Merge pull request #60 from FSoft-AI4Code/feat/run-on-claude-codex-sub

Support running on Claude and Codex subscriptions

Author: anhnh2002

DEVELOPMENT.md

@@ -14,7 +14,10 @@ codewiki/
 │   │   └── adapters/         # External integrations
 │   ├── src/                  # Web application
 │   │   ├── be/               # Backend (dependency analysis, agents)
-│   │   │   ├── agent_orchestrator.py
+│   │   │   ├── backend.py            # LLMBackend abstraction + factory
+│   │   │   ├── pydantic_ai_backend.py  # API-key backend (OpenAI/Anthropic/Bedrock/Azure)
+│   │   │   ├── caw_backend.py        # Subscription backend (claude / codex CLI via caw)
+│   │   │   ├── caw_toolkit.py        # CodeWiki tools exposed to caw via MCP
 │   │   │   ├── agent_tools/
 │   │   │   ├── cluster_modules.py
 │   │   │   ├── dependency_analyzer/
@@ -83,11 +86,15 @@ pip install -r requirements.txt
 - Feature-oriented module partitioning
 - Topological sorting for dependency ordering
 
-#### 3. Agent System (`src/be/agent_orchestrator.py`)
+#### 3. Agent System (`src/be/backend.py`, `pydantic_ai_backend.py`, `caw_backend.py`)
 
-- Recursive agent-based documentation generation
-- Dynamic delegation for complex modules
-- Cross-module reference management
+- ``LLMBackend`` abstracts the API-key and CLI-subscription paths.
+- ``PydanticAIBackend`` runs the per-module agent via pydantic-ai (used by
+  ``openai-compatible`` / ``anthropic`` / ``bedrock`` / ``azure-openai``).
+- ``CawBackend`` routes the per-module agent through the ``claude`` /
+  ``codex`` CLI via the ``caw`` library (used by the ``claude-code`` /
+  ``codex`` providers).  CodeWiki's tools are exposed to the CLI via an MCP
+  server defined in ``caw_toolkit.py``.
 
 #### 4. Agent Tools (`src/be/agent_tools/`)
 
@@ -168,7 +175,7 @@ class AgentInstructions:
 5. **Use in relevant components**:
    - File filtering → `dependency_analyzer/ast_parser.py`
    - Prompts → `be/prompt_template.py`
-   - Agent creation → `be/agent_orchestrator.py`
+   - Agent creation → `be/pydantic_ai_backend.py` (API path) or `be/caw_backend.py` (subscription path)
 
 ---
 

README.md

@@ -19,9 +19,14 @@
   <a href="#quick-start"><strong>Quick Start</strong></a> •
   <a href="#cli-commands"><strong>CLI Commands</strong></a> •
   <a href="#documentation-output"><strong>Output Structure</strong></a> •
+  <a href="./docs/index.html"><strong>Repo Docs</strong></a> •
   <a href="https://arxiv.org/abs/2510.24428"><strong>Paper</strong></a>
 </p>
 
+<p align="center">
+  📚 <strong>CodeWiki documents itself</strong> — browse the generated documentation for this repository at <a href="./docs/index.html"><code>docs/index.html</code></a> (open in a browser, or visit the hosted GitHub Pages site).
+</p>
+
 <p align="center">
   <img src="./img/framework-overview.png" alt="CodeWiki Framework" width="600" style="border: 2px solid #e1e4e8; border-radius: 12px; padding: 20px;"/>
 </p>
@@ -42,11 +47,21 @@ codewiki --version
 
 ### 2. Configure Your Environment
 
-CodeWiki supports multiple LLM providers: **OpenAI-compatible**, **Anthropic**, **AWS Bedrock**, and **Azure OpenAI**.
+CodeWiki supports multiple LLM providers: **OpenAI-compatible**, **Anthropic**, **AWS Bedrock**, **Azure OpenAI**, plus subscription mode via **Claude Code** and **Codex** CLIs (no API key required).
 
 ```bash
+# OpenAI-compatible
+codewiki config set \
+  --provider openai-compatible \
+  --api-key YOUR_API_KEY \
+  --base-url https://api.anthropic.com \
+  --main-model claude-sonnet-4 \
+  --cluster-model claude-sonnet-4 \
+  --fallback-model glm-4p5
+
 # Anthropic
 codewiki config set \
+  --provider anthropic \
   --api-key YOUR_API_