cg-code-graph 0.11.0__tar.gz → 0.12.0__tar.gz

{cg_code_graph-0.11.0/cg_code_graph.egg-info → cg_code_graph-0.12.0}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: cg-code-graph
-Version: 0.11.0
+Version: 0.12.0
 Summary: cg: a code graph of routes, calls, data access and platform boundaries for agents and humans (CLI + MCP server)
 Author: cyberchronos00
 License-Expression: MIT
@@ -207,7 +207,7 @@ The 7 plan gaps are the admin update path, `UpdateBookRequest`, `Book::$fillable
 
 ### 1. A CLI for impact questions
 
-`reaches`, `impact`, `downstream`, `path`, `writers`, `siblings`, `routes`, `search`, `api-calls`, `channels`, `tests`
+`reaches`, `impact`, `downstream`, `path`, `writers`, `siblings`, `routes`, `search`, `snippet`, `api-calls`, `channels`, `tests`
 and `platforms` all work on one SQLite graph, and across repos once the frontend and backend are linked. Every hop shows its evidence:
 
 ```text
@@ -315,7 +315,7 @@ Full reference: [docs/cli.md](docs/cli.md) · value facts: [docs/value-facts.md]
 
 ### 2. An MCP server for AI agents
 
-The same queries as MCP tools (`reaches`, `impact`, `callers`, `siblings`, `path`, `downstream`, `routes`, `search`, `api_calls`,
+The same queries as MCP tools (`reaches`, `impact`, `callers`, `siblings`, `path`, `downstream`, `routes`, `search`, `snippet`, `api_calls`,
 `channels`, `tests_covering`, `resolutions`, `plan_check`, `index`, `coverage`, `starters`, `platform_divergence`, …), so an agent can check the blast radius before it edits. Replies are compact,
 use repo-relative paths, and `plan_check` starts with a summary (`details=true` for the full report). Every reply also
 carries a machine-readable `completeness` object, so the agent knows when an answer covers the whole repository and
@@ -386,7 +386,7 @@ without configuration; `cg config show` lists what was applied ([docs/configurat
 | Django REST Framework | resolved | routers, ViewSets (+ `@action`), APIView/generic views, `permission_classes`, serializer fields |
 | FastAPI / Starlette | resolved | `FastAPI` / `APIRouter` / `Starlette` / `Router` objects, `include_router` / `mount` prefix chains across files (prefixes from constants and settings attributes such as `settings.API_V1_STR`), verb decorators, `api_route`, `add_api_route`, Starlette `routes=[Route, Mount, WebSocketRoute]`, websockets, path parameters, `Depends()` / `Security()` dependencies (parameters, `Annotated` aliases, `dependencies=`) as route access with what each checks (raised 401 / 403, security schemes, nested dependencies), `name=` for `url_path_for`; apps / routers received as parameters or fixtures or returned by factories, Starlette `Host`, `@cbv` / `InferringRouter`, classy-fastapi `Routable` |
 | Flask | resolved | `Flask` / `Blueprint` objects, `register_blueprint` (`url_prefix` from the blueprint or the registration, nested blueprints), `@route` / verb decorators, `add_url_rule` incl. `MethodView.as_view()` (and its `methods`), endpoint-only rules, werkzeug `Rule` / `Submount`, the built-in static route, views defined in an app factory, apps received as parameters or pytest fixtures, flask-restful / flask-restx resources, `subdomain=` / `defaults=`, `<int:id>` parameters, `blueprint.endpoint` names for `url_for`, view decorators (`login_required`) as route access |
-| Celery / Channels | resolved | tasks + `.delay`/`.apply_async` dispatches; websocket routing to consumers |
+| Celery / RQ / Dramatiq / Channels | resolved | tasks + `.delay`/`.apply_async` dispatches; job and queue endpoints by task name across repos (`send_task`, RQ import paths, Dramatiq actors) with worker processes from Procfile / compose; websocket routing to consumers |
 | Dart | resolved (package:analyzer parse, declared types) + heuristic fallback | libraries/parts, classes, methods, functions, calls with import resolution |
 | Flutter | resolved + heuristic fallback | widgets/State, bloc/cubit events → handlers → states → UI, Navigator/go_router/auto_route pages, HTTP calls (package:http, Dio, dart:io, Retrofit/Chopper), WebSockets, json_serializable/freezed and hand-written JSON keys |
 | Rust | **exact** with rust-analyzer (SCIP); **heuristic** without | crates, modules, `pub` API, traits → impls (dyn/generic dispatch), bins, tests, benches, examples, `build.rs`, FFI, `unsafe`, `#[cfg(feature)]` gates, env keys, `#[tokio::main]`, axum/actix routes ([docs/native.md](docs/native.md)) |
@@ -401,7 +401,7 @@ without configuration; `cg config show` lists what was applied ([docs/configurat
 | Desktop processes | Electron `ipcMain` / `ipcRenderer` / `webContents.send` and `contextBridge.exposeInMainWorld`; Tauri `invoke` → `#[tauri::command]` | `endpoint:electron-ipc:<channel>`, `endpoint:electron-preload:<key>#<member>`, `endpoint:tauri:<command>` with SENDS_TO / RECEIVED_BY across the processes, process roles (main / preload / renderer, webview / core) on module nodes; checks for channels nobody receives and unregistered commands ([docs/bridges.md](docs/bridges.md#desktop-process-boundaries-electron-and-tauri)) |
