code-graph-rag 0.2.24__tar.gz → 0.3.13__tar.gz

{code_graph_rag-0.2.24/code_graph_rag.egg-info → code_graph_rag-0.3.13}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: code-graph-rag
-Version: 0.2.24
+Version: 0.3.13
 Summary: The ultimate RAG for your monorepo. Query, understand, and edit multi-language codebases with the power of AI and knowledge graphs
 License-Expression: MIT
 Keywords: rag,retrieval-augmented-generation,knowledge-graph,code-analysis,tree-sitter,mcp,mcp-server,llm,graph-database,semantic-search,codebase,memgraph,developer-tools,monorepo
@@ -38,6 +38,7 @@ Requires-Dist: defusedxml>=0.7.1
 Requires-Dist: huggingface-hub[hf-xet]>=1.7.2
 Requires-Dist: pathspec>=1.0.4
 Requires-Dist: pygments>=2.20.0
+Requires-Dist: certifi>=2026.10.23
 Requires-Dist: regex>=2024.11.6
 Requires-Dist: cgr-graph-rs<0.2,>=0.1.0; platform_python_implementation == "CPython" and ((sys_platform == "linux" and (platform_machine == "x86_64" or platform_machine == "aarch64")) or sys_platform == "darwin" or (sys_platform == "win32" and platform_machine == "AMD64"))
 Provides-Extra: test
@@ -78,10 +79,13 @@ Requires-Dist: tree-sitter-sql>=0.3.11; extra == "treesitter-full"
 Requires-Dist: tree-sitter-markdown>=0.5.1; extra == "treesitter-full"
 Provides-Extra: semantic
 Requires-Dist: qdrant-client>=1.9.0; extra == "semantic"
+Requires-Dist: grpcio>=1.84.1; extra == "semantic"
+Requires-Dist: sqlite-vec>=0.1.9; ((sys_platform == "linux" and (platform_machine == "x86_64" or platform_machine == "aarch64")) or sys_platform == "darwin" or (sys_platform == "win32" and platform_machine == "AMD64")) and extra == "semantic"
 Requires-Dist: torch>=2.6.0; extra == "semantic"
 Requires-Dist: transformers>=5.0.0; extra == "semantic"
 Provides-Extra: milvus
 Requires-Dist: pymilvus[milvus-lite]>=3.0.0; extra == "milvus"
+Requires-Dist: grpcio>=1.84.1; extra == "milvus"
 Provides-Extra: sqlite-vec
 Requires-Dist: sqlite-vec>=0.1.9; extra == "sqlite-vec"
 Provides-Extra: ast-grep
@@ -135,12 +139,12 @@ With semantic code search (UniXcoder embeddings):
 pip install 'code-graph-rag[semantic]'
 ```
 
-Qdrant is the default vector store for semantic search. To use Milvus Lite,
-install `code-graph-rag[semantic,milvus]`, then set
+The vectors go to a SQLite file beside the embedded graph
+(`~/.cgr/vectors.sqlite`), so semantic search needs no server and no Docker
+either. To use Qdrant instead, set `CGR_VECTOR_STORE_BACKEND=qdrant`. To use
+Milvus Lite, install `code-graph-rag[semantic,milvus]`, then set
 `CGR_VECTOR_STORE_BACKEND=milvus` and `MILVUS_URI=./.milvus_code_embeddings.db`
-before indexing. To keep the vectors in a SQLite file beside the embedded graph
-(`~/.cgr/vectors.sqlite`, no server), install `code-graph-rag[semantic,sqlite-vec]`
-and set `CGR_VECTOR_STORE_BACKEND=sqlite`.
+before indexing.
 
 To compute embeddings on an OpenAI-compatible endpoint (OpenAI, Ollama, vLLM)
 instead of locally, set `CGR_EMBEDDING_PROVIDER=openai` with

{code_graph_rag-0.2.24 → code_graph_rag-0.3.13}/PYPI_README.md

@@ -42,12 +42,12 @@ With semantic code search (UniXcoder embeddings):
 pip install 'code-graph-rag[semantic]'
 ```
 
-Qdrant is the default vector store for semantic search. To use Milvus Lite,
-install `code-graph-rag[semantic,milvus]`, then set
+The vectors go to a SQLite file beside the embedded graph
+(`~/.cgr/vectors.sqlite`), so semantic search needs no server and no Docker
+either. To use Qdrant instead, set `CGR_VECTOR_STORE_BACKEND=qdrant`. To use
+Milvus Lite, install `code-graph-rag[semantic,milvus]`, then set
 `CGR_VECTOR_STORE_BACKEND=milvus` and `MILVUS_URI=./.milvus_code_embeddings.db`
-before indexing. To keep the vectors in a SQLite file beside the embedded graph
-(`~/.cgr/vectors.sqlite`, no server), install `code-graph-rag[semantic,sqlite-vec]`
-and set `CGR_VECTOR_STORE_BACKEND=sqlite`.
