codewiki 0.1.0__tar.gz

codewiki-0.1.0/PKG-INFO

@@ -0,0 +1,289 @@
+Metadata-Version: 2.4
+Name: codewiki
+Version: 0.1.0
+Summary: Local-first code wiki platform with AST graph analysis, GraphRAG, and LiteLLM.
+Author-email: PorunC <09982.misaka@gmail.com>
+Project-URL: Homepage, https://github.com/PorunC/CodeWiki
+Project-URL: Repository, https://github.com/PorunC/CodeWiki
+Project-URL: Issues, https://github.com/PorunC/CodeWiki/issues
+Keywords: code-wiki,graphrag,fastapi,llm,documentation
+Classifier: Development Status :: 3 - Alpha
+Classifier: Environment :: Console
+Classifier: Framework :: FastAPI
+Classifier: Intended Audience :: Developers
+Classifier: Operating System :: OS Independent
+Classifier: Programming Language :: Python :: 3
+Classifier: Programming Language :: Python :: 3.12
+Classifier: Topic :: Documentation
+Classifier: Topic :: Software Development :: Documentation
+Classifier: Topic :: Software Development :: Libraries :: Application Frameworks
+Requires-Python: >=3.12
+Description-Content-Type: text/markdown
+Requires-Dist: aiosqlite>=0.20.0
+Requires-Dist: click>=8.1.0
+Requires-Dist: fastapi>=0.115.0
+Requires-Dist: graspologic>=3.4.4; python_version < "3.13"
+Requires-Dist: httpx[socks]>=0.28.1
+Requires-Dist: litellm>=1.60.0
+Requires-Dist: mermaid-parser-py>=0.0.4
+Requires-Dist: networkx>=3.4.0
+Requires-Dist: pathspec>=0.12.1
+Requires-Dist: pydantic>=2.10.0
+Requires-Dist: pydantic-settings>=2.7.0
+Requires-Dist: python-dotenv>=1.0.1
+Requires-Dist: sqlite-vec>=0.1.9
+Requires-Dist: sqlalchemy>=2.0.35
+Requires-Dist: tree-sitter>=0.23.0
+Requires-Dist: tree-sitter-c>=0.24.2
+Requires-Dist: tree-sitter-c-sharp>=0.23.5
+Requires-Dist: tree-sitter-cpp>=0.23.4
+Requires-Dist: tree-sitter-go>=0.25.0
+Requires-Dist: tree-sitter-java>=0.23.5
+Requires-Dist: tree-sitter-javascript>=0.25.0
+Requires-Dist: tree-sitter-python>=0.25.0
+Requires-Dist: tree-sitter-rust>=0.24.2
+Requires-Dist: tree-sitter-typescript>=0.23.2
+Requires-Dist: uvicorn[standard]>=0.34.0
+Provides-Extra: dev
+Requires-Dist: pytest>=8.3.0; extra == "dev"
+Requires-Dist: pytest-asyncio>=0.25.0; extra == "dev"
+Requires-Dist: ruff>=0.8.0; extra == "dev"
+
+# CodeWiki
+
+Single-user CodeWiki platform for AST-based code graph analysis, GraphRAG retrieval,
+source-grounded wiki generation, and LiteLLM-powered Q&A.
+
+## Current Scope
+
+- FastAPI backend with repository management, analysis runs, GraphRAG, wiki, ask, graph,
+  file, run, and settings APIs.
+- React/Vite frontend with repository management plus graph explorer, wiki reader,
+  ask, and settings pages.
+- AST-backed code graph extraction for Python, TypeScript/TSX, JavaScript/JSX, Java,
+  Go, Rust, C, C++, and C#.
+- Deterministic graph edges for imports, exports, definitions, inheritance,
+  implementations, calls, route handlers, source references, and configuration usage.
+- GraphRAG retrieval with source chunks, optional embeddings, community summaries,
+  and cached LLM runs.
+- DeepWiki-style wiki generation with catalog planning, detailed page generation,
+  source citations, automatic diagrams, multi-language translation, and incremental
+  updates.
+- Pure frontend wiki exports: interactive standalone HTML and Obsidian vault ZIP.
+- Design notes live in `docs/design.md`.
+
+## Installation
+
+Install the Python package from PyPI:
+
+```bash
+pip install codewiki
+codewiki --help
+```
+
+Start the backend API after installation:
+
+```bash
+codewiki serve
+```
+
+The source checkout is still the recommended way to run the full backend + Vite
+frontend development stack.
+
+## Wiki Workflow
+