 | AI harnesses | MCP servers (FastMCP / MCPServer / low-level) and clients, OpenAI / Anthropic tool schemas, Agents SDK, LangChain, LlamaIndex tools, hand-written agent loops (Python) | `endpoint:llm_tool:<name>` / `endpoint:mcp_tool:<server>/<name>` (resources, prompts) with the handler as an `llm_tool` entry, `agent:<name>` with OFFERS_TOOL / HANDS_OFF_TO; `cg tools` lists handlers, tables reached, agents, no_receiver / no_sender, dynamic dispatch and model calls ([docs/ai-tools.md](docs/ai-tools.md)) |
 | External systems | Laravel database connections, env keys read by code (`DB_*`, `DATABASE_URL`, `REDIS_*`, `SMTP_*`, `MONGO*`, `AMQP_*`, `LDAP_*`, `SFTP_*`, `S3_*` ...), Python settings (`DATABASES`, `CACHES`, `CELERY_BROKER_URL` ...), `.env.example` values, docker-compose services, DSNs | `external:<protocol>:<host:port>` (or `env:<KEY>`) with CONNECTS_TO from code and connections, CONFIGURED_BY, CREDENTIAL_FROM (location only, never the value), TLS; one node per system across linked repos; `cg external` ([docs/external.md](docs/external.md)) |
-| Protocol links | HTTP calls / routes, Pusher channels, NestJS messages, Bull / Laravel / Celery jobs, application events, bridges and IPC in one view; python-socketio / Flask-SocketIO events | `endpoint:<protocol>:<name>` with SENDS_TO / RECEIVED_BY and MATCHES_ENDPOINT (path, MQTT, NATS, AMQP topic, glob and template matchers in a registry plugins extend); `cg protocols` lists senders, receivers, guards and the checks no_receiver, no_sender, ambiguous, schema_mismatch, unguarded; `.cg.yaml` `protocols.external` for known outside parties ([docs/protocols.md](docs/protocols.md)) |
+| Protocol links | HTTP calls / routes, Pusher channels, NestJS messages, Bull / Laravel / Celery jobs, Celery / RQ / Dramatiq / Bull / Laravel / Symfony Messenger job and queue endpoints, application events, Kafka / RabbitMQ / Redis pub/sub and streams / MQTT / NATS producers and consumers, bridges and IPC in one view; python-socketio / Flask-SocketIO events | `endpoint:<protocol>:<name>` with SENDS_TO / RECEIVED_BY and MATCHES_ENDPOINT (path, MQTT, NATS, AMQP, glob and template matchers in a registry plugins extend); `cg protocols` lists senders, receivers, guards and the checks no_receiver, no_sender, ambiguous, schema_mismatch, unguarded, no_consumer; `.cg.yaml` `protocols.external` for known outside parties ([docs/protocols.md](docs/protocols.md)) |
 | Generated and copied files | detected (`.gitattributes`, generator banners, framework build paths, generator file names, Capacitor / Cordova copy targets, `.openapi-generator/FILES`) | kept out of the graph and listed by `cg coverage` by reason; copies map back to their source; `--include-generated` indexes them labelled `attrs.generated` ([docs/generated.md](docs/generated.md)) |
 
 ## Prerequisites per language
@@ -565,7 +565,8 @@ The scope as of v0.3, so you know how far each answer reaches. The full list is
   (`getattr`, DI containers, Riverpod/Provider lookups without a type) fall back to `heuristic` or stay unresolved.
   Python dispatch tables, plugin lists, callbacks and registering decorators are followed as function references
   ([docs/python.md](docs/python.md#entry-points-and-function-references)).
-  GraphQL APIs (graphene/strawberry) and Django template rendering are not modelled.
+  GraphQL root fields (graphene / strawberry / ariadne) are protocol endpoints ([docs/protocols.md](docs/protocols.md#graphql-root-fields));
+  object-type fields and Django template rendering are not modelled.
 - **Next to index:** seeders, `Artisan::command` closures, observers fired by model writes, Nuxt server routes, and
   navigation edges (`NuxtLink`, `navigateTo`).
 - **Broadcast channels** are read from `Broadcast::channel` and `broadcastOn()`; names cg cannot evaluate keep a
@@ -625,14 +626,14 @@ Ideas we are exploring after v0.3. Feedback on priorities is welcome.
 - Swift: `URLComponents` and helper-built URLs, `Info.plist` / `.xcconfig` base URLs, OS-version conditions, App Intents /
   widget entries, value navigation through variables ([docs/swift.md](docs/swift.md#not-covered-yet)).
 - Protocol links: extraction for MQTT, NATS, AMQP, Kafka and Redis pub/sub (matchers registered), Socket.IO outside
-  Python, raw WebSocket / SSE message names, gRPC / GraphQL / webhooks (epic #29; [docs/protocols.md](docs/protocols.md#not-covered-yet)).
+  Python, raw WebSocket / SSE message names, GraphQL object-type fields, webhooks (epic #29; [docs/protocols.md](docs/protocols.md#not-covered-yet)).
 - AI harnesses: TypeScript MCP servers / clients and the Vercel AI SDK, LangGraph graphs, agent runners, tools declared
   inside functions ([docs/ai-tools.md](docs/ai-tools.md#not-covered-yet)).
 - External systems: client constructors with literal arguments (`psycopg.connect(host=)`, `new Redis()`, ...),
   Spring / Rails / Kubernetes configuration, env reads through config schemas ([docs/external.md](docs/external.md#not-covered-yet)).
 - Web / native bridges beyond Capacitor, React Native, Flutter channels and Pigeon: React Native events, Capacitor
   `notifyListeners`, Cordova plugins and native UI components ([docs/bridges.md](docs/bridges.md#not-covered-yet)).
-- More HTTP clients beyond fetch, axios, ofetch and ky, and response-field modelling for the TypeScript client (setting → API response → client state); GraphQL APIs.
+- More HTTP clients beyond fetch, axios, ofetch and ky, and response-field modelling for the TypeScript client (setting → API response → client state).