+before indexing.
 
 To compute embeddings on an OpenAI-compatible endpoint (OpenAI, Ollama, vLLM)
 instead of locally, set `CGR_EMBEDDING_PROVIDER=openai` with

{code_graph_rag-0.2.24 → code_graph_rag-0.3.13}/README.md

@@ -61,8 +61,11 @@ Code-Graph-RAG parses a multi-language codebase with Tree-sitter, sharpened by c
 ## Latest News 🔥
 
 <!-- SECTION:latest_news -->
-- **Security**: Linked dependency advisories in the security notice to the global database.
-- **Build**: Enabled building the gotypes helper on Go 1.27.2 with x/mod v0.42.0.
+- **Security Enhancements**: Incorporates an OSS Scanner build image and threat model.
+- **Data Export**: Ensures all node properties within the graph are exported.
+- **SQL Fixes**: Addresses several SQL-related issues including grammar errors and dead-code roots.
+- **Rust Fixes**: Improves indexing of methods and handling of types in Rust code.
+- **Build Fixes**: Repairs build issues related to eMudhra root certificates and distrusted components.
 <!-- /SECTION:latest_news -->
 
 See [NEWS.md](NEWS.md) for the full history.
@@ -189,6 +192,7 @@ Code-Graph-RAG runs as an [MCP](https://modelcontextprotocol.io) server so Claud
 - [Real-Time Updates](docs/guide/realtime-updates.md)
 - [C/C++ Semantic Mode](docs/guide/cpp-semantic-mode.md)
 - [MCP Server](docs/guide/mcp-server.md)
+- [Porting Between Languages](docs/guide/porting.md)
 
 **Architecture**
 - [Overview](docs/architecture/overview.md)

{code_graph_rag-0.2.24 → code_graph_rag-0.3.13/code_graph_rag.egg-info}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: code-graph-rag
-Version: 0.2.24
+Version: 0.3.13
 Summary: The ultimate RAG for your monorepo. Query, understand, and edit multi-language codebases with the power of AI and knowledge graphs
 License-Expression: MIT
 Keywords: rag,retrieval-augmented-generation,knowledge-graph,code-analysis,tree-sitter,mcp,mcp-server,llm,graph-database,semantic-search,codebase,memgraph,developer-tools,monorepo
@@ -38,6 +38,7 @@ Requires-Dist: defusedxml>=0.7.1
 Requires-Dist: huggingface-hub[hf-xet]>=1.7.2
 Requires-Dist: pathspec>=1.0.4
 Requires-Dist: pygments>=2.20.0
+Requires-Dist: certifi>=2026.10.23
 Requires-Dist: regex>=2024.11.6
 Requires-Dist: cgr-graph-rs<0.2,>=0.1.0; platform_python_implementation == "CPython" and ((sys_platform == "linux" and (platform_machine == "x86_64" or platform_machine == "aarch64")) or sys_platform == "darwin" or (sys_platform == "win32" and platform_machine == "AMD64"))
 Provides-Extra: test
@@ -78,10 +79,13 @@ Requires-Dist: tree-sitter-sql>=0.3.11; extra == "treesitter-full"
 Requires-Dist: tree-sitter-markdown>=0.5.1; extra == "treesitter-full"
 Provides-Extra: semantic
 Requires-Dist: qdrant-client>=1.9.0; extra == "semantic"
+Requires-Dist: grpcio>=1.84.1; extra == "semantic"
+Requires-Dist: sqlite-vec>=0.1.9; ((sys_platform == "linux" and (platform_machine == "x86_64" or platform_machine == "aarch64")) or sys_platform == "darwin" or (sys_platform == "win32" and platform_machine == "AMD64")) and extra == "semantic"
 Requires-Dist: torch>=2.6.0; extra == "semantic"
 Requires-Dist: transformers>=5.0.0; extra == "semantic"
 Provides-Extra: milvus
 Requires-Dist: pymilvus[milvus-lite]>=3.0.0; extra == "milvus"
+Requires-Dist: grpcio>=1.84.1; extra == "milvus"
 Provides-Extra: sqlite-vec
 Requires-Dist: sqlite-vec>=0.1.9; extra == "sqlite-vec"
 Provides-Extra: ast-grep
@@ -135,12 +139,12 @@ With semantic code search (UniXcoder embeddings):
 pip install 'code-graph-rag[semantic]'
 ```
 
-Qdrant is the default vector store for semantic search. To use Milvus Lite,
-install `code-graph-rag[semantic,milvus]`, then set
+The vectors go to a SQLite file beside the embedded graph
+(`~/.cgr/vectors.sqlite`), so semantic search needs no server and no Docker
+either. To use Qdrant instead, set `CGR_VECTOR_STORE_BACKEND=qdrant`. To use
+Milvus Lite, install `code-graph-rag[semantic,milvus]`, then set
 `CGR_VECTOR_STORE_BACKEND=milvus` and `MILVUS_URI=./.milvus_code_embeddings.db`
-before indexing. To keep the vectors in a SQLite file beside the embedded graph
-(`~/.cgr/vectors.sqlite`, no server), install `code-graph-rag[semantic,sqlite-vec]`
-and set `CGR_VECTOR_STORE_BACKEND=sqlite`.
