contextl 1.2.23 → 1.2.25

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "contextl",
-  "version": "1.2.23",
+  "version": "1.2.25",
   "description": "contextl — finds the most relevant files in your codebase for any change request. MCP server for AI coding agents.",
   "keywords": [
     "mcp",
@@ -31,7 +31,7 @@
   "files": [
     "bin/",
     "python/",
-    "README.md"
+    "README.md", "requirements.txt"
   ],
 
   "engines": {

package/python/import_parser.py

@@ -19,6 +19,18 @@ import sys
 from pathlib import Path, PurePosixPath
 from dataclasses import dataclass, field
 
+PYTHON_STDLIB_AND_COMMON_PACKAGES = {
+    "os", "sys", "re", "math", "json", "datetime", "typing", "collections",
+    "itertools", "functools", "pathlib", "logging", "argparse", "subprocess",
+    "unittest", "pytest", "numpy", "pandas", "requests", "urllib", "asyncio",
+    "time", "random", "hashlib", "base64", "csv", "xml", "html", "http",
+    "concurrent", "multiprocessing", "threading", "sqlite3", "flask", "django",
+    "fastapi", "sqlalchemy", "pydantic", "dataclasses", "uuid", "socket",
+    "struct", "tempfile", "shutil", "glob", "copy", "warnings", "traceback",
+    "contextlib", "io", "sysconfig", "importlib", "pkgutil", "runpy",
+    "celery", "redis", "psycopg2", "boto3", "jinja2", "pytest_mock",
+}
+
 from scanner import ScannedFile, ScanResult
 
 
@@ -142,7 +154,7 @@ def _ts_imports_js_ts(root: "Node") -> list[str]:
                     return child.text.decode("utf-8", errors="replace")
             # Fallback: strip quotes manually
             raw = node.text.decode("utf-8", errors="replace")
-            return raw.strip("'\"`")
+            return raw.strip('"\'`')
         return None
 
     def visit(node: "Node"):
@@ -553,24 +565,35 @@ def _find_file_in_repo(
         return None
 
     if extension == ".py":
-        py_ext = candidate_str + ".py"
-        if py_ext in file_index:
-            return py_ext
-        py_idx = candidate_str + "/__init__.py"
-        if py_idx in file_index:
-            return py_idx
-        suffix = "/" + py_ext
-        for path in file_index:
-            if path.endswith(suffix) or path == py_ext:
-                return path
-                
-        # Fallback for Python cross-language imports (e.g. from tests to .ts/.js core logic)
-        for ext in [".ts", ".tsx", ".js", ".jsx", ".java", ".go", ".rs", ".cpp", ".cc", ".cxx", ".h", ".hpp", ".c"]:
-            alt_suffix = "/" + candidate_str + ext
-            alt_matches = [p for p in file_index if p.endswith(alt_suffix) or p == alt_suffix.lstrip("/")]
-            if len(alt_matches) == 1:
-                return alt_matches[0]
-                
+        def _resolve_python(cand: str) -> str | None:
+            py_ext = cand + ".py"
+            if py_ext in file_index:
+                return py_ext
+            py_idx = cand + "/__init__.py"
+            if py_idx in file_index:
+                return py_idx
+            suffix = "/" + py_ext
+            for path in file_index:
+                if path.endswith(suffix) or path == py_ext:
+                    return path
+                    
+            # Fallback for Python cross-language imports (e.g. from tests to .ts/.js core logic)
+            for ext in [".ts", ".tsx", ".js", ".jsx", ".java", ".go", ".rs", ".cpp", ".cc", ".cxx", ".h", ".hpp", ".c"]:
+                alt_suffix = "/" + cand + ext
+                alt_matches = [p for p in file_index if p.endswith(alt_suffix) or p == alt_suffix.lstrip("/")]
+                if len(alt_matches) == 1:
+                    return alt_matches[0]
+            return None
+            
+        res = _resolve_python(candidate_str)
+        if res is not None:
+            return res
+            
+        # Try stripping the last component (in case it's a symbol-level import like `from a.b import C`)
+        if "/" in candidate_str:
+            stripped = candidate_str.rsplit("/", 1)[0]
+            return _resolve_python(stripped)
+            
         return None
     for ext in [".tsx", ".ts", ".jsx", ".js"]:
         with_ext = candidate_str + ext
@@ -872,6 +895,10 @@ def parse_imports(scan_result: ScanResult) -> ParseResult:
             else:
                 if scanned_file.extension == ".rs" and locals().get("is_likely_external", False):
                     continue
+                if scanned_file.extension == ".py":
+                    base_pkg = raw.split(".")[0]
+                    if base_pkg in PYTHON_STDLIB_AND_COMMON_PACKAGES:
+                        continue
                 result.unresolved.append((scanned_file.path, raw))
 
