codegraphcontext 0.6.9__tar.gz → 0.6.11__tar.gz

{codegraphcontext-0.6.9/src/codegraphcontext.egg-info → codegraphcontext-0.6.11}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: codegraphcontext
-Version: 0.6.9
+Version: 0.6.11
 Summary: An MCP server that indexes local code into a graph database to provide context to AI assistants.
 Author-email: Shashank Shekhar Singh <shashankshekharsingh1205@gmail.com>
 License: MIT License
@@ -388,6 +388,8 @@ CodeGraphContext supports multiple graph database backends to suit your environm
 | **Speed** | ⚡ Extremely Fast | ⚡ Fast | 🚀 Scalable | 🌐 Network-dependent |
 | **Persistence**| Yes (to disk) | Yes (to disk) | Yes (to disk) | Yes (server-side) |
 
+Embedded KuzuDB/LadybugDB default to a **4 GiB** buffer pool (`CGC_EMBEDDED_BUFFER_POOL_MB`). Set that env var to another MiB value, or `0` for the library default (~80% of system RAM).
+
 ---
 
 ## SCIP indexing (optional)
@@ -873,6 +875,14 @@ Once the server is running, you can interact with it through your AI assistant u
 
 ---
 
+## Troubleshooting
+
+Hitting a lock error, a "never finishes indexing" loop, or a silent feature?
+See [docs/TROUBLESHOOTING.md](docs/TROUBLESHOOTING.md) — every entry maps a
+real reported issue to its fix or workaround.
+
+---
+
 ## Contributing
 
 Contributions are welcome! 🎉  

{codegraphcontext-0.6.9 → codegraphcontext-0.6.11}/README.md

@@ -289,6 +289,8 @@ CodeGraphContext supports multiple graph database backends to suit your environm
 | **Speed** | ⚡ Extremely Fast | ⚡ Fast | 🚀 Scalable | 🌐 Network-dependent |
 | **Persistence**| Yes (to disk) | Yes (to disk) | Yes (to disk) | Yes (server-side) |
 
+Embedded KuzuDB/LadybugDB default to a **4 GiB** buffer pool (`CGC_EMBEDDED_BUFFER_POOL_MB`). Set that env var to another MiB value, or `0` for the library default (~80% of system RAM).
+
 ---
 
 ## SCIP indexing (optional)
@@ -774,6 +776,14 @@ Once the server is running, you can interact with it through your AI assistant u
 
 ---
 
+## Troubleshooting
+
+Hitting a lock error, a "never finishes indexing" loop, or a silent feature?
+See [docs/TROUBLESHOOTING.md](docs/TROUBLESHOOTING.md) — every entry maps a
+real reported issue to its fix or workaround.
+
+---
+
 ## Contributing
 
 Contributions are welcome! 🎉  

{codegraphcontext-0.6.9 → codegraphcontext-0.6.11}/pyproject.toml

@@ -1,6 +1,6 @@
 [project]
 name = "codegraphcontext"
-version = "0.6.9"
+version = "0.6.11"
 description = "An MCP server that indexes local code into a graph database to provide context to AI assistants."
 authors = [{ name = "Shashank Shekhar Singh", email = "shashankshekharsingh1205@gmail.com" }]
 readme = "README.md"

{codegraphcontext-0.6.9 → codegraphcontext-0.6.11}/src/codegraphcontext/cli/cli_helpers.py

@@ -342,6 +342,21 @@ def index_helper(path: str, context: Optional[str] = None, no_progress: bool = F
         db_manager.close_driver()
         raise typer.Exit(code=1)
 
+    # Register the repo in its named context BEFORE the already-indexed skip
+    # guard below: the early return used to sit above this call, so a graph
+    # populated by any path that skipped registration (MCP add_code_to_graph,
+    # an interrupted run) could never reconcile config.yaml through the normal
+    # command — 'cgc context list' showed "Repos Linked: 0" forever (#1317).
+    # Registration is idempotent, so running it on every invocation is safe.
+    # The resolved context name is used so a *default* named context registers
+    # too, not only an explicit --context flag.
+    if ctx.mode == "named":
+        register_context_name = context or getattr(ctx, "context_name", None)
+        if register_context_name:
+            if not register_repo_in_context(register_context_name, str(path_obj), auto_create=False):
+                db_manager.close_driver()
+                raise typer.Exit(code=1)
+
     indexed_repos = code_finder.list_indexed_repositories()
     repo_exists = any_repo_matches_path(indexed_repos, path_obj)
 
@@ -400,11 +415,6 @@ def index_helper(path: str, context: Optional[str] = None, no_progress: bool = F
         except Exception as e:
             console.print(f"[yellow]Warning: Could not check file count: {e}. Proceeding with indexing...[/yellow]")
 
