
+            # Only add if not already defined (first match / most specific wins)
+            if alias_key not in alias_map:
+                alias_map[alias_key] = target_val
+
+    # Detect NPM Workspaces (monorepos)
+    for pkg in root.rglob("package.json"):
+        if any(p in ignore for p in pkg.parts):
+            continue
         try:
             import json
-            # naive regex to strip comments from tsconfig before parsing
-            text = tsconfig.read_text()
-            text = re.sub(r'//.*?\n|/\*.*?\*/', '', text, flags=re.S)
-            data = json.loads(text)
-            paths = data.get("compilerOptions", {}).get("paths", {})
-
-            for alias, targets in paths.items():
-                if not targets:
-                    continue
-                # "@/*": ["./src/*"] → strip trailing /* from both
-                clean_alias = alias.rstrip("/*").rstrip("*")
-                clean_target = targets[0].rstrip("/*").rstrip("*").lstrip("./")
-
-                tsconfig_dir = tsconfig.parent.relative_to(root)
-                prefix = str(tsconfig_dir / clean_target).lstrip("./")
-                alias_map[clean_alias + "/"] = prefix + "/" if prefix else ""
+            data = json.loads(pkg.read_text(encoding="utf-8"))
+            name = data.get("name")
+            if name and isinstance(name, str):
+                pkg_dir = pkg.parent
+                try:
+                    repo_rel = str(pkg_dir.relative_to(root)).replace("\\", "/")
+                    if repo_rel == ".":
+                        continue # Skip root package.json
+                    # For absolute imports like "@meshtastic/sdk-react"
+                    alias_map[name] = repo_rel
+                    # For sub-path imports like "@meshtastic/sdk-react/src/foo"
+                    alias_map[name + "/"] = repo_rel + "/"
+                except ValueError:
+                    pass
         except Exception:
-            continue
+            pass
 
     if not alias_map:
-        # Heuristic fallback: if all files share a common top-level dir, use that
+        # Heuristic fallback: single top-level source dir
         top_dirs = {Path(f.path).parts[0] for f in scan_result.files if Path(f.path).parts}
         if len(top_dirs) == 1:
             top = list(top_dirs)[0]
@@ -274,6 +394,27 @@ def parse_imports(scan_result: ScanResult) -> ParseResult:
     # Auto-detect alias map (e.g. @/ → frontend/)
     alias_map = _detect_alias_map(scan_result)
 
+    # Pre-pass for Java: build package map and implicit same-package relationships
+    java_packages = {}
+    for scanned_file in scan_result.files:
+        if scanned_file.extension == ".java":
+            try:
+                source_code = Path(scanned_file.absolute_path).read_text(encoding="utf-8")
+                pkg_match = JAVA_PACKAGE_PATTERN.search(source_code)
+                if pkg_match:
+                    pkg = pkg_match.group(1)
+                    java_packages.setdefault(pkg, []).append(scanned_file.path)
+            except Exception:
+                pass
+
+    for pkg, files in java_packages.items():
+        for source_file in files:
+            for target_file in files:
+                if source_file != target_file:
+                    result.relationships.append(
+                        ImportRelationship(source=source_file, target=target_file, raw_import=f"implicit_package:{pkg}")
+                    )
+
     for scanned_file in scan_result.files:
         try:
             source_code = Path(scanned_file.absolute_path).read_text(encoding="utf-8")
@@ -289,6 +430,14 @@ def parse_imports(scan_result: ScanResult) -> ParseResult:
 
             # Resolve to a candidate path string
             if scanned_file.extension == ".java":
+                if raw.endswith(".*"):
+                    pkg = raw[:-2]
+                    for target_file in java_packages.get(pkg, []):
+                        if target_file != scanned_file.path:
+                            result.relationships.append(
+                                ImportRelationship(source=scanned_file.path, target=target_file, raw_import=raw)
+                            )
+                    continue
                 candidate = raw.replace(".", "/")
             elif scanned_file.extension == ".py":
                 if raw.startswith("."):

package/python/obsidian_export.py

@@ -19,68 +19,95 @@ from graph_builder import build_graph, RepoGraph
 
 PYTHON_DOCSTRING_PATTERN = re.compile(r'^\s*["\']{3}(.*?)["\']{3}', re.DOTALL)
 JS_DOC_PATTERN = re.compile(r'^\s*/\*\*(.*?)\*/', re.DOTALL)
+OUTLINE_PATTERN = re.compile(r'^\s*(?:export\s+)?(?:class|def|function|interface|type)\s+([a-zA-Z0-9_]+)', re.MULTILINE)
 