+1. Register and analyze a repository.
+2. Build GraphRAG source chunks, optionally with embeddings.
+3. Generate a wiki catalog.
+4. Generate wiki pages from the catalog.
+5. Use update/regenerate flows when code changes.
+
+Wiki pages are generated from deterministic graph facts and retrieved source chunks.
+The page prompt enforces a gather/think/write workflow and includes ReadFile evidence so
+the model must stay close to real source files. Source references are validated before a
+page is promoted to `generated`; otherwise the page is saved as `draft` with validation
+errors.
+
+Mermaid diagrams are generated server-side from validated graph facts. Invalid diagrams
+are filtered out instead of failing the whole page, so a bad graph block should not turn
+a good wiki page into a draft.
+
+## Wiki Languages
+
+The base wiki language is generated first. Other languages are produced by translating
+the base catalog and pages while preserving slugs, source references, code identifiers,
+links, and Markdown structure.
+
+Set configured translation languages in `.env`:
+
+```bash
+CODEWIKI_WIKI_BASE_LANGUAGE=en
+CODEWIKI_WIKI_TRANSLATION_LANGUAGES=zh
+```
+
+The frontend wiki page has an English/Chinese language switch above the left catalog
+navigation. If a requested non-base language is missing, the backend generates the base
+wiki first and then translates it.
+
+## Wiki Export
+
+The frontend wiki toolbar can export the currently selected language as:
+
+- Interactive HTML: a standalone static page with catalog navigation, page switching,
+  rendered Markdown, source sections, related pages, and Mermaid rendering.
+- Obsidian vault: a ZIP containing Markdown pages, wiki links, source metadata, and
+  minimal `.obsidian` settings.
+
+Exports are built entirely in the browser from already-loaded wiki data and do not
+require a backend export API.
+
+## LLM Configuration
+
+Copy `.env.example` and fill in a default model profile:
+
+```bash
+cp .env.example .env
+```
+
+The default profile is used for every task unless a task-specific profile overrides it.
+This is the simplest "use one model for everything" setup:
+
+```bash
+CODEWIKI_LLM__MODE=sdk
+CODEWIKI_LLM__DEFAULT__MODEL=provider/strong-coding-model
+CODEWIKI_LLM__DEFAULT__PROVIDER_TYPE=
+CODEWIKI_LLM__DEFAULT__ENDPOINT=
+CODEWIKI_LLM__DEFAULT__API_KEY=
+CODEWIKI_LLM__TIMEOUT_SECONDS=120
+CODEWIKI_LLM__MAX_RETRIES=3
+CODEWIKI_LLM__CACHE_ENABLED=true
+```
+
+Each LLM task can override model, provider type, endpoint, and API key:
+
+```bash
+# Fast/cheap catalog planning
+CODEWIKI_LLM__PROFILES__CATALOG__MODEL=
+CODEWIKI_LLM__PROFILES__CATALOG__PROVIDER_TYPE=
+CODEWIKI_LLM__PROFILES__CATALOG__ENDPOINT=
+CODEWIKI_LLM__PROFILES__CATALOG__API_KEY=
+
+# Strong source-grounded wiki page generation
+CODEWIKI_LLM__PROFILES__PAGE__MODEL=
+CODEWIKI_LLM__PROFILES__PAGE__PROVIDER_TYPE=
+CODEWIKI_LLM__PROFILES__PAGE__ENDPOINT=
+CODEWIKI_LLM__PROFILES__PAGE__API_KEY=
+
+# Translation
+CODEWIKI_LLM__PROFILES__TRANSLATION__MODEL=
+CODEWIKI_LLM__PROFILES__TRANSLATION__PROVIDER_TYPE=
+CODEWIKI_LLM__PROFILES__TRANSLATION__ENDPOINT=
+CODEWIKI_LLM__PROFILES__TRANSLATION__API_KEY=
+
+# Ask / QA
+CODEWIKI_LLM__PROFILES__QA__MODEL=
+CODEWIKI_LLM__PROFILES__QA__PROVIDER_TYPE=
+CODEWIKI_LLM__PROFILES__QA__ENDPOINT=
+CODEWIKI_LLM__PROFILES__QA__API_KEY=
+
+# Embeddings, used when GraphRAG vector indexing is enabled
+CODEWIKI_LLM__PROFILES__EMBEDDING__MODEL=
+CODEWIKI_LLM__PROFILES__EMBEDDING__PROVIDER_TYPE=
+CODEWIKI_LLM__PROFILES__EMBEDDING__ENDPOINT=
+CODEWIKI_LLM__PROFILES__EMBEDDING__API_KEY=
+```
+
+Provider examples depend on LiteLLM. For OpenAI-compatible endpoints, set an endpoint