     return result
@@ -1484,11 +1511,18 @@ def extract_skeleton(file_path: str) -> dict:
     try:
         path = Path(file_path)
         ext = path.suffix.lower()
+        
+        lang_map = {
+            ".py": "python", ".ts": "typescript", ".tsx": "tsx",
+            ".js": "javascript", ".jsx": "jsx", ".java": "java",
+            ".rs": "rust", ".go": "go", ".cpp": "cpp",
+        }
+        derived_language = lang_map.get(ext, ext.lstrip("."))
 
         if not _TS_AVAILABLE:
             return {
                 "file": str(path),
-                "language": ext.lstrip("."),
+                "language": derived_language,
                 "error": "tree-sitter not installed — run: pip install tree-sitter tree-sitter-python tree-sitter-typescript ...",
                 "classes": [],
                 "functions": [],
@@ -1500,7 +1534,7 @@ def extract_skeleton(file_path: str) -> dict:
         if parser is None:
             return {
                 "file": str(path),
-                "language": ext.lstrip("."),
+                "language": derived_language,
                 "error": f"Unsupported file type: {ext}",
                 "classes": [],
                 "functions": [],
@@ -1515,7 +1549,7 @@ def extract_skeleton(file_path: str) -> dict:
         if handler is None:
             return {
                 "file": str(path),
-                "language": ext.lstrip("."),
+                "language": derived_language,
                 "error": f"No skeleton handler for {ext}",
                 "classes": [],
                 "functions": [],
@@ -1523,12 +1557,6 @@ def extract_skeleton(file_path: str) -> dict:
                 "docstrings": {},
             }
 
-        lang_map = {
-            ".py": "python", ".ts": "typescript", ".tsx": "tsx",
-            ".js": "javascript", ".jsx": "jsx", ".java": "java",
-            ".rs": "rust", ".go": "go", ".cpp": "cpp",
-        }
-
         if ext == ".py":
             body = handler(tree.root_node, source_bytes)
         else:
@@ -1536,14 +1564,22 @@ def extract_skeleton(file_path: str) -> dict:
 
         return {
             "file": str(path),
-            "language": lang_map.get(ext, ext.lstrip(".")),
+            "language": derived_language,
             **body,
         }
 
     except Exception as e:
+        # Fallback if path is not yet defined
+        lang = "unknown"
+        if 'derived_language' in locals():
+            lang = derived_language
+        elif 'file_path' in locals():
+            ext = Path(file_path).suffix.lower()
+            lang = lang_map.get(ext, ext.lstrip(".")) if 'lang_map' in locals() else ext.lstrip(".")
+
         return {
             "file": str(file_path),
-            "language": "unknown",
+            "language": lang,
             "error": str(e),
             "classes": [],
             "functions": [],

package/python/mcp_server.py

@@ -106,7 +106,10 @@ def _get_graph(repo_path: str):
         now = time.time()
         entry = _cache.get(resolved)
         