 
 ## Documentation
 

{cg_code_graph-0.11.0 → cg_code_graph-0.12.0}/README.md

@@ -183,7 +183,7 @@ The 7 plan gaps are the admin update path, `UpdateBookRequest`, `Book::$fillable
 
 ### 1. A CLI for impact questions
 
-`reaches`, `impact`, `downstream`, `path`, `writers`, `siblings`, `routes`, `search`, `api-calls`, `channels`, `tests`
+`reaches`, `impact`, `downstream`, `path`, `writers`, `siblings`, `routes`, `search`, `snippet`, `api-calls`, `channels`, `tests`
 and `platforms` all work on one SQLite graph, and across repos once the frontend and backend are linked. Every hop shows its evidence:
 
 ```text
@@ -291,7 +291,7 @@ Full reference: [docs/cli.md](docs/cli.md) · value facts: [docs/value-facts.md]
 
 ### 2. An MCP server for AI agents
 
-The same queries as MCP tools (`reaches`, `impact`, `callers`, `siblings`, `path`, `downstream`, `routes`, `search`, `api_calls`,
+The same queries as MCP tools (`reaches`, `impact`, `callers`, `siblings`, `path`, `downstream`, `routes`, `search`, `snippet`, `api_calls`,
 `channels`, `tests_covering`, `resolutions`, `plan_check`, `index`, `coverage`, `starters`, `platform_divergence`, …), so an agent can check the blast radius before it edits. Replies are compact,
 use repo-relative paths, and `plan_check` starts with a summary (`details=true` for the full report). Every reply also
 carries a machine-readable `completeness` object, so the agent knows when an answer covers the whole repository and
@@ -362,7 +362,7 @@ without configuration; `cg config show` lists what was applied ([docs/configurat
 | Django REST Framework | resolved | routers, ViewSets (+ `@action`), APIView/generic views, `permission_classes`, serializer fields |
 | FastAPI / Starlette | resolved | `FastAPI` / `APIRouter` / `Starlette` / `Router` objects, `include_router` / `mount` prefix chains across files (prefixes from constants and settings attributes such as `settings.API_V1_STR`), verb decorators, `api_route`, `add_api_route`, Starlette `routes=[Route, Mount, WebSocketRoute]`, websockets, path parameters, `Depends()` / `Security()` dependencies (parameters, `Annotated` aliases, `dependencies=`) as route access with what each checks (raised 401 / 403, security schemes, nested dependencies), `name=` for `url_path_for`; apps / routers received as parameters or fixtures or returned by factories, Starlette `Host`, `@cbv` / `InferringRouter`, classy-fastapi `Routable` |
 | Flask | resolved | `Flask` / `Blueprint` objects, `register_blueprint` (`url_prefix` from the blueprint or the registration, nested blueprints), `@route` / verb decorators, `add_url_rule` incl. `MethodView.as_view()` (and its `methods`), endpoint-only rules, werkzeug `Rule` / `Submount`, the built-in static route, views defined in an app factory, apps received as parameters or pytest fixtures, flask-restful / flask-restx resources, `subdomain=` / `defaults=`, `<int:id>` parameters, `blueprint.endpoint` names for `url_for`, view decorators (`login_required`) as route access |
-| Celery / Channels | resolved | tasks + `.delay`/`.apply_async` dispatches; websocket routing to consumers |
+| Celery / RQ / Dramatiq / Channels | resolved | tasks + `.delay`/`.apply_async` dispatches; job and queue endpoints by task name across repos (`send_task`, RQ import paths, Dramatiq actors) with worker processes from Procfile / compose; websocket routing to consumers |
 | Dart | resolved (package:analyzer parse, declared types) + heuristic fallback | libraries/parts, classes, methods, functions, calls with import resolution |
 | Flutter | resolved + heuristic fallback | widgets/State, bloc/cubit events → handlers → states → UI, Navigator/go_router/auto_route pages, HTTP calls (package:http, Dio, dart:io, Retrofit/Chopper), WebSockets, json_serializable/freezed and hand-written JSON keys |
 | Rust | **exact** with rust-analyzer (SCIP); **heuristic** without | crates, modules, `pub` API, traits → impls (dyn/generic dispatch), bins, tests, benches, examples, `build.rs`, FFI, `unsafe`, `#[cfg(feature)]` gates, env keys, `#[tokio::main]`, axum/actix routes ([docs/native.md](docs/native.md)) |
@@ -377,7 +377,7 @@ without configuration; `cg config show` lists what was applied ([docs/configurat
 | Desktop processes | Electron `ipcMain` / `ipcRenderer` / `webContents.send` and `contextBridge.exposeInMainWorld`; Tauri `invoke` → `#[tauri::command]` | `endpoint:electron-ipc:<channel>`, `endpoint:electron-preload:<key>#<member>`, `endpoint:tauri:<command>` with SENDS_TO / RECEIVED_BY across the processes, process roles (main / preload / renderer, webview / core) on module nodes; checks for channels nobody receives and unregistered commands ([docs/bridges.md](docs/bridges.md#desktop-process-boundaries-electron-and-tauri)) |
 | AI harnesses | MCP servers (FastMCP / MCPServer / low-level) and clients, OpenAI / Anthropic tool schemas, Agents SDK, LangChain, LlamaIndex tools, hand-written agent loops (Python) | `endpoint:llm_tool:<name>` / `endpoint:mcp_tool:<server>/<name>` (resources, prompts) with the handler as an `llm_tool` entry, `agent:<name>` with OFFERS_TOOL / HANDS_OFF_TO; `cg tools` lists handlers, tables reached, agents, no_receiver / no_sender, dynamic dispatch and model calls ([docs/ai-tools.md](docs/ai-tools.md)) |
 | External systems | Laravel database connections, env keys read by code (`DB_*`, `DATABASE_URL`, `REDIS_*`, `SMTP_*`, `MONGO*`, `AMQP_*`, `LDAP_*`, `SFTP_*`, `S3_*` ...), Python settings (`DATABASES`, `CACHES`, `CELERY_BROKER_URL` ...), `.env.example` values, docker-compose services, DSNs | `external:<protocol>:<host:port>` (or `env:<KEY>`) with CONNECTS_TO from code and connections, CONFIGURED_BY, CREDENTIAL_FROM (location only, never the value), TLS; one node per system across linked repos; `cg external` ([docs/external.md](docs/external.md)) |
-| Protocol links | HTTP calls / routes, Pusher channels, NestJS messages, Bull / Laravel / Celery jobs, application events, bridges and IPC in one view; python-socketio / Flask-SocketIO events | `endpoint:<protocol>:<name>` with SENDS_TO / RECEIVED_BY and MATCHES_ENDPOINT (path, MQTT, NATS, AMQP topic, glob and template matchers in a registry plugins extend); `cg protocols` lists senders, receivers, guards and the checks no_receiver, no_sender, ambiguous, schema_mismatch, unguarded; `.cg.yaml` `protocols.external` for known outside parties ([docs/protocols.md](docs/protocols.md)) |
+| Protocol links | HTTP calls / routes, Pusher channels, NestJS messages, Bull / Laravel / Celery jobs, Celery / RQ / Dramatiq / Bull / Laravel / Symfony Messenger job and queue endpoints, application events, Kafka / RabbitMQ / Redis pub/sub and streams / MQTT / NATS producers and consumers, bridges and IPC in one view; python-socketio / Flask-SocketIO events | `endpoint:<protocol>:<name>` with SENDS_TO / RECEIVED_BY and MATCHES_ENDPOINT (path, MQTT, NATS, AMQP, glob and template matchers in a registry plugins extend); `cg protocols` lists senders, receivers, guards and the checks no_receiver, no_sender, ambiguous, schema_mismatch, unguarded, no_consumer; `.cg.yaml` `protocols.external` for known outside parties ([docs/protocols.md](docs/protocols.md)) |
 | Generated and copied files | detected (`.gitattributes`, generator banners, framework build paths, generator file names, Capacitor / Cordova copy targets, `.openapi-generator/FILES`) | kept out of the graph and listed by `cg coverage` by reason; copies map back to their source; `--include-generated` indexes them labelled `attrs.generated` ([docs/generated.md](docs/generated.md)) |
 
 ## Prerequisites per language
@@ -541,7 +541,8 @@ The scope as of v0.3, so you know how far each answer reaches. The full list is
   (`getattr`, DI containers, Riverpod/Provider lookups without a type) fall back to `heuristic` or stay unresolved.