+and API key. For native LiteLLM providers, set `PROVIDER_TYPE` and model according to
+LiteLLM's provider naming.
+
+Failed LLM provider calls are recorded in `llm_run` with `status=error`; API responses
+return a `run_id` where possible so failures can be traced without exposing API keys.
+
+## Development
+
+```bash
+# Install backend and frontend dependencies
+make install
+
+# Start FastAPI and Vite
+make start
+
+# Stop local dev servers on the configured ports
+make kill
+```
+
+Default local URLs:
+
+- Backend: `http://127.0.0.1:8000`
+- Frontend: `http://127.0.0.1:5173`
+
+Useful checks:
+
+```bash
+make lint
+make test
+make build
+```
+
+## CLI
+
+```bash
+# Register or inspect repositories
+codewiki repos add . --name my-repo
+codewiki repos list
+codewiki repos scan .
+
+# Full analysis and GraphRAG
+codewiki analyze .
+codewiki graphrag build .
+codewiki graphrag build . --embeddings
+
+# Wiki generation
+codewiki wiki catalog .
+codewiki wiki pages .
+codewiki wiki update . --language en
+codewiki wiki page overview .
+
+# Incremental graph update, with wiki regeneration enabled by default
+codewiki update .
+
+# GraphRAG grounded Q&A
+codewiki ask "How does the main workflow fit together?"
+codewiki ask --repo my-repo "Where are wiki pages generated?"
+```
+
+Most commands accept a repository id, id prefix, registered name, path, or Git URL.
+Use `--json` on CLI commands when machine-readable output is useful.
+
+## HTTP API Highlights
+
+| Method | Path | Purpose |
+|---|---|---|
+| `POST` | `/api/repos/{repo_id}/wiki/catalog?language=en` | Generate a wiki catalog |
+| `POST` | `/api/repos/{repo_id}/wiki/pages/generate?language=en` | Generate all wiki pages |
+| `POST` | `/api/repos/{repo_id}/wiki/pages/update?language=en` | Incrementally update stale/missing pages |
+| `POST` | `/api/repos/{repo_id}/wiki/pages/{slug}/regenerate?language=en` | Regenerate one page |
+| `POST` | `/api/repos/{repo_id}/wiki/translate` | Translate catalog and pages |
+| `GET` | `/api/repos/{repo_id}/wiki?language=en` | Read the wiki catalog and pages |
+| `POST` | `/api/repos/{repo_id}/ask` | Ask a GraphRAG-grounded question |
+
+## Supported AST Languages
+
+| Language | Parser | Extracted facts |
+|---|---|---|
+| Python | tree-sitter capture parser | imports, classes, functions, methods, decorators, calls, references, FastAPI-style endpoints |
+| TypeScript / TSX | tree-sitter capture parser | imports/exports, classes, interfaces, type aliases, functions, methods, calls, route endpoints |
+| JavaScript / JSX | tree-sitter capture parser | imports/exports, classes, functions, methods, calls, route endpoints |
+| Java | tree-sitter capture parser | package/imports, classes, interfaces, records, enums, methods, constructors, inheritance, implementations, Spring-style endpoints |
+| Go | tree-sitter capture parser | package/imports, structs, interfaces, type aliases, functions, receiver methods, calls, router-style endpoints |
+| Rust | tree-sitter capture parser | imports, structs, enums, traits, impls, functions, methods, calls |
+| C | tree-sitter capture parser | includes, structs, functions, calls |
+| C++ | tree-sitter capture parser | includes, classes, structs, functions, methods, inheritance, calls |
+| C# | tree-sitter capture parser | usings, namespaces, classes, interfaces, methods, inheritance, calls |
+
+## Notes
+
+The core contract is that code facts come from deterministic scanners and AST parsers
+first. GraphRAG and LLM workflows consume those facts for retrieval, synthesis, and wiki
+generation rather than inventing structure.