+before indexing.
 
 To compute embeddings on an OpenAI-compatible endpoint (OpenAI, Ollama, vLLM)
 instead of locally, set `CGR_EMBEDDING_PROVIDER=openai` with

{code_graph_rag-0.2.24 → code_graph_rag-0.3.13}/code_graph_rag.egg-info/SOURCES.txt

@@ -52,6 +52,8 @@ codebase_rag/models.py
 codebase_rag/parser_fingerprint.py
 codebase_rag/parser_loader.py
 codebase_rag/path_filters.py
+codebase_rag/port_cli.py
+codebase_rag/port_parity.py
 codebase_rag/project_selection.py
 codebase_rag/prompts.py
 codebase_rag/readme_sections.py
@@ -148,6 +150,7 @@ codebase_rag/constants/lang_tooling.py
 codebase_rag/constants/languages.py
 codebase_rag/constants/mcp.py
 codebase_rag/constants/notebook.py
+codebase_rag/constants/port.py
 codebase_rag/constants/providers.py
 codebase_rag/constants/security.py
 codebase_rag/constants/sqlite_vectors.py
@@ -223,6 +226,8 @@ codebase_rag/parsers/positional_params.py
 codebase_rag/parsers/python_source_roots.py
 codebase_rag/parsers/rpc_exposure.py
 codebase_rag/parsers/semantic_call_join.py
+codebase_rag/parsers/sql_ddl.py
+codebase_rag/parsers/sql_recovery.py
 codebase_rag/parsers/stdlib_extractor.py
 codebase_rag/parsers/string_call.py
 codebase_rag/parsers/structure_processor.py

{code_graph_rag-0.2.24 → code_graph_rag-0.3.13}/code_graph_rag.egg-info/requires.txt

@@ -18,6 +18,7 @@ defusedxml>=0.7.1
 huggingface-hub[hf-xet]>=1.7.2
 pathspec>=1.0.4
 pygments>=2.20.0
+certifi>=2026.10.23
 regex>=2024.11.6
 
 [:platform_python_implementation == "CPython" and ((sys_platform == "linux" and (platform_machine == "x86_64" or platform_machine == "aarch64")) or sys_platform == "darwin" or (sys_platform == "win32" and platform_machine == "AMD64"))]
@@ -35,6 +36,7 @@ pymgclient>=1.6.0
 
 [milvus]
 pymilvus[milvus-lite]>=3.0.0
+grpcio>=1.84.1
 
 [neo4j]
 neo4j>=5.28.1
@@ -44,9 +46,13 @@ jedi>=0.19
 
 [semantic]
 qdrant-client>=1.9.0
+grpcio>=1.84.1
 torch>=2.6.0
 transformers>=5.0.0
 
+[semantic:(sys_platform == "linux" and (platform_machine == "x86_64" or platform_machine == "aarch64")) or sys_platform == "darwin" or (sys_platform == "win32" and platform_machine == "AMD64")]
+sqlite-vec>=0.1.9
+
 [sqlite-vec]
 sqlite-vec>=0.1.9
 

{code_graph_rag-0.2.24 → code_graph_rag-0.3.13}/codebase_rag/cli.py

@@ -59,6 +59,7 @@ from .editor_links import (
 from .graph_cli import cli as graph_cli
 from .graph_dialects import DIALECT_MEMGRAPH
 from .parser_loader import load_parsers
+from .port_cli import cli as port_cli
 from .project_selection import ProjectChoice, choose_project, scope_refusal
 from .services.gloss_cleanup import prune_orphaned_glosses
 from .services.graph_diff import DiffError, diff_indexes, diff_is_empty
@@ -72,7 +73,7 @@ from .services.provenance import (
 )
 from .stack import StackManager
 from .stack.cli import cli as daemon_cli
-from .stack.constants import SERVICE_QDRANT, StackState
+from .stack.constants import SERVICE_MEMGRAPH, SERVICE_QDRANT, StackState
 from .stack.manager import StackError
 from .sync_guard import RUNNING_SYNC_TOKENS as _RUNNING_SYNC_TOKENS
 from .sync_guard import OwnerCheck as _OwnerCheck
@@ -775,15 +776,25 @@ def _resolve_active_projects(projects: str | None, default_project: str) -> list
     return [default_project]
 
 
-def _maybe_start_stack() -> None:
+def _maybe_start_stack(skip_embeddings: bool = False) -> None:
     # The Docker stack serves Memgraph, and Qdrant when the vector settings
     # pick the bundled one. The embedded graph is a file and Neo4j is the
     # user's own server, so for them the stack runs only for the vectors:
     # only its Qdrant is started and checked, since every Memgraph check
-    # needs a driver those installs need not have.