   Python dispatch tables, plugin lists, callbacks and registering decorators are followed as function references
   ([docs/python.md](docs/python.md#entry-points-and-function-references)).
-  GraphQL APIs (graphene/strawberry) and Django template rendering are not modelled.
+  GraphQL root fields (graphene / strawberry / ariadne) are protocol endpoints ([docs/protocols.md](docs/protocols.md#graphql-root-fields));
+  object-type fields and Django template rendering are not modelled.
 - **Next to index:** seeders, `Artisan::command` closures, observers fired by model writes, Nuxt server routes, and
   navigation edges (`NuxtLink`, `navigateTo`).
 - **Broadcast channels** are read from `Broadcast::channel` and `broadcastOn()`; names cg cannot evaluate keep a
@@ -601,14 +602,14 @@ Ideas we are exploring after v0.3. Feedback on priorities is welcome.
 - Swift: `URLComponents` and helper-built URLs, `Info.plist` / `.xcconfig` base URLs, OS-version conditions, App Intents /
   widget entries, value navigation through variables ([docs/swift.md](docs/swift.md#not-covered-yet)).
 - Protocol links: extraction for MQTT, NATS, AMQP, Kafka and Redis pub/sub (matchers registered), Socket.IO outside
-  Python, raw WebSocket / SSE message names, gRPC / GraphQL / webhooks (epic #29; [docs/protocols.md](docs/protocols.md#not-covered-yet)).
+  Python, raw WebSocket / SSE message names, GraphQL object-type fields, webhooks (epic #29; [docs/protocols.md](docs/protocols.md#not-covered-yet)).
 - AI harnesses: TypeScript MCP servers / clients and the Vercel AI SDK, LangGraph graphs, agent runners, tools declared
   inside functions ([docs/ai-tools.md](docs/ai-tools.md#not-covered-yet)).
 - External systems: client constructors with literal arguments (`psycopg.connect(host=)`, `new Redis()`, ...),
   Spring / Rails / Kubernetes configuration, env reads through config schemas ([docs/external.md](docs/external.md#not-covered-yet)).
 - Web / native bridges beyond Capacitor, React Native, Flutter channels and Pigeon: React Native events, Capacitor
   `notifyListeners`, Cordova plugins and native UI components ([docs/bridges.md](docs/bridges.md#not-covered-yet)).
-- More HTTP clients beyond fetch, axios, ofetch and ky, and response-field modelling for the TypeScript client (setting → API response → client state); GraphQL APIs.
+- More HTTP clients beyond fetch, axios, ofetch and ky, and response-field modelling for the TypeScript client (setting → API response → client state).
 
 ## Documentation
 

{cg_code_graph-0.11.0 → cg_code_graph-0.12.0/cg_code_graph.egg-info}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: cg-code-graph
-Version: 0.11.0
+Version: 0.12.0
 Summary: cg: a code graph of routes, calls, data access and platform boundaries for agents and humans (CLI + MCP server)
 Author: cyberchronos00
 License-Expression: MIT
@@ -207,7 +207,7 @@ The 7 plan gaps are the admin update path, `UpdateBookRequest`, `Book::$fillable
 
 ### 1. A CLI for impact questions
 
-`reaches`, `impact`, `downstream`, `path`, `writers`, `siblings`, `routes`, `search`, `api-calls`, `channels`, `tests`
+`reaches`, `impact`, `downstream`, `path`, `writers`, `siblings`, `routes`, `search`, `snippet`, `api-calls`, `channels`, `tests`
 and `platforms` all work on one SQLite graph, and across repos once the frontend and backend are linked. Every hop shows its evidence:
 
 ```text
@@ -315,7 +315,7 @@ Full reference: [docs/cli.md](docs/cli.md) · value facts: [docs/value-facts.md]
 
 ### 2. An MCP server for AI agents
 
-The same queries as MCP tools (`reaches`, `impact`, `callers`, `siblings`, `path`, `downstream`, `routes`, `search`, `api_calls`,
+The same queries as MCP tools (`reaches`, `impact`, `callers`, `siblings`, `path`, `downstream`, `routes`, `search`, `snippet`, `api_calls`,
 `channels`, `tests_covering`, `resolutions`, `plan_check`, `index`, `coverage`, `starters`, `platform_divergence`, …), so an agent can check the blast radius before it edits. Replies are compact,
 use repo-relative paths, and `plan_check` starts with a summary (`details=true` for the full report). Every reply also
 carries a machine-readable `completeness` object, so the agent knows when an answer covers the whole repository and
@@ -386,7 +386,7 @@ without configuration; `cg config show` lists what was applied ([docs/configurat
 | Django REST Framework | resolved | routers, ViewSets (+ `@action`), APIView/generic views, `permission_classes`, serializer fields |
 | FastAPI / Starlette | resolved | `FastAPI` / `APIRouter` / `Starlette` / `Router` objects, `include_router` / `mount` prefix chains across files (prefixes from constants and settings attributes such as `settings.API_V1_STR`), verb decorators, `api_route`, `add_api_route`, Starlette `routes=[Route, Mount, WebSocketRoute]`, websockets, path parameters, `Depends()` / `Security()` dependencies (parameters, `Annotated` aliases, `dependencies=`) as route access with what each checks (raised 401 / 403, security schemes, nested dependencies), `name=` for `url_path_for`; apps / routers received as parameters or fixtures or returned by factories, Starlette `Host`, `@cbv` / `InferringRouter`, classy-fastapi `Routable` |