codewiki-0.1.0/README.md

@@ -0,0 +1,238 @@
+# CodeWiki
+
+Single-user CodeWiki platform for AST-based code graph analysis, GraphRAG retrieval,
+source-grounded wiki generation, and LiteLLM-powered Q&A.
+
+## Current Scope
+
+- FastAPI backend with repository management, analysis runs, GraphRAG, wiki, ask, graph,
+  file, run, and settings APIs.
+- React/Vite frontend with repository management plus graph explorer, wiki reader,
+  ask, and settings pages.
+- AST-backed code graph extraction for Python, TypeScript/TSX, JavaScript/JSX, Java,
+  Go, Rust, C, C++, and C#.
+- Deterministic graph edges for imports, exports, definitions, inheritance,
+  implementations, calls, route handlers, source references, and configuration usage.
+- GraphRAG retrieval with source chunks, optional embeddings, community summaries,
+  and cached LLM runs.
+- DeepWiki-style wiki generation with catalog planning, detailed page generation,
+  source citations, automatic diagrams, multi-language translation, and incremental
+  updates.
+- Pure frontend wiki exports: interactive standalone HTML and Obsidian vault ZIP.
+- Design notes live in `docs/design.md`.
+
+## Installation
+
+Install the Python package from PyPI:
+
+```bash
+pip install codewiki
+codewiki --help
+```
+
+Start the backend API after installation:
+
+```bash
+codewiki serve
+```
+
+The source checkout is still the recommended way to run the full backend + Vite
+frontend development stack.
+
+## Wiki Workflow
+
+1. Register and analyze a repository.
+2. Build GraphRAG source chunks, optionally with embeddings.
+3. Generate a wiki catalog.
+4. Generate wiki pages from the catalog.
+5. Use update/regenerate flows when code changes.
+
+Wiki pages are generated from deterministic graph facts and retrieved source chunks.
+The page prompt enforces a gather/think/write workflow and includes ReadFile evidence so
+the model must stay close to real source files. Source references are validated before a
+page is promoted to `generated`; otherwise the page is saved as `draft` with validation
+errors.
+
+Mermaid diagrams are generated server-side from validated graph facts. Invalid diagrams
+are filtered out instead of failing the whole page, so a bad graph block should not turn
+a good wiki page into a draft.
+
+## Wiki Languages
+
+The base wiki language is generated first. Other languages are produced by translating
+the base catalog and pages while preserving slugs, source references, code identifiers,
+links, and Markdown structure.
+
+Set configured translation languages in `.env`:
+
+```bash
+CODEWIKI_WIKI_BASE_LANGUAGE=en
+CODEWIKI_WIKI_TRANSLATION_LANGUAGES=zh
+```
+
+The frontend wiki page has an English/Chinese language switch above the left catalog
+navigation. If a requested non-base language is missing, the backend generates the base
+wiki first and then translates it.
+
+## Wiki Export
+
+The frontend wiki toolbar can export the currently selected language as:
+
+- Interactive HTML: a standalone static page with catalog navigation, page switching,
+  rendered Markdown, source sections, related pages, and Mermaid rendering.
+- Obsidian vault: a ZIP containing Markdown pages, wiki links, source metadata, and
+  minimal `.obsidian` settings.
+
+Exports are built entirely in the browser from already-loaded wiki data and do not
+require a backend export API.
+
+## LLM Configuration
+
+Copy `.env.example` and fill in a default model profile:
+
+```bash
+cp .env.example .env
+```
+
+The default profile is used for every task unless a task-specific profile overrides it.
+This is the simplest "use one model for everything" setup:
+
+```bash
+CODEWIKI_LLM__MODE=sdk
+CODEWIKI_LLM__DEFAULT__MODEL=provider/strong-coding-model
+CODEWIKI_LLM__DEFAULT__PROVIDER_TYPE=
+CODEWIKI_LLM__DEFAULT__ENDPOINT=
+CODEWIKI_LLM__DEFAULT__API_KEY=
+CODEWIKI_LLM__TIMEOUT_SECONDS=120
+CODEWIKI_LLM__MAX_RETRIES=3
+CODEWIKI_LLM__CACHE_ENABLED=true
+```
+
+Each LLM task can override model, provider type, endpoint, and API key:
+
+```bash
+# Fast/cheap catalog planning
+CODEWIKI_LLM__PROFILES__CATALOG__MODEL=
+CODEWIKI_LLM__PROFILES__CATALOG__PROVIDER_TYPE=
+CODEWIKI_LLM__PROFILES__CATALOG__ENDPOINT=
+CODEWIKI_LLM__PROFILES__CATALOG__API_KEY=
+
+# Strong source-grounded wiki page generation
+CODEWIKI_LLM__PROFILES__PAGE__MODEL=
+CODEWIKI_LLM__PROFILES__PAGE__PROVIDER_TYPE=
+CODEWIKI_LLM__PROFILES__PAGE__ENDPOINT=
+CODEWIKI_LLM__PROFILES__PAGE__API_KEY=
+
+# Translation
+CODEWIKI_LLM__PROFILES__TRANSLATION__MODEL=
+CODEWIKI_LLM__PROFILES__TRANSLATION__PROVIDER_TYPE=
+CODEWIKI_LLM__PROFILES__TRANSLATION__ENDPOINT=
+CODEWIKI_LLM__PROFILES__TRANSLATION__API_KEY=
+
+# Ask / QA
+CODEWIKI_LLM__PROFILES__QA__MODEL=
+CODEWIKI_LLM__PROFILES__QA__PROVIDER_TYPE=
+CODEWIKI_LLM__PROFILES__QA__ENDPOINT=
+CODEWIKI_LLM__PROFILES__QA__API_KEY=
+
+# Embeddings, used when GraphRAG vector indexing is enabled
+CODEWIKI_LLM__PROFILES__EMBEDDING__MODEL=
+CODEWIKI_LLM__PROFILES__EMBEDDING__PROVIDER_TYPE=
+CODEWIKI_LLM__PROFILES__EMBEDDING__ENDPOINT=
+CODEWIKI_LLM__PROFILES__EMBEDDING__API_KEY=
+```
+
+Provider examples depend on LiteLLM. For OpenAI-compatible endpoints, set an endpoint
+and API key. For native LiteLLM providers, set `PROVIDER_TYPE` and model according to
+LiteLLM's provider naming.
+
+Failed LLM provider calls are recorded in `llm_run` with `status=error`; API responses
+return a `run_id` where possible so failures can be traced without exposing API keys.
+
+## Development
+
+```bash
+# Install backend and frontend dependencies
+make install
+
+# Start FastAPI and Vite
+make start
+
+# Stop local dev servers on the configured ports
+make kill
+```
+
+Default local URLs:
+
+- Backend: `http://127.0.0.1:8000`
+- Frontend: `http://127.0.0.1:5173`
+
+Useful checks:
+
+```bash
+make lint
+make test
+make build
+```
+
+## CLI
+
+```bash