-        # Fast path: bypass filesystem walk if checked within last 5 seconds
+        # Fast path: bypass filesystem walk if checked within last 5 seconds.
+        # This intentionally creates a 5s "blind window" where new/deleted/modified
+        # files are not instantly detected, drastically reducing disk I/O on large repos
+        # when the MCP client rapid-fires multiple queries concurrently.
         if entry and (now - entry.last_checked_time < 5.0):
             return entry.scan, entry.repo_graph
 

package/python/query_engine.py

@@ -17,6 +17,7 @@ No LLM. No embeddings. Pure text + graph.
 import math
 import re
 from dataclasses import dataclass, field
+from impact_analysis import _is_test_file
 
 from scanner import scan_repo
 from import_parser import parse_imports
@@ -93,12 +94,16 @@ def _keyword_score(query_terms: list[str], file_path: str, idf_weights: dict[str
     score = 0.0
 
     for term in query_terms:
-        # Exact match
+        # Exact match on filename
         if term in filename_toks:
             score += 2.0 * idf_weights.get(term, 1.0)   # Massive signal
             matched.append(term)
-        # Substring match on filename
-        elif len(term) >= 2 and any(term in ft for ft in filename_toks) or any(len(ft) >= 2 and ft in term for ft in filename_toks):
+        # Substring or abbreviation match on filename
+        elif len(term) >= 2 and any(
+            term in ft or ft in term or ft.startswith(term) or term.startswith(ft) or
+            all(c in iter(ft) for c in term)
+            for ft in filename_toks
+        ):
             score += 1.5 * idf_weights.get(term, 1.0)
             matched.append(term)
         # Exact match on path
@@ -106,8 +111,12 @@ def _keyword_score(query_terms: list[str], file_path: str, idf_weights: dict[str
             score += 0.8 * idf_weights.get(term, 1.0)   # Strong signal
             if term not in matched:
                 matched.append(term)
-        # Substring match on path
-        elif len(term) >= 2 and any(term in pt for pt in path_toks) or any(len(pt) >= 2 and pt in term for pt in path_toks):
+        # Substring or abbreviation match on path
+        elif len(term) >= 2 and any(
+            term in pt or pt in term or pt.startswith(term) or term.startswith(pt) or
+            all(c in iter(pt) for c in term)
+            for pt in path_toks
+        ):
             score += 0.4 * idf_weights.get(term, 1.0)
             if term not in matched:
                 matched.append(term)
@@ -247,7 +256,7 @@ def _content_score(query_terms: list[str], file_path: str, file_content: str, id
 def _apply_neighbor_bonus(
     scores: dict[str, float],
     repo_graph: RepoGraph,
-    boost: float = 0.35,
+    boost: float = 0.2,
     depth: int = 1,
 ) -> dict[str, float]:
     """
@@ -353,8 +362,9 @@ def query(
         content_scores[path] = cscore
         matched_terms[path] = list(set(kterms + cterms))
         
-        # Base combination: 50/50 keyword vs content
-        combined = (kscore * 0.5) + (cscore * 0.5)
+        # Base combination: heavily weight filename/path matches (70%) over content matches (30%)
+        # This prevents transitive consumers of a file from out-ranking the file itself
+        combined = (kscore * 0.7) + (cscore * 0.3)
         
         # Tiebreaker: Slightly penalize deep paths so src/x.py wins over build/npm/python/x.py
         depth_penalty = 0.0001 * path.count("/")
@@ -383,7 +393,6 @@ def query(
 
     # Down-rank test files unless the user specifically searched for tests
     is_test_query = "test" in q.lower() or "spec" in q.lower()
-    from impact_analysis import _is_test_file
     for r in results:
         if _is_test_file(r.path) and not is_test_query:
             r.total_score *= 0.5

package/requirements.txt

@@ -0,0 +1,10 @@
+networkx
+mcp>=1.0.0
+tree-sitter>=0.22.0
+tree-sitter-python>=0.22.0
+tree-sitter-javascript>=0.22.0
+tree-sitter-typescript>=0.22.0
+tree-sitter-java>=0.22.0
+tree-sitter-rust>=0.22.0
+tree-sitter-go>=0.22.0
+tree-sitter-cpp>=0.22.0