-def _extract_explanation(file_path: str, extension: str, root_dir: str) -> str:
-    """Reads the top-level docstring or JSDoc block from the file."""
+def _extract_file_context(file_path: str, extension: str, root_dir: str) -> tuple[str, list[str], str]:
+    """Returns (explanation, outline_symbols, snippet)."""
+    explanation = "*No explanation provided in source code.*"
+    outline = []
+    snippet = ""
     try:
         content = (Path(root_dir) / file_path).read_text(encoding="utf-8")
+        
+        # 1. Snippet (first 20 lines)
+        lines = content.splitlines()
+        snippet = "\n".join(lines[:20])
+        if len(lines) > 20:
+            snippet += "\n..."
+            
+        # 2. Explanation
         if extension == ".py":
             match = PYTHON_DOCSTRING_PATTERN.search(content)
             if match:
-                return match.group(1).strip()
+                explanation = match.group(1).strip()
         elif extension in {".ts", ".tsx", ".js", ".jsx", ".java"}:
             match = JS_DOC_PATTERN.search(content)
             if match:
-                # Clean up leading asterisks from JSDoc lines
-                lines = match.group(1).split('\n')
-                cleaned = [line.strip().lstrip('*').strip() for line in lines]
-                return "\n".join(cleaned).strip()
+                clean_lines = [line.strip().lstrip('*').strip() for line in match.group(1).split('\n')]
+                explanation = "\n".join(clean_lines).strip()
+                
+        # 3. Outline
+        for match in OUTLINE_PATTERN.finditer(content):
+            outline.append(match.group(1))
+            
     except Exception:
         pass
-    return "*No explanation provided in source code.*"
+        
+    return explanation, outline, snippet
 
 
 def export_obsidian_vault(repo_graph: RepoGraph, output_dir: str) -> str:
     """
-    Generates a markdown file for every node in the graph, with
-    Obsidian wikilinks connecting them based on imports.
+    Generates a markdown file for every node in the graph in a hierarchical folder structure, 
+    with Obsidian wikilinks connecting them based on imports.
     """
     out_path = Path(output_dir).resolve()
     
-    # Create the output directory (clean it if it exists)
     if out_path.exists():
         shutil.rmtree(out_path)
     out_path.mkdir(parents=True)
 
     for path, node in repo_graph.nodes.items():
-        # We replace slashes with dashes to keep a flat folder structure
-        # so Obsidian graph view looks clean, but keep original name as title
-        safe_name = path.replace("/", "-").replace("\\", "-")
-        md_file = out_path / f"{safe_name}.md"
+        md_file = out_path / f"{path}.md"
+        md_file.parent.mkdir(parents=True, exist_ok=True)
         
         dependencies = repo_graph.get_dependencies(path)
         dependents = repo_graph.get_dependents(path)
         
-        explanation = _extract_explanation(path, node.extension, repo_graph.root)
+        explanation, outline, snippet = _extract_file_context(path, node.extension, repo_graph.root)
         
         content = [
-            f"# {path}",
+            f"# {Path(path).name}",
+            "",
+            "## Architecture Metrics",
+            f"- **Path:** `{path}`",
+            f"- **Extension:** `{node.extension}`",
+            f"- **Size:** {node.size_bytes} bytes",
+            f"- **Centrality Score:** {node.centrality:.4f}",
+            f"- **In-Degree (Imported By):** {len(dependents)}",
+            f"- **Out-Degree (Imports):** {len(dependencies)}",
             "",
             "## Explanation",
             explanation,
             "",
-            "## Metrics",
-            f"**Extension:** `{node.extension}`",
-            f"**Size:** {node.size_bytes} bytes",
-            f"**Centrality Score:** {node.centrality:.4f}",
-            "",
-            "## Imports (Dependencies)",
+            "## Structural Outline",
         ]
         
+        if outline:
+            for symbol in outline:
+                content.append(f"- `{symbol}`")
+        else:
+            content.append("*No major classes or functions detected.*")
+            
+        content.extend([
+            "",
+            "## Imports (Dependencies)"
+        ])
+        
         if dependencies:
             for dep in sorted(dependencies):
-                dep_safe = dep.replace("/", "-").replace("\\", "-")
-                content.append(f"- [[{dep_safe}]]")
+                # Use full relative path for accurate linking in hierarchical vaults
+                content.append(f"- [[{dep}.md|{dep}]]")
         else:
             content.append("*No internal imports*")
             
@@ -91,11 +118,18 @@ def export_obsidian_vault(repo_graph: RepoGraph, output_dir: str) -> str:
         
         if dependents:
             for dep in sorted(dependents):
-                dep_safe = dep.replace("/", "-").replace("\\", "-")
-                content.append(f"- [[{dep_safe}]]")
+                content.append(f"- [[{dep}.md|{dep}]]")
         else:
             content.append("*Not imported by any file*")
             