 | Flask | resolved | `Flask` / `Blueprint` objects, `register_blueprint` (`url_prefix` from the blueprint or the registration, nested blueprints), `@route` / verb decorators, `add_url_rule` incl. `MethodView.as_view()` (and its `methods`), endpoint-only rules, werkzeug `Rule` / `Submount`, the built-in static route, views defined in an app factory, apps received as parameters or pytest fixtures, flask-restful / flask-restx resources, `subdomain=` / `defaults=`, `<int:id>` parameters, `blueprint.endpoint` names for `url_for`, view decorators (`login_required`) as route access |
-| Celery / Channels | resolved | tasks + `.delay`/`.apply_async` dispatches; websocket routing to consumers |
+| Celery / RQ / Dramatiq / Channels | resolved | tasks + `.delay`/`.apply_async` dispatches; job and queue endpoints by task name across repos (`send_task`, RQ import paths, Dramatiq actors) with worker processes from Procfile / compose; websocket routing to consumers |
 | Dart | resolved (package:analyzer parse, declared types) + heuristic fallback | libraries/parts, classes, methods, functions, calls with import resolution |
 | Flutter | resolved + heuristic fallback | widgets/State, bloc/cubit events → handlers → states → UI, Navigator/go_router/auto_route pages, HTTP calls (package:http, Dio, dart:io, Retrofit/Chopper), WebSockets, json_serializable/freezed and hand-written JSON keys |
 | Rust | **exact** with rust-analyzer (SCIP); **heuristic** without | crates, modules, `pub` API, traits → impls (dyn/generic dispatch), bins, tests, benches, examples, `build.rs`, FFI, `unsafe`, `#[cfg(feature)]` gates, env keys, `#[tokio::main]`, axum/actix routes ([docs/native.md](docs/native.md)) |
@@ -401,7 +401,7 @@ without configuration; `cg config show` lists what was applied ([docs/configurat
 | Desktop processes | Electron `ipcMain` / `ipcRenderer` / `webContents.send` and `contextBridge.exposeInMainWorld`; Tauri `invoke` → `#[tauri::command]` | `endpoint:electron-ipc:<channel>`, `endpoint:electron-preload:<key>#<member>`, `endpoint:tauri:<command>` with SENDS_TO / RECEIVED_BY across the processes, process roles (main / preload / renderer, webview / core) on module nodes; checks for channels nobody receives and unregistered commands ([docs/bridges.md](docs/bridges.md#desktop-process-boundaries-electron-and-tauri)) |
 | AI harnesses | MCP servers (FastMCP / MCPServer / low-level) and clients, OpenAI / Anthropic tool schemas, Agents SDK, LangChain, LlamaIndex tools, hand-written agent loops (Python) | `endpoint:llm_tool:<name>` / `endpoint:mcp_tool:<server>/<name>` (resources, prompts) with the handler as an `llm_tool` entry, `agent:<name>` with OFFERS_TOOL / HANDS_OFF_TO; `cg tools` lists handlers, tables reached, agents, no_receiver / no_sender, dynamic dispatch and model calls ([docs/ai-tools.md](docs/ai-tools.md)) |
 | External systems | Laravel database connections, env keys read by code (`DB_*`, `DATABASE_URL`, `REDIS_*`, `SMTP_*`, `MONGO*`, `AMQP_*`, `LDAP_*`, `SFTP_*`, `S3_*` ...), Python settings (`DATABASES`, `CACHES`, `CELERY_BROKER_URL` ...), `.env.example` values, docker-compose services, DSNs | `external:<protocol>:<host:port>` (or `env:<KEY>`) with CONNECTS_TO from code and connections, CONFIGURED_BY, CREDENTIAL_FROM (location only, never the value), TLS; one node per system across linked repos; `cg external` ([docs/external.md](docs/external.md)) |
-| Protocol links | HTTP calls / routes, Pusher channels, NestJS messages, Bull / Laravel / Celery jobs, application events, bridges and IPC in one view; python-socketio / Flask-SocketIO events | `endpoint:<protocol>:<name>` with SENDS_TO / RECEIVED_BY and MATCHES_ENDPOINT (path, MQTT, NATS, AMQP topic, glob and template matchers in a registry plugins extend); `cg protocols` lists senders, receivers, guards and the checks no_receiver, no_sender, ambiguous, schema_mismatch, unguarded; `.cg.yaml` `protocols.external` for known outside parties ([docs/protocols.md](docs/protocols.md)) |
+| Protocol links | HTTP calls / routes, Pusher channels, NestJS messages, Bull / Laravel / Celery jobs, Celery / RQ / Dramatiq / Bull / Laravel / Symfony Messenger job and queue endpoints, application events, Kafka / RabbitMQ / Redis pub/sub and streams / MQTT / NATS producers and consumers, bridges and IPC in one view; python-socketio / Flask-SocketIO events | `endpoint:<protocol>:<name>` with SENDS_TO / RECEIVED_BY and MATCHES_ENDPOINT (path, MQTT, NATS, AMQP, glob and template matchers in a registry plugins extend); `cg protocols` lists senders, receivers, guards and the checks no_receiver, no_sender, ambiguous, schema_mismatch, unguarded, no_consumer; `.cg.yaml` `protocols.external` for known outside parties ([docs/protocols.md](docs/protocols.md)) |
 | Generated and copied files | detected (`.gitattributes`, generator banners, framework build paths, generator file names, Capacitor / Cordova copy targets, `.openapi-generator/FILES`) | kept out of the graph and listed by `cg coverage` by reason; copies map back to their source; `--include-generated` indexes them labelled `attrs.generated` ([docs/generated.md](docs/generated.md)) |
 
 ## Prerequisites per language
@@ -565,7 +565,8 @@ The scope as of v0.3, so you know how far each answer reaches. The full list is
   (`getattr`, DI containers, Riverpod/Provider lookups without a type) fall back to `heuristic` or stay unresolved.
   Python dispatch tables, plugin lists, callbacks and registering decorators are followed as function references
   ([docs/python.md](docs/python.md#entry-points-and-function-references)).
-  GraphQL APIs (graphene/strawberry) and Django template rendering are not modelled.