+    # needs a driver those installs need not have. A run that writes no
+    # embeddings, or writes them to another backend, needs no Qdrant, so it
+    # never starts one: for the embedded graph that means no Docker at all.
+    wants_vectors = not (skip_embeddings or settings.SKIP_EMBEDDINGS)
     if settings.GRAPH_BACKEND == DIALECT_MEMGRAPH:
-        mgr = StackManager()
+        mgr = (
+            StackManager()
+            if wants_vectors
+            and settings.VECTOR_STORE_BACKEND == cs.VectorStoreBackend.QDRANT
+            else StackManager(services=(SERVICE_MEMGRAPH,))
+        )
     else:
+        if not wants_vectors:
+            return
         mgr = StackManager(services=(SERVICE_QDRANT,))
         if not mgr.serves_vectors():
             return
@@ -1479,7 +1490,7 @@ def start(
     _refuse_empty_workspace(workspace_config, update_graph, active_projects)
 
     if not no_start_stack:
-        _maybe_start_stack()
+        _maybe_start_stack(skip_embeddings=no_embeddings)
 
     effective_batch_size = settings.resolve_batch_size(batch_size)
 
@@ -2166,6 +2177,18 @@ def graph_command(ctx: typer.Context) -> None:
     _run_delegated_group(graph_cli, ctx)
 
 
+@app.command(
+    name=ch.CLICommandName.PORT,
+    help=ch.CMD_PORT,
+    short_help=ch.CMD_PORT,
+    add_help_option=False,
+    context_settings=_DELEGATED_GROUP_CONTEXT,
+    rich_help_panel=ch.PANEL_USE,
+)
+def port_command(ctx: typer.Context) -> None:
+    _run_delegated_group(port_cli, ctx)
+
+
 @app.command(
     name=ch.CLICommandName.CHECK,
     help=ch.CMD_CHECK,

{code_graph_rag-0.2.24 → code_graph_rag-0.3.13}/codebase_rag/cli_help.py

@@ -23,6 +23,7 @@ class CLICommandName(StrEnum):
     CHECK = "check"
     RENAME = "rename"
     CONTEXT = "context"
+    PORT = "port"
     STOP = "stop"
     STATUS = "status"
     HELP = "help"
@@ -206,6 +207,43 @@ EXAMPLES_CONTEXT = (
 )
 HELP_CONTEXT_TARGET = "Qualified name, bare name, path:line, or a free-text task."
 HELP_CONTEXT_BUDGET = "Token budget for the slice."
+CMD_PORT = "Judge a port between languages: pair the functions of two implementations and compare their calls and arity, as JSON, no LLM."
+CMD_PORT_GROUP = CMD_PORT
+CMD_PORT_CHECK = (
+    "Pair the functions and methods under OLD_SCOPE with those under "
+    "NEW_SCOPE (paths relative to the repository root) and judge each pair "
+    "MATCH, DIFFERS or UNKNOWN on the first-party calls it makes and its "
+    "positional arity; unpaired and ambiguous definitions are listed."
+)
+EPILOG_PORT = "Run 'cgr help port COMMAND' for command-specific help."
+# `\b` opens a paragraph click prints as written instead of rewrapping it,
+# which keeps one example per line.
+EPILOG_PORT_CHECK = (
+    "\b\nExamples:\n  cgr port check src/legacy src/rewrite\n"
+    "  cgr port check codebase_rag/embedded_graph rust/cgr_graph_rs/src "
+    "--fail-on-found\n"
+    "  cgr port check py rs --pair py.lexer._lex=rs.lexer.tokenize\n\n"
+    "Exit status: 0 with the JSON report; 1 with --fail-on-found when a pair "
+    "DIFFERS; 2 when the scopes overlap or a --pair is malformed or repeated; "
+    "3 when a project is not indexed; 4 when a scope holds no function or "
+    "method, or a --pair names none or several."
+)
+HELP_PORT_NEW_PROJECT = (
+    "Project holding NEW_SCOPE, when the new implementation was indexed as a "
+    "project of its own (default: the project of OLD_SCOPE)."