+# Register or inspect repositories
+codewiki repos add . --name my-repo
+codewiki repos list
+codewiki repos scan .
+
+# Full analysis and GraphRAG
+codewiki analyze .
+codewiki graphrag build .
+codewiki graphrag build . --embeddings
+
+# Wiki generation
+codewiki wiki catalog .
+codewiki wiki pages .
+codewiki wiki update . --language en
+codewiki wiki page overview .
+
+# Incremental graph update, with wiki regeneration enabled by default
+codewiki update .
+
+# GraphRAG grounded Q&A
+codewiki ask "How does the main workflow fit together?"
+codewiki ask --repo my-repo "Where are wiki pages generated?"
+```
+
+Most commands accept a repository id, id prefix, registered name, path, or Git URL.
+Use `--json` on CLI commands when machine-readable output is useful.
+
+## HTTP API Highlights
+
+| Method | Path | Purpose |
+|---|---|---|
+| `POST` | `/api/repos/{repo_id}/wiki/catalog?language=en` | Generate a wiki catalog |
+| `POST` | `/api/repos/{repo_id}/wiki/pages/generate?language=en` | Generate all wiki pages |
+| `POST` | `/api/repos/{repo_id}/wiki/pages/update?language=en` | Incrementally update stale/missing pages |
+| `POST` | `/api/repos/{repo_id}/wiki/pages/{slug}/regenerate?language=en` | Regenerate one page |
+| `POST` | `/api/repos/{repo_id}/wiki/translate` | Translate catalog and pages |
+| `GET` | `/api/repos/{repo_id}/wiki?language=en` | Read the wiki catalog and pages |
+| `POST` | `/api/repos/{repo_id}/ask` | Ask a GraphRAG-grounded question |
+
+## Supported AST Languages
+
+| Language | Parser | Extracted facts |
+|---|---|---|
+| Python | tree-sitter capture parser | imports, classes, functions, methods, decorators, calls, references, FastAPI-style endpoints |
+| TypeScript / TSX | tree-sitter capture parser | imports/exports, classes, interfaces, type aliases, functions, methods, calls, route endpoints |
+| JavaScript / JSX | tree-sitter capture parser | imports/exports, classes, functions, methods, calls, route endpoints |
+| Java | tree-sitter capture parser | package/imports, classes, interfaces, records, enums, methods, constructors, inheritance, implementations, Spring-style endpoints |
+| Go | tree-sitter capture parser | package/imports, structs, interfaces, type aliases, functions, receiver methods, calls, router-style endpoints |
+| Rust | tree-sitter capture parser | imports, structs, enums, traits, impls, functions, methods, calls |
+| C | tree-sitter capture parser | includes, structs, functions, calls |
+| C++ | tree-sitter capture parser | includes, classes, structs, functions, methods, inheritance, calls |
+| C# | tree-sitter capture parser | usings, namespaces, classes, interfaces, methods, inheritance, calls |
+
+## Notes
+
+The core contract is that code facts come from deterministic scanners and AST parsers
+first. GraphRAG and LLM workflows consume those facts for retrieval, synthesis, and wiki
+generation rather than inventing structure.