+        content.extend([
+            "",
+            "## Source Code Snippet",
+            f"```{node.extension.lstrip('.')}",
+            snippet,
+            "```"
+        ])
+            
         md_file.write_text("\n".join(content), encoding="utf-8")
         
     return str(out_path)

package/python/query_engine.py

@@ -161,6 +161,13 @@ def _content_score(query_terms: list[str], file_path: str, file_content: str, id
                 for part in re.findall(r"[a-z0-9]+", sub):
                     if len(part) > 3:
                         java_methods.add(part)
+                        
+    py_defs = set()
+    if file_path.endswith(".py"):
+        defs = re.findall(r"\b(?:def|async def)\s+([a-zA-Z0-9_]+)\s*\(", file_content)
+        for d in defs:
+            if d.lower() not in {"main", "init", "get", "set"}:
+                py_defs.add(d.lower())
             
     if is_ts:
         exports = re.findall(r"\bexport\s+(?:const|let|var|class|interface|function|default)\s+([a-zA-Z0-9_]+)", file_content)
@@ -213,7 +220,7 @@ def _content_score(query_terms: list[str], file_path: str, file_content: str, id
             multiplier = 8.0
         elif t in declared_classes:
             multiplier = class_mult
-        elif t in java_methods:
+        elif t in java_methods or t in py_defs:
             multiplier = 4.0
         elif t in annotations:
             multiplier = ann_mult
@@ -340,9 +347,10 @@ def query(
         if df == 0:
             idf_weights[term] = 0.0
         else:
-            idf_weights[term] = math.log(total_files / (1 + df)) if total_files > 0 else 1.0
+            idf_weights[term] = max(0.1, math.log(total_files / (1 + df))) if total_files > 0 else 1.0
 
     # --- Pass 1: keyword + content scores per file ---
+    base_scores: dict[str, float] = {}
     keyword_scores: dict[str, float] = {}
     content_scores: dict[str, float] = {}
     matched_terms: dict[str, list[str]] = {}
@@ -355,14 +363,15 @@ def query(
         keyword_scores[path] = kscore
         content_scores[path] = cscore
         matched_terms[path] = list(set(kterms + cterms))
-
-    # --- Pass 2: combine into base score ---
-    base_scores: dict[str, float] = {}
-    for path in repo_graph.nodes:
-        base_scores[path] = (
-            keyword_scores.get(path, 0.0) * 0.5 +
-            content_scores.get(path, 0.0) * 0.5
-        )
+        
+        # Base combination: 50/50 keyword vs content
+        combined = (kscore * 0.5) + (cscore * 0.5)
+        
+        # Tiebreaker: Slightly penalize deep paths so src/x.py wins over build/npm/python/x.py
+        depth_penalty = 0.0001 * path.count("/")
+        combined -= depth_penalty
+        
+        base_scores[path] = combined
 
     # --- Pass 3: neighbor bonus ---
     boosted_scores = _apply_neighbor_bonus(base_scores, repo_graph)

package/python/scanner.py

@@ -7,24 +7,23 @@ Filters by supported extensions for the MVP (Next.js / React / TypeScript).
 """
 
 import os
+import pathlib
 from pathlib import Path
 from dataclasses import dataclass, field
 
 
 SUPPORTED_EXTENSIONS = {".tsx", ".ts", ".jsx", ".js", ".py", ".java"}
 
-IGNORED_DIRS = {
-    "node_modules",
-    ".git",
-    ".next",
-    "dist",
-    "build",
-    ".cache",
-    "__pycache__",
-    ".turbo",
-    "coverage",
+ignore_dirs = {
+    ".git", "node_modules", "__pycache__", "venv", ".venv",
+    "env", ".env", ".next", ".nuxt", "out", "build", "dist",
+    "coverage", ".nyc_output", ".pytest_cache", "npm", "pypi"
 }
 
+def _should_ignore(path: str) -> bool:
+    parts = pathlib.Path(path).parts
+    return any(part in ignore_dirs for part in parts)
+
 
 @dataclass
 class ScannedFile:
@@ -87,7 +86,7 @@ def scan_repo(repo_path: str) -> ScanResult:
 
     for dirpath, dirnames, filenames in os.walk(root):
         # Prune ignored directories in-place so os.walk skips them
-        dirnames[:] = [d for d in dirnames if d not in IGNORED_DIRS]
+        dirnames[:] = [d for d in dirnames if d not in ignore_dirs]
 
         for filename in filenames:
             filepath = Path(dirpath) / filename