+  GraphQL root fields (graphene / strawberry / ariadne) are protocol endpoints ([docs/protocols.md](docs/protocols.md#graphql-root-fields));
+  object-type fields and Django template rendering are not modelled.
 - **Next to index:** seeders, `Artisan::command` closures, observers fired by model writes, Nuxt server routes, and
   navigation edges (`NuxtLink`, `navigateTo`).
 - **Broadcast channels** are read from `Broadcast::channel` and `broadcastOn()`; names cg cannot evaluate keep a
@@ -625,14 +626,14 @@ Ideas we are exploring after v0.3. Feedback on priorities is welcome.
 - Swift: `URLComponents` and helper-built URLs, `Info.plist` / `.xcconfig` base URLs, OS-version conditions, App Intents /
   widget entries, value navigation through variables ([docs/swift.md](docs/swift.md#not-covered-yet)).
 - Protocol links: extraction for MQTT, NATS, AMQP, Kafka and Redis pub/sub (matchers registered), Socket.IO outside
-  Python, raw WebSocket / SSE message names, gRPC / GraphQL / webhooks (epic #29; [docs/protocols.md](docs/protocols.md#not-covered-yet)).
+  Python, raw WebSocket / SSE message names, GraphQL object-type fields, webhooks (epic #29; [docs/protocols.md](docs/protocols.md#not-covered-yet)).
 - AI harnesses: TypeScript MCP servers / clients and the Vercel AI SDK, LangGraph graphs, agent runners, tools declared
   inside functions ([docs/ai-tools.md](docs/ai-tools.md#not-covered-yet)).
 - External systems: client constructors with literal arguments (`psycopg.connect(host=)`, `new Redis()`, ...),
   Spring / Rails / Kubernetes configuration, env reads through config schemas ([docs/external.md](docs/external.md#not-covered-yet)).
 - Web / native bridges beyond Capacitor, React Native, Flutter channels and Pigeon: React Native events, Capacitor
   `notifyListeners`, Cordova plugins and native UI components ([docs/bridges.md](docs/bridges.md#not-covered-yet)).
-- More HTTP clients beyond fetch, axios, ofetch and ky, and response-field modelling for the TypeScript client (setting → API response → client state); GraphQL APIs.
+- More HTTP clients beyond fetch, axios, ofetch and ky, and response-field modelling for the TypeScript client (setting → API response → client state).
 
 ## Documentation
 

{cg_code_graph-0.11.0 → cg_code_graph-0.12.0}/cg_code_graph.egg-info/SOURCES.txt

@@ -9,19 +9,29 @@ cg_code_graph.egg-info/entry_points.txt
 cg_code_graph.egg-info/requires.txt
 cg_code_graph.egg-info/top_level.txt
 codegraph/__init__.py
+codegraph/agent_rules.py
+codegraph/agents.py
 codegraph/aitools.py
+codegraph/android_ipc.py
+codegraph/apple_ipc.py
 codegraph/apps.py
 codegraph/blindspots.py
 codegraph/bridges.py
+codegraph/brokers.py
 codegraph/cli.py
 codegraph/concepts.py
 codegraph/config.py
 codegraph/coverage.py
+codegraph/dart_isolates.py
 codegraph/doctor.py
 codegraph/external.py
+codegraph/graphql.py
 codegraph/indexer.py
+codegraph/jobs.py
 codegraph/link.py
 codegraph/lint_async.py
+codegraph/local_ipc.py
+codegraph/local_sockets.py
 codegraph/mcp_server.py
 codegraph/parity.py
 codegraph/parity_structure.py
@@ -32,6 +42,9 @@ codegraph/platforms.py
 codegraph/process_runs.py
 codegraph/query.py
 codegraph/realtime.py
+codegraph/realtime_events.py
+codegraph/realtime_native.py
+codegraph/realtime_ws.py
 codegraph/roundtrip.py
 codegraph/routes.py
 codegraph/rpc.py
@@ -39,6 +52,7 @@ codegraph/sockets.py
 codegraph/starters.py
 codegraph/tests_index.py
 codegraph/udp_apps.py
+codegraph/webhooks.py
 codegraph/xcode.py
 codegraph/core/__init__.py
 codegraph/core/cache.py
@@ -96,6 +110,7 @@ codegraph/plugins/nextjs/plugin.py
 codegraph/plugins/nuxt/__init__.py
 codegraph/plugins/nuxt/plugin.py
 codegraph/plugins/php/__init__.py
+codegraph/plugins/php/external.py
 codegraph/plugins/php/gating.py
 codegraph/plugins/php/plugin.py
 codegraph/plugins/php/strings.py
@@ -134,6 +149,7 @@ codegraph/plugins/swift/packages.py
 codegraph/plugins/swift/plugin.py
 codegraph/plugins/ts/__init__.py
 codegraph/plugins/ts/baseurl.py
+codegraph/plugins/ts/nav.py
 codegraph/plugins/ts/plugin.py
 codegraph/plugins/ts/extractor/extract.mjs
 codegraph/plugins/ts/extractor/fw.mjs
@@ -179,9 +195,14 @@ codegraph/viz/static/vendor/layout-base.js
 codegraph/viz/tools/package-lock.json
 codegraph/viz/tools/package.json
 codegraph/viz/tools/shoot.mjs
+tests/test_agents_install.py
+tests/test_agents_md.py
 tests/test_ai_tools.py
+tests/test_android_ipc.py
+tests/test_apple_ipc.py
 tests/test_bridges.py
 tests/test_bridges_pigeon_monorepo.py
+tests/test_brokers.py
 tests/test_cache_clean.py
 tests/test_cache_invalidation.py
 tests/test_channels_tests.py
@@ -190,15 +211,19 @@ tests/test_completeness.py
 tests/test_config_extras.py
 tests/test_coverage.py
 tests/test_coverage_ts_hint.py
+tests/test_dart_isolates.py
 tests/test_dispatch_interfaces.py
 tests/test_django_flutter.py
 tests/test_enum_constants.py
 tests/test_external.py
 tests/test_gating.py
 tests/test_generated.py
+tests/test_graphql.py
 tests/test_impact_overrides.py
 tests/test_index_isolation.py
 tests/test_inherited_overrides.py
+tests/test_jobs.py
+tests/test_js_refs.py
 tests/test_kotlin.py
 tests/test_kotlin_binding.py
 tests/test_kotlin_exact.py
@@ -207,23 +232,29 @@ tests/test_kotlin_mixed.py
 tests/test_kotlin_nav_consts.py
 tests/test_kotlin_props.py
 tests/test_laravel_determinism.py
+tests/test_laravel_resources.py
 tests/test_lint_async.py
+tests/test_local_ipc.py
+tests/test_local_sockets.py