+)
+HELP_PORT_PAIR = (
+    "Pair two definitions explicitly, OLD=NEW, each a qualified name or a "
+    "dotted suffix naming exactly one definition of its side. Repeatable."
+)
+HELP_PORT_INCLUDE_TESTS = (
+    "Pair test functions too; by default they are left out, classified as "
+    "tests-reaching classifies them."
+)
+HELP_PORT_FAIL_ON_FOUND = (
+    "Exit with status 1 when any pair DIFFERS. UNKNOWN pairs and unpaired "
+    "definitions do not fail the check."
+)
 CMD_TRACE_INGEST = "Resolve a trace file against a project and write dynamic edges"
 CMD_TRACE_CONVERT = "Convert a V8 .cpuprofile (node --cpu-prof) to a trace file"
 
@@ -678,6 +716,7 @@ CLI_COMMANDS: dict[CLICommandName, str] = {
     CLICommandName.CHECK: CMD_CHECK,
     CLICommandName.RENAME: CMD_RENAME,
     CLICommandName.CONTEXT: CMD_CONTEXT,
+    CLICommandName.PORT: CMD_PORT,
     CLICommandName.WORKSPACE: CMD_WORKSPACE,
     CLICommandName.STOP: CMD_STOP,
     CLICommandName.STATUS: CMD_STATUS,

{code_graph_rag-0.2.24 → code_graph_rag-0.3.13}/codebase_rag/config.py

@@ -452,8 +452,11 @@ class AppConfig(BaseSettings):
     QDRANT_UPSERT_RETRIES: int = Field(default=3, gt=0)
     QDRANT_RETRY_BASE_DELAY: float = Field(default=0.5, gt=0)
     QDRANT_BATCH_SIZE: int = Field(default=50, gt=0)
+    # The `sqlite` store is the default for the reason the embedded graph is:
+    # a file under CGR_HOME needs no server and no Docker. Qdrant and Milvus
+    # stay one setting away.
     VECTOR_STORE_BACKEND: cs.VectorStoreBackend = Field(
-        cs.VectorStoreBackend.QDRANT, validation_alias="CGR_VECTOR_STORE_BACKEND"
+        cs.VectorStoreBackend.SQLITE, validation_alias="CGR_VECTOR_STORE_BACKEND"
     )
     MILVUS_URI: str = "./.milvus_code_embeddings.db"
     MILVUS_TOKEN: str | None = None
@@ -469,6 +472,18 @@ class AppConfig(BaseSettings):
     SQLITE_VECTOR_DIM: int = 768
     SQLITE_VECTOR_TOP_K: int = 5
 
+    @property
+    def qdrant_set_without_backend(self) -> bool:
+        """Whether Qdrant is configured but the vector store was left to default.
+
+        Qdrant was the default vector store before the `sqlite` one replaced
+        it, so this is how an install upgraded in place looks: its embeddings
+        live in Qdrant while cgr now opens an empty file.
+        """
+        return cs.SETTING_VECTOR_STORE_BACKEND not in self.model_fields_set and any(
+            name in self.model_fields_set for name in cs.QDRANT_CONNECTION_SETTINGS
+        )
+
     @property
     def vector_db_file(self) -> Path:
         if self.SQLITE_VECTOR_PATH is not None:

{code_graph_rag-0.2.24 → code_graph_rag-0.3.13}/codebase_rag/constants/__init__.py

@@ -30,6 +30,7 @@ from .lang_tooling import *  # noqa: F403
 from .languages import *  # noqa: F403
 from .mcp import *  # noqa: F403
 from .notebook import *  # noqa: F403
+from .port import *  # noqa: F403
 from .providers import *  # noqa: F403
 from .security import *  # noqa: F403
 from .sqlite_vectors import *  # noqa: F403

{code_graph_rag-0.2.24 → code_graph_rag-0.3.13}/codebase_rag/constants/ast_rust.py

@@ -113,6 +113,14 @@ TS_RS_LOOP_EXPRESSION = "loop_expression"
 # A Rust call node whose callee is descended for chain flattening: a plain call
 # or a turbofish generic_function (`f::<T>()`).
 RS_CALL_OR_GENERIC_FN = (TS_RS_CALL_EXPRESSION, TS_GENERIC_FUNCTION)
+TS_RS_GENERIC_TYPE_WITH_TURBOFISH = "generic_type_with_turbofish"
+# A constructor path with type arguments: `Tree::<u8>::Leaf`'s qualifier,
+# `Wrapper::<u8> { .. }`'s name, `Pair::<u8>(..)`'s callee.