codewiki-0.1.0/backend/__init__.py

@@ -0,0 +1,2 @@
+"""Backend package for Code Wiki Platform."""
+

codewiki-0.1.0/backend/app/__init__.py

@@ -0,0 +1,2 @@
+"""FastAPI application package."""
+

codewiki-0.1.0/backend/app/api/__init__.py

@@ -0,0 +1,2 @@
+"""HTTP API routers."""
+

codewiki-0.1.0/backend/app/api/ask.py

@@ -0,0 +1,37 @@
+from fastapi import APIRouter, HTTPException
+
+from backend.app.config import get_settings
+from backend.app.database import get_store
+from backend.app.schemas.ask import AskRequest, AskResponse
+from backend.app.services.graphrag import GraphRAGRetriever
+from backend.app.services.llm_gateway import LLMGateway
+from backend.app.services.llm_run_recorder import LLMCallError
+from backend.app.services.question_answerer import QuestionAnswerer
+
+router = APIRouter()
+
+
+@router.post("/{repo_id}/ask")
+async def ask_repo(repo_id: str, payload: AskRequest) -> AskResponse:
+    settings = get_settings()
+    store = get_store()
+    answerer = QuestionAnswerer(
+        GraphRAGRetriever(store=store, settings=settings),
+        LLMGateway(settings),
+        store=store,
+    )
+    try:
+        return await answerer.answer(repo_id, payload)
+    except LLMCallError as exc:
+        raise HTTPException(
+            status_code=502,
+            detail={
+                "message": str(exc),
+                "task_type": exc.task_type,
+                "run_id": exc.run_id,
+            },
+        ) from exc
+    except ValueError as exc:
+        message = str(exc)
+        status_code = 404 if message.startswith("Repository not found") else 400
+        raise HTTPException(status_code=status_code, detail=message) from exc

codewiki-0.1.0/backend/app/api/dependencies.py

@@ -0,0 +1,63 @@
+from functools import cached_property, lru_cache
+from typing import Annotated
+
+from fastapi import Depends
+
+from backend.app.config import Settings, get_settings
+from backend.app.database import SQLiteStore, get_store
+from backend.app.services.graphrag import GraphRAGRetriever
+from backend.app.services.incremental import IncrementalUpdater
+from backend.app.services.llm_gateway import LLMGateway
+from backend.app.services.wiki import WikiGenerator
+
+
+class ServiceContainer:
+    def __init__(self, *, settings: Settings, store: SQLiteStore) -> None:
+        self.settings = settings
+        self.store = store
+
+    @cached_property
+    def llm_gateway(self) -> LLMGateway:
+        return LLMGateway(self.settings)
+
+    @cached_property
+    def graph_retriever(self) -> GraphRAGRetriever:
+        return GraphRAGRetriever(store=self.store, settings=self.settings)
+
+    @cached_property
+    def wiki_generator(self) -> WikiGenerator:
+        return WikiGenerator(
+            self.graph_retriever,
+            self.llm_gateway,
+            store=self.store,
+            settings=self.settings,
+        )
+
+    @cached_property
+    def incremental_updater(self) -> IncrementalUpdater:
+        return IncrementalUpdater(
+            store=self.store,
+            graphrag=self.graph_retriever,
+        )
+
+
+@lru_cache
+def get_service_container() -> ServiceContainer:
+    return ServiceContainer(settings=get_settings(), store=get_store())
+
+
+def get_store_dependency() -> SQLiteStore:
+    return get_service_container().store
+
+
+def get_wiki_generator() -> WikiGenerator:
+    return get_service_container().wiki_generator
+
+
+def get_incremental_updater() -> IncrementalUpdater:
+    return get_service_container().incremental_updater
+
+
+StoreDep = Annotated[SQLiteStore, Depends(get_store_dependency)]
+WikiGeneratorDep = Annotated[WikiGenerator, Depends(get_wiki_generator)]
+IncrementalUpdaterDep = Annotated[IncrementalUpdater, Depends(get_incremental_updater)]