 tests/test_mcp.py
 tests/test_mcp_index.py
 tests/test_native_cfamily.py
 tests/test_native_concurrency.py
 tests/test_native_rust.py
+tests/test_nuxt_import_dirs.py
 tests/test_nuxt_rootsrc.py
 tests/test_packaging.py
 tests/test_parity.py
 tests/test_parity_structure.py
 tests/test_php_props.py
+tests/test_plain_js.py
 tests/test_plans.py
 tests/test_platform_gaps.py
 tests/test_platform_type_variants.py
 tests/test_platforms.py
 tests/test_platforms_apple.py
 tests/test_presets_config.py
+tests/test_process_ipc.py
 tests/test_process_runs.py
 tests/test_protocols.py
 tests/test_python_app_tests_module.py
@@ -238,10 +269,14 @@ tests/test_python_tests.py
 tests/test_python_web_hosts.py
 tests/test_python_web_routes.py
 tests/test_python_web_routes_more.py
+tests/test_realtime.py
+tests/test_realtime_native.py
 tests/test_roundtrip.py
+tests/test_route_tables.py
 tests/test_routes_search_explain.py
 tests/test_rpc.py
 tests/test_scip_health.py
+tests/test_snippet.py
 tests/test_sockets.py
 tests/test_starters_budget.py
 tests/test_swift.py
@@ -264,4 +299,7 @@ tests/test_ts_file_specs.py
 tests/test_ts_frameworks.py
 tests/test_ts_nuxt.py
 tests/test_values.py
-tests/test_viz.py
+tests/test_viz.py
+tests/test_vue_nav.py
+tests/test_webhooks.py
+tests/test_websockets.py

{cg_code_graph-0.11.0 → cg_code_graph-0.12.0}/codegraph/__init__.py

@@ -1,2 +1,2 @@
 """code-graph: deterministic code knowledge graph (beta)."""
-__version__ = "0.11.0"
+__version__ = "0.12.0"

cg_code_graph-0.12.0/codegraph/agent_rules.py

@@ -0,0 +1,57 @@
+"""Canonical "how to read this codebase with cg" guidance.
+
+Single source for this repo's AGENTS.md and for the blocks that `cg agents` installs into a
+project's AGENTS.md / CLAUDE.md / Cursor rules, so the wording never drifts. TEXT is the body;
+wrap it between BEGIN_MARK and END_MARK to get a block that can be replaced in place."""
+
+BEGIN_MARK = "<!-- BEGIN cg agent rules (managed by `cg agents`; edit cg, not here) -->"
+END_MARK = "<!-- END cg agent rules -->"
+
+TEXT = """\
+# Reading this codebase with cg
+
+This repository ships `cg`, a code-graph tool. Read the code by symbol through the index
+instead of opening whole files.
+
+## Build the index first
+    pip install -e . -q                     # make `cg` this checkout's version
+    cg index . --db out/graph.db --name repo
+There is no incremental index: re-run `cg index` after large changes.
+
+## Find code (get file:line; avoid ls / cat / broad grep)
+    cg search <name> --db out/graph.db        # nodes by name or FQN substring
+    cg impact <symbol> --db out/graph.db      # callers up to entry points (reverse)
+    cg downstream <symbol> --db out/graph.db  # what the symbol calls and reaches (forward)
+    cg node <symbol> --db out/graph.db        # one node's location, kind and edges
+`<symbol>` is a name, an FQN, `Class.method`, or a node id. Ambiguous names list candidates.
+
+## Read source (never cat a whole file; never more than ~200 lines at once)
+    cg snippet <symbol> --db out/graph.db                 # just that symbol's source, numbered
+    cg snippet <symbol> --db out/graph.db --context 5     # a few lines around it
+    sed -n '120,180p' path/to/file                      # a plain line range when needed
+`cg snippet` prints a `path:start-end` header then the body; `--max-lines` caps long bodies.
+
+## Follow an edge only when your change reaches it
+Use `cg impact` and `cg downstream` to decide whether a caller or callee matters before
+opening it. Chase a dependency only if your edit changes what crosses that edge.
+
+## After editing, re-query instead of rereading
+Re-run `cg index`, then `cg snippet` or `cg impact` on the changed spot to confirm the new
+shape, rather than reopening files.
+
+## Keep output small
+Pipe long output through `head` or `tail`:
+    cg impact <symbol> --db out/graph.db | head -n 40
+While iterating, run one file: `pytest -q -x tests/test_x.py`.
+Run the full suite once at the end: `pytest -q 2>&1 | tail -n 30`.
+"""
+
+
+def document() -> str:
+    """The repo-root AGENTS.md content (TEXT, newline-terminated)."""
+    return TEXT
+
+
+def block() -> str:
+    """The marked block `cg agents` writes into a project file; replaced in place on re-run."""
+    return BEGIN_MARK + "\n" + TEXT + END_MARK + "\n"

cg_code_graph-0.12.0/codegraph/agents.py

@@ -0,0 +1,150 @@
+"""`cg agents`: opt-in install of the cg usage guidance (and the cg MCP server entry) into a
+project's AGENTS.md / CLAUDE.md / Cursor rules. It only ever writes a clearly marked block and
+never writes silently: every run previews the exact changes and asks before writing (--dry-run
+previews without writing, --yes skips the prompt). Re-running replaces the block in place; remove
+deletes it and restores the surrounding bytes. MCP changes touch only the `cg` server entry."""
+from __future__ import annotations
+
+import difflib
+import json
+import re
+from pathlib import Path
+
+from . import agent_rules as AR
+
+# project guidance files `cg agents` can manage, by --target key
+TARGETS = {"agents": "AGENTS.md", "claude": "CLAUDE.md", "cursor": ".cursor/rules/cg.mdc"}
+DEFAULT_MCP_FILE = ".cursor/mcp.json"
+MCP_ENTRY = {"command": "cg-mcp", "args": ["--db", "out/graph.db"]}
+
+_CORE = re.compile(re.escape(AR.BEGIN_MARK) + r".*?" + re.escape(AR.END_MARK), re.S)
+
+
+def _core_text() -> str:
+    """The managed block without a trailing newline (BEGIN … END)."""