-    if context and ctx.mode == "named":
-        if not register_repo_in_context(context, str(path_obj), auto_create=False):
-            db_manager.close_driver()
-            raise typer.Exit(code=1)
-
     console.print(f"Starting indexing for: {path_obj}")
 
     try:

{codegraphcontext-0.6.9 → codegraphcontext-0.6.11}/src/codegraphcontext/cli/main.py

@@ -711,6 +711,7 @@ def bundle_export(
     no_stats: bool = typer.Option(False, "--no-stats", help="Skip statistics generation"),
     sign_key: Optional[str] = typer.Option(None, "--sign-key", envvar="CGC_BUNDLE_SIGN_KEY", help="HMAC signing key (or set CGC_BUNDLE_SIGN_KEY)"),
     encrypt_password: Optional[str] = typer.Option(None, "--encrypt-password", envvar="CGC_BUNDLE_PASSWORD", help="Encrypt bundle with this password (or set CGC_BUNDLE_PASSWORD)", hide_input=True),
+    exclude_labels: Optional[str] = typer.Option(None, "--exclude-labels", help="Comma-separated node labels to leave out of the bundle, with their edges (e.g. DbTable,ExternalClass) (#1323)"),
     context: Optional[str] = typer.Option(None, "--context", "-c", help="Specific context to use"),
 ):
     """
@@ -760,6 +761,14 @@ def bundle_export(
             include_stats=not no_stats,
             sign_key=sign_key,
             encrypt_password=encrypt_password,
+            # bundle_export is also called programmatically (the golden
+            # harness does), where unpassed Typer options arrive as OptionInfo
+            # objects rather than their defaults — same trap the CLI meta-test
+            # documents for other commands.
+            exclude_labels=[
+                l for l in (exclude_labels if isinstance(exclude_labels, str) else "").split(",")
+                if l.strip()
+            ],
         )
         
         if success:
@@ -1137,6 +1146,7 @@ def export_shortcut(
     no_stats: bool = typer.Option(False, "--no-stats", help="Skip generating statistics in the bundle"),
     sign_key: Optional[str] = typer.Option(None, "--sign-key", envvar="CGC_BUNDLE_SIGN_KEY", help="HMAC signing key (or set CGC_BUNDLE_SIGN_KEY)"),
     encrypt_password: Optional[str] = typer.Option(None, "--encrypt-password", envvar="CGC_BUNDLE_PASSWORD", help="Encrypt bundle with this password (or set CGC_BUNDLE_PASSWORD)", hide_input=True),
+    exclude_labels: Optional[str] = typer.Option(None, "--exclude-labels", help="Comma-separated node labels to leave out of the bundle, with their edges (e.g. DbTable,ExternalClass) (#1323)"),
     context: Optional[str] = typer.Option(None, "--context", "-c", help="Specific context to use"),
 ):
     """Shortcut for 'cgc bundle export'"""
@@ -1146,6 +1156,7 @@ def export_shortcut(
         no_stats=no_stats,
         sign_key=sign_key,
         encrypt_password=encrypt_password,
+        exclude_labels=exclude_labels,
         context=context,
     )
 

{codegraphcontext-0.6.9 → codegraphcontext-0.6.11}/src/codegraphcontext/core/cgc_bundle.py

@@ -199,6 +199,7 @@ class CGCBundle:
         include_stats: bool = True,
         sign_key: Optional[str] = None,
         encrypt_password: Optional[str] = None,
+        exclude_labels: Optional[List[str]] = None,
     ) -> Tuple[bool, str]:
         """
         Export the current graph (or a specific repository) to a .cgc bundle.
@@ -225,6 +226,12 @@ class CGCBundle:
                 # Step 1: Extract metadata base
                 info_logger("Extracting metadata...")
                 metadata = self._extract_metadata(repo_path)
+                if exclude_labels:
+                    # Recorded for transparency: a consumer can tell a bundle
+                    # was exported without these node types (#1323).
+                    metadata["excluded_labels"] = sorted(
+                        {l.strip() for l in exclude_labels if l.strip()}
+                    )
                 
                 # Step 2: Extract schema
                 info_logger("Extracting schema...")
@@ -234,7 +241,7 @@ class CGCBundle:
                 
                 # Step 3: Extract nodes
                 info_logger("Extracting nodes...")
-                node_count = self._extract_nodes(temp_path / "nodes.jsonl", repo_path)
+                node_count = self._extract_nodes(temp_path / "nodes.jsonl", repo_path, exclude_labels=exclude_labels)
                 if node_count == 0:
                     return False, (
                         "No nodes to export. Index the repository first or verify "
@@ -244,6 +251,7 @@ class CGCBundle:
                 # Step 4: Extract edges
                 info_logger("Extracting edges...")
                 edge_count = self._extract_edges(temp_path / "edges.jsonl", repo_path)
+                # (edge filtering uses the id set _extract_nodes recorded)
                 