+RS_GENERIC_CONSTRUCTOR_PATHS = (
+    TS_GENERIC_TYPE,
+    TS_RS_GENERIC_TYPE_WITH_TURBOFISH,
+    TS_GENERIC_FUNCTION,
+)
 TS_RS_TUPLE_STRUCT_PATTERN = "tuple_struct_pattern"
 TS_RS_TYPE_ARGUMENTS = "type_arguments"
 TS_RS_TRY_EXPRESSION = "try_expression"

{code_graph_rag-0.2.24 → code_graph_rag-0.3.13}/codebase_rag/constants/core.py

@@ -262,6 +262,9 @@ PRUNE_PENDING_FILENAME = ".cgr-prune-pending"
 # PR #3358). Not in CGR_STATE_FILENAMES: no cgr ever wrote one into the tree,
 # so there is nothing to move out of it.
 EMBEDDINGS_PENDING_PROJECT_TEMPLATE = ".cgr-embeddings-pending-{digest}"
+# The vector store backend a project's points were last settled in, one file
+# per project for the same reason; likewise never written into the tree.
+EMBEDDINGS_STORE_PROJECT_TEMPLATE = ".cgr-embeddings-store-{digest}"
 # The marker an earlier build shared between every project of a checkout: a
 # JSON list of the projects owing a pass, or an empty file. Read so that each
 # still retries once, then split into the per-project files above.

{code_graph_rag-0.2.24 → code_graph_rag-0.3.13}/codebase_rag/constants/languages.py

@@ -401,7 +401,7 @@ LANGUAGE_METADATA: dict[SupportedLanguage, LanguageMetadata] = {
     ),
     SupportedLanguage.SQL: LanguageMetadata(
         LanguageStatus.DEV,
-        "Stored functions (CREATE FUNCTION), schema-qualified names, invocations between routines. CREATE PROCEDURE and in-depth PL/pgSQL bodies await upstream grammar support: the published grammar parses plain SQL statements only",
+        "Stored functions (CREATE FUNCTION), each one a dead-code root, schema-qualified names, invocations between routines, and the DDL that runs a routine by name (trigger EXECUTE FUNCTION, aggregate SFUNC/FINALFUNC, cast WITH FUNCTION, operator FUNCTION, type INPUT/OUTPUT, HANDLER/VALIDATOR, SUPPORT). CREATE PROCEDURE and in-depth PL/pgSQL bodies await upstream grammar support: the published grammar parses plain SQL statements only",
         "SQL (PostgreSQL)",
     ),
     SupportedLanguage.JULIA: LanguageMetadata(

{code_graph_rag-0.2.24 → code_graph_rag-0.3.13}/codebase_rag/constants/mcp.py

@@ -246,6 +246,12 @@ MCP_ROOT_NOT_INDEXED = (
     "No project is indexed for this server's root {path}; run "
     "index_repository or update_repository first, or pass `project`."
 )
+# flow_verdict, explain_traceback and rank_root_causes take no `project`, so
+# their refusal cannot advise passing one (issue #3193).
+MCP_FIXED_ROOT_NOT_INDEXED = (
+    "No project is indexed for this server's root {path}; run "
+    "index_repository or update_repository first."
+)
 # An empty answer for a name the graph does not hold read as "nothing calls
 # it" / "no test reaches it" (issue #2461).
 MCP_UNKNOWN_TARGET = "{qualified_name!r} is not in the graph."

code_graph_rag-0.3.13/codebase_rag/constants/port.py

@@ -0,0 +1,88 @@
+# Cross-language parity (`cgr port check`, epic #3413): how two
+# implementations of the same code are paired and judged.
+
+from enum import StrEnum
+
+from .graph import EdgeResolution, NodeLabel
+
+
+class ParityVerdict(StrEnum):
+    """A dimension's or a pair's verdict. `UNKNOWN` is what the graph cannot
+    settle (an edge it only guessed, a signature it did not record), and is
+    never reported as `MATCH`."""
+
+    MATCH = "MATCH"
+    DIFFERS = "DIFFERS"
+    UNKNOWN = "UNKNOWN"
+
+
+# A pair's verdict is the worst of its dimensions'.
+PARITY_VERDICT_RANK: dict[ParityVerdict, int] = {
+    ParityVerdict.MATCH: 0,
+    ParityVerdict.UNKNOWN: 1,
+    ParityVerdict.DIFFERS: 2,
+}
+
+
+class PairedBy(StrEnum):
+    """How a pair was found, strongest first: a `--pair` the user gave; the
+    same normalised name at the same module path relative to each scope (and
+    in the same class, for a method); the same name in the same container
+    (the class, or the module file's stem) when the layouts differ; or the
+    same name alone with the containers differing, which is the pairing to
+    verify."""