+    return AR.block().rstrip("\n")
+
+
+def plan_text(path: Path, remove: bool) -> dict:
+    """Decide the new content of one guidance file. action: insert | update | remove | noop.
+    Everything outside the marked block is preserved; a re-run replaces the block in place and
+    remove restores the bytes the matching insert added (for newline-terminated files)."""
+    old = path.read_text(encoding="utf-8") if path.exists() else ""
+    m = _CORE.search(old)
+    core = _core_text()
+    if remove:
+        if not m:
+            return {"path": str(path), "kind": "text", "action": "noop", "old": old, "new": old}
+        s, e = m.start(), m.end()
+        while e < len(old) and old[e] == "\n":          # drop the block's trailing newlines
+            e += 1
+        j = s
+        while j > 0 and old[j - 1] == "\n":             # drop separator newlines before it
+            j -= 1
+        s = j + 1 if j > 0 else 0                        # but keep one newline after real content
+        new = old[:s] + old[e:]
+        return {"path": str(path), "kind": "text", "action": "remove", "old": old, "new": new}
+    if m:
+        new = old[:m.start()] + core + old[m.end():]
+        return {"path": str(path), "kind": "text", "action": "update", "old": old, "new": new}
+    if not old:
+        new = core + "\n"
+    elif old.endswith("\n\n"):
+        new = old + core + "\n"
+    elif old.endswith("\n"):
+        new = old + "\n" + core + "\n"
+    else:
+        new = old + "\n\n" + core + "\n"
+    return {"path": str(path), "kind": "text", "action": "insert", "old": old, "new": new}
+
+
+def plan_mcp(path: Path, remove: bool) -> dict:
+    """Add / update / remove only the `cg` entry under mcpServers; every other entry and key is
+    kept. Returns action noop when there is nothing to change, or error when the file is not JSON."""
+    old = path.read_text(encoding="utf-8") if path.exists() else ""
+    if old.strip():
+        try:
+            data = json.loads(old)
+        except ValueError as ex:
+            return {"path": str(path), "kind": "mcp", "action": "error", "old": old, "new": old,
+                    "error": f"not valid JSON ({ex}); left unchanged"}
+    else:
+        data = {}
+    if not isinstance(data, dict):
+        return {"path": str(path), "kind": "mcp", "action": "error", "old": old, "new": old,
+                "error": "top level is not a JSON object; left unchanged"}
+    servers = data.get("mcpServers")
+    if not isinstance(servers, dict):
+        servers = {}
+    if remove:
+        if "cg" not in servers:
+            return {"path": str(path), "kind": "mcp", "action": "noop", "old": old, "new": old}
+        servers = {k: v for k, v in servers.items() if k != "cg"}
+        action = "remove"
+    else:
+        action = "update" if "cg" in servers else "insert"
+        servers = {**servers, "cg": MCP_ENTRY}
+    data = {**data, "mcpServers": servers}
+    new = json.dumps(data, indent=2) + "\n"
+    return {"path": str(path), "kind": "mcp", "action": action, "old": old, "new": new}
+
+
+def preview(plan: dict) -> str:
+    """A unified diff of one plan (the exact proposed change), or a one-line note for noop/error."""
+    if plan["action"] == "noop":
+        return f"{plan['path']}: no change (block not present)" if plan["kind"] == "text" \
+            else f"{plan['path']}: no change (no cg entry)"
+    if plan["action"] == "error":
+        return f"{plan['path']}: {plan['error']}"
+    diff = difflib.unified_diff(plan["old"].splitlines(True), plan["new"].splitlines(True),
+                                fromfile=plan["path"], tofile=plan["path"], n=1)
+    body = "".join(diff).rstrip("\n")
+    return f"{plan['path']}  [{plan['action']}]\n{body}"
+
+
+def build_plans(root: Path, targets: list[str], mcp: bool, mcp_file: str | None, remove: bool) -> list[dict]:
+    plans = [plan_text(root / TARGETS[t], remove) for t in targets]
+    if mcp:
+        plans.append(plan_mcp(Path(mcp_file) if mcp_file else root / DEFAULT_MCP_FILE, remove))
+    return plans
+
+
+def run(action: str, root: str = ".", targets: list[str] | None = None, all_targets: bool = False,
+        mcp: bool = False, mcp_file: str | None = None, dry_run: bool = False, assume_yes: bool = False,
+        input_fn=input, out=print) -> int:
+    """Entry point for the CLI. action: install | update | remove | show. Returns a process exit code."""
+    rootp = Path(root)
+    remove = action == "remove"
+    if all_targets:
+        sel = list(TARGETS)
+    elif targets:
+        sel = targets
+    else:                                   # default: the guidance files that already exist
+        sel = [t for t, f in TARGETS.items() if (rootp / f).exists()]
+    if not sel and not mcp:
+        out("nothing selected: pass --target agents|claude|cursor (repeatable), --all, or --mcp "
+            "(none of AGENTS.md / CLAUDE.md / .cursor/rules/cg.mdc exist here)")
+        return 2
+    plans = build_plans(rootp, sel, mcp, mcp_file, remove)
+    for p in plans:
+        out(preview(p))
+    if any(p["action"] == "error" for p in plans):
+        return 2
+    changing = [p for p in plans if p["action"] not in ("noop",)]
+    if action == "show" or dry_run:
+        out("(dry run: nothing written)" if dry_run else "(show: nothing written)")
+        return 0
+    if not changing:
+        out("nothing to do")
+        return 0
+    if not assume_yes:
+        resp = input_fn(f"apply {len(changing)} change(s)? [y/N] ")
+        if (resp or "").strip().lower() not in ("y", "yes"):
+            out("aborted; nothing written")
+            return 1
+    for p in changing:
+        fp = Path(p["path"])
+        fp.parent.mkdir(parents=True, exist_ok=True)
+        fp.write_text(p["new"], encoding="utf-8")
+        out(f"wrote {p['path']} ({p['action']})")
+    return 0