                 # Step 5: Generate statistics and assemble standardized metadata
                 if include_stats:
@@ -667,10 +675,19 @@ class CGCBundle:
             """,
         ]
     
-    def _extract_nodes(self, output_file: Path, repo_path: Optional[Path]) -> int:
-        """Extract all nodes to JSONL format."""
+    def _extract_nodes(self, output_file: Path, repo_path: Optional[Path], exclude_labels: Optional[List[str]] = None) -> int:
+        """Extract all nodes to JSONL format.
+
+        ``exclude_labels`` drops nodes carrying any of the given labels
+        (#1323 — e.g. DbTable/ExternalClass datasource metadata that path
+        filters can never reach). Their ids are recorded so
+        ``_extract_edges`` drops the touching edges too, keeping the bundle
+        free of dangling endpoints.
+        """
         count = 0
         seen_nodes = set()
+        excluded = {l.strip() for l in (exclude_labels or []) if l.strip()}
+        self._excluded_node_ids = set()
         
         with self.db_manager.get_driver(self._active_graph).session() as session:
             if repo_path:
@@ -691,6 +708,12 @@ class CGCBundle:
                         if node_key in seen_nodes:
                             continue
                         seen_nodes.add(node_key)
+
+                        if excluded and any(l in excluded for l in labels):
+                            self._excluded_node_ids.add(
+                                self._stable_id_key(self._export_id_of(node, node_dict))
+                            )
+                            continue
                     
                         # Clean up absolute path prefix to keep it relative
                         if repo_path:
@@ -718,6 +741,24 @@ class CGCBundle:
         
         return count
     
+    @staticmethod
+    def _stable_id_key(node_id) -> str:
+        """A hashable, backend-agnostic key for a node id (dict for Kùzu,
+        string for Neo4j/FalkorDB)."""
+        if isinstance(node_id, dict):
+            return json.dumps(node_id, sort_keys=True, default=str)
+        return str(node_id)
+
+    @staticmethod
+    def _export_id_of(node, node_dict):
+        if '_id' in node_dict:
+            return node_dict['_id']
+        if hasattr(node, 'element_id'):
+            return node.element_id
+        if hasattr(node, 'id'):
+            return str(node.id)
+        return None
+
     def _extract_edges(self, output_file: Path, repo_path: Optional[Path]) -> int:
         """Extract all relationships to JSONL format."""
         count = 0
@@ -779,6 +820,16 @@ class CGCBundle:
                             else:
                                 to_id = str(id(target))
 
+                        # Drop edges touching a label-excluded node (#1323):
+                        # nodes.jsonl no longer carries the endpoint, and a
+                        # dangling reference is an import failure waiting.
+                        excluded_ids = getattr(self, "_excluded_node_ids", None)
+                        if excluded_ids and (
+                            self._stable_id_key(from_id) in excluded_ids
+                            or self._stable_id_key(to_id) in excluded_ids
+                        ):
+                            continue
+
                         # Clean up absolute path prefix inside edge properties
                         if repo_path:
                             repo_str = repo_path.resolve().as_posix()

{codegraphcontext-0.6.9 → codegraphcontext-0.6.11}/src/codegraphcontext/core/database_embedded_kuzu.py

@@ -23,6 +23,40 @@ from codegraphcontext.core.graph_query import GraphQueryInterface
 from ..utils.cypher_ddl import is_schema_ddl
 from codegraphcontext.utils.debug_log import debug_log, info_logger, error_logger, warning_logger
 
+# Default cap for ladybug/kuzu buffer pools (bytes). Both libraries otherwise
+# default to ~80% of system RAM, which lets long-running gateways grow huge.
+DEFAULT_EMBEDDED_BUFFER_POOL_BYTES = 4 * 1024**3
+
+
+def resolve_embedded_buffer_pool_size() -> int:
+    """
+    Resolve ``buffer_pool_size`` for embedded ``Database(...)`` constructors.
+
+    Reads ``CGC_EMBEDDED_BUFFER_POOL_MB`` (integer MiB). Unset uses a 4 GiB
+    default. ``0`` opts back into the library default (~80% of system memory).
+    Invalid values log a warning and fall back to 4 GiB.
+    """
+    raw = os.getenv("CGC_EMBEDDED_BUFFER_POOL_MB")
+    if raw is None or str(raw).strip() == "":
+        return DEFAULT_EMBEDDED_BUFFER_POOL_BYTES
+    try:
+        mb = int(str(raw).strip())
+    except ValueError:
+        warning_logger(
+            f"Invalid CGC_EMBEDDED_BUFFER_POOL_MB={raw!r}; "
+            f"using default {DEFAULT_EMBEDDED_BUFFER_POOL_BYTES} bytes"
+        )
+        return DEFAULT_EMBEDDED_BUFFER_POOL_BYTES
+    if mb < 0:
+        warning_logger(
+            f"Invalid CGC_EMBEDDED_BUFFER_POOL_MB={raw!r}; "
+            f"using default {DEFAULT_EMBEDDED_BUFFER_POOL_BYTES} bytes"
+        )
+        return DEFAULT_EMBEDDED_BUFFER_POOL_BYTES
+    if mb == 0:
+        return 0
+    return mb * 1024**2
+
 