+
+    EXPLICIT = "explicit"
+    MODULE = "module"
+    CONTAINER = "container"
+    NAME = "name"
+
+
+# The automatic tiers, tried in this order on what the previous one left.
+PORT_PAIRING_TIERS = (PairedBy.MODULE, PairedBy.CONTAINER, PairedBy.NAME)
+
+
+PORT_CALLABLE_LABELS = frozenset({NodeLabel.FUNCTION.value, NodeLabel.METHOD.value})
+
+# Edges the resolver only guessed: a difference resting on them alone is
+# UNKNOWN, since the other side may simply have bound the same call exactly.
+PORT_UNCERTAIN_RESOLUTIONS = frozenset(
+    {EdgeResolution.HEURISTIC.value, EdgeResolution.OVERLOAD.value}
+)
+
+# Folded before names are compared, so `_read_number`, `readNumber` and
+# `ReadNumber` agree. The boundary pattern splits camelCase and the end of an
+# acronym (`HTTPServer` -> `HTTP_Server`).
+PORT_NAME_STRIP_CHARS = "_"
+PORT_CAMEL_BOUNDARY_PATTERN = r"(?<=[a-z0-9])(?=[A-Z])|(?<=[A-Z])(?=[A-Z][a-z])"
+PORT_NAME_JOIN = "_"
+
+# File stems that name a package rather than its contents (`__init__.py`,
+# Rust `mod.rs` and `lib.rs`, `index.ts`): a function's container is then the
+# directory holding the file.
+PORT_PACKAGE_ENTRY_STEMS = frozenset({"__init__", "mod", "lib", "main", "index"})
+
+PORT_PAIR_SEPARATOR = "="
+PORT_SCOPE_ROOTS = frozenset({"", "."})
+
+# Overlapping scopes and a malformed or repeated `--pair` are usage errors,
+# reported under click's own status for one; a scope or pair naming nothing
+# in the graph exits with GRAPH_EXIT_UNKNOWN_TARGET like `cgr graph`.
+PORT_EXIT_USAGE = 2
+PORT_EXIT_DIFFERS = 1
+
+PORT_ERR_EMPTY_SCOPE = (
+    "No function or method of project '{project}' lies under '{scope}'. "
+    "A scope is a path relative to the repository root, as `cgr graph "
+    "definition` prints it."
+)
+PORT_ERR_OVERLAPPING_SCOPES = (
+    "The scopes '{old}' and '{new}' overlap in project '{project}': each "
+    "definition must belong to one side."
+)
+PORT_ERR_PAIR_FORMAT = "--pair takes OLD=NEW, got '{value}'."
+PORT_ERR_PAIR_TARGET = (
+    "--pair: '{target}' names {count} definitions under '{scope}'; give a "
+    "qualified name, or a dotted suffix of one, that names exactly one."
+)
+PORT_ERR_PAIR_TWICE = "--pair: '{qualified_name}' is paired more than once."

{code_graph_rag-0.2.24 → code_graph_rag-0.3.13}/codebase_rag/constants/providers.py

@@ -191,6 +191,13 @@ QDRANT_DEFAULT_DB_PATH = "./.qdrant_code_embeddings"
 # The settings field that holds it. Whether it was supplied is read from the
 # fields the settings sources set, since a supplied value can equal the default.
 SETTING_QDRANT_DB_PATH = "QDRANT_DB_PATH"
+# Settings whose presence, with VECTOR_STORE_BACKEND left unset, marks an
+# install that stored its embeddings in Qdrant when Qdrant was the default.
+SETTING_VECTOR_STORE_BACKEND = "VECTOR_STORE_BACKEND"
+QDRANT_CONNECTION_SETTINGS = ("QDRANT_URL", SETTING_QDRANT_DB_PATH, "QDRANT_API_KEY")
+# The backend a project synced before cgr stamped its vector store was embedded
+# into: the default then.
+EMBEDDINGS_STORE_UNSTAMPED = VectorStoreBackend.QDRANT
 
 SEMANTIC_DEPENDENCIES = (
     MODULE_PYMILVUS,

{code_graph_rag-0.2.24 → code_graph_rag-0.3.13}/codebase_rag/editing/rename.py

@@ -471,25 +471,39 @@ def _chain_links(
     if parser is None:
         return []
     root = parser.parse(source).root_node
+    # Each link's own site starts at its name (issue #3166); the links of
+    # its chain are the calls that start where the site's call starts.
+    starts = {call.start_point for call in calls_starting_at(root, line - 1, col)}
+    starts = starts or {(line - 1, col)}
     found: list[tuple[int, int]] = []
     stack: list[Node] = [root]
     while stack:
         node = stack.pop()
-        if node.start_point == (line - 1, col):
-            func = node.child_by_field_name(cs.FIELD_FUNCTION)
-            member = func.children[-1] if func is not None and func.children else None
-            if (
-                member is not None
-                and member.type in _MEMBER_NAME_TYPES
-                and member.text is not None
-                and member.text.decode(cs.ENCODING_UTF8, errors="replace") == name
-            ):
-                found.append((member.start_point[0] + 1, member.start_point[1]))
-        if node.start_point[0] <= line - 1 <= node.end_point[0]:
+        if (
+            node.start_point in starts
+            and (member := _called_member(node, name)) is not None
+        ):
+            found.append((member.start_point[0] + 1, member.start_point[1]))
+        if any(node.start_point[0] <= row <= node.end_point[0] for row, _ in starts):
             stack.extend(node.children)
     return sorted(found)
 
 
+def _called_member(call: Node, name: str) -> Node | None:
+    # The member name a call invokes (`name` in `obj.name(...)`), if it is
+    # `name`; None for a bare call or another member.
+    func = call.child_by_field_name(cs.FIELD_FUNCTION)
+    member = func.children[-1] if func is not None and func.children else None
+    if (
+        member is not None
+        and member.type in _MEMBER_NAME_TYPES
+        and member.text is not None
+        and member.text.decode(cs.ENCODING_UTF8, errors="replace") == name
+    ):
+        return member
+    return None
+
+
 def _jsx_closing_token(
     source: bytes,
     language: cs.SupportedLanguage | None,
@@ -614,8 +628,10 @@ def _last_identifier(
         # (a grammar whose call node has no `function` field, such as Java's
         # method_invocation) keeps the fallback.
         return _STALE_CALL
-    if callee is not None and callee[0] == start:
-        start, end = callee
+    if callee is not None and callee[0] <= start < callee[1]:
+        # The callee ends the token search, whether the site starts with it
+        # or at its member name (issue #3166): never inside the arguments.
+        end = callee[1]
     text = source[start:end].decode(cs.ENCODING_UTF8, errors="replace")
     if callee is None:
         # No grammar: cut at the last opening parenthesis so the arguments

{code_graph_rag-0.2.24 → code_graph_rag-0.3.13}/codebase_rag/editing/sites.py

@@ -17,6 +17,7 @@ from .. import constants as cs
 from .. import cypher_queries as cq
 from .. import graph_query
 from ..graph_query import QueryFn
+from ..parsers.utils import call_site_start
 
 # Resolutions that bound a site by guesswork: an operation rewrites through
 # them only when the caller accepts the risk with `allow_heuristic`.
@@ -123,13 +124,14 @@ def call_node_at(
 
 
 def calls_starting_at(root: Node, row: int, col: int) -> list[Node]:
+    # A member call's site starts at its name (issue #3166); a graph indexed
+    # before that recorded the call node's own start, so both name the call.
     calls: list[Node] = []
     stack: list[Node] = [root]
     while stack:
         node = stack.pop()
-        if (
-            node.start_point == (row, col)
-            and node.child_by_field_name(cs.FIELD_FUNCTION) is not None
+        if node.child_by_field_name(cs.FIELD_FUNCTION) is not None and (
+            node.start_point == (row, col) or call_site_start(node) == (row, col)
         ):
             calls.append(node)
         if node.start_point[0] <= row <= node.end_point[0]:

{code_graph_rag-0.2.24 → code_graph_rag-0.3.13}/codebase_rag/exceptions.py

@@ -119,12 +119,14 @@ SQLITE_VECTOR_DIM_MISMATCH = (
     "or re-index with --clean to rebuild the store with the new size."
 )
 SQLITE_VEC_NOT_INSTALLED = (
-    "CGR_VECTOR_STORE_BACKEND=sqlite needs sqlite-vec, which is not installed. "
-    "Install it with `pip install 'code-graph-rag[sqlite-vec]'` (or "
-    "`uv sync --extra sqlite-vec`)."
+    "The sqlite vector store (the default, CGR_VECTOR_STORE_BACKEND=sqlite) "
+    "needs sqlite-vec, which is not installed. Install it with "
+    "`pip install 'code-graph-rag[sqlite-vec]'` (or `uv sync --extra "
+    "sqlite-vec`); the `semantic` extra includes it."
 )
 SQLITE_VEC_EXTENSIONS_UNSUPPORTED = (
-    "CGR_VECTOR_STORE_BACKEND=sqlite loads the sqlite-vec extension, but this "
+    "The sqlite vector store (the default, CGR_VECTOR_STORE_BACKEND=sqlite) "
+    "loads the sqlite-vec extension, but this "
     "Python's sqlite3 module was built without extension loading. Use a Python "
     "that has it (uv-managed Pythons do), or set CGR_VECTOR_STORE_BACKEND=qdrant."
 )