 @dataclass(frozen=True)
 class EmbeddedBackendSpec:
@@ -117,8 +151,18 @@ class EmbeddedGraphManager(GraphQueryInterface):
                     max_retries = 5
                     for attempt in range(max_retries):
                         try:
-                            info_logger(f"Initializing {spec.display_name} at {self.db_path}")
-                            self._db = backend.Database(self.db_path)
+                            buffer_pool_size = resolve_embedded_buffer_pool_size()
+                            if buffer_pool_size == 0:
+                                pool_msg = "library default (~80% of system memory)"
+                            else:
+                                pool_msg = f"{buffer_pool_size} bytes"
+                            info_logger(
+                                f"Initializing {spec.display_name} at {self.db_path} "
+                                f"(buffer_pool_size={pool_msg})"
+                            )
+                            self._db = backend.Database(
+                                self.db_path, buffer_pool_size=buffer_pool_size
+                            )
 
                             # Initialise connection pool
                             self._pool = queue.Queue()

{codegraphcontext-0.6.9 → codegraphcontext-0.6.11}/src/codegraphcontext/tools/languages/elisp.py

@@ -242,11 +242,26 @@ class ElispTreeSitterParser:
         return None
 
     def _first_symbol_in_quote(self, node: Any) -> Optional[str]:
-        if node.type != "quote":
+        # (quote symbol) -> symbol
+        if node.type == "quote":
+            for child in self._named_children(node):
+                if child.type == "symbol":
+                    return self._get_node_text(child)
             return None
-        for child in self._named_children(node):
-            if child.type == "symbol":
-                return self._get_node_text(child)
+        # #'symbol (function quote): tree-sitter-elisp parses `(funcall #'fn ...)`
+        # as (special_form (symbol "function") (symbol "fn")). Without this branch
+        # funcall/apply targets are dropped, so function_calls never records them
+        # and elisp call tests fail (KeyError / missing call edges).
+        if node.type == "special_form" and self._form_head(node) == "function":
+            named = self._named_children(node)
+            if len(named) >= 2 and named[1].type == "symbol":
+                return self._get_node_text(named[1])
+            return None
+        # Some grammar builds expose #'/quote as a `quoted_symbol` node wrapping a
+        # single symbol child; cover that variant too.
+        named = self._named_children(node)
+        if len(named) == 1 and named[0].type == "symbol":
+            return self._get_node_text(named[0])
         return None
 
     def _quoted_symbol_argument(self, node: Any, index: int = 1) -> Optional[str]:

{codegraphcontext-0.6.9 → codegraphcontext-0.6.11/src/codegraphcontext.egg-info}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: codegraphcontext
-Version: 0.6.9
+Version: 0.6.11
 Summary: An MCP server that indexes local code into a graph database to provide context to AI assistants.
 Author-email: Shashank Shekhar Singh <shashankshekharsingh1205@gmail.com>
 License: MIT License
@@ -388,6 +388,8 @@ CodeGraphContext supports multiple graph database backends to suit your environm
 | **Speed** | ⚡ Extremely Fast | ⚡ Fast | 🚀 Scalable | 🌐 Network-dependent |
 | **Persistence**| Yes (to disk) | Yes (to disk) | Yes (to disk) | Yes (server-side) |
 
+Embedded KuzuDB/LadybugDB default to a **4 GiB** buffer pool (`CGC_EMBEDDED_BUFFER_POOL_MB`). Set that env var to another MiB value, or `0` for the library default (~80% of system RAM).
+
 ---
 
 ## SCIP indexing (optional)
@@ -873,6 +875,14 @@ Once the server is running, you can interact with it through your AI assistant u
 
 ---
 
+## Troubleshooting
+
+Hitting a lock error, a "never finishes indexing" loop, or a silent feature?
+See [docs/TROUBLESHOOTING.md](docs/TROUBLESHOOTING.md) — every entry maps a
+real reported issue to its fix or workaround.
+
+---
+
 ## Contributing
 
 Contributions are welcome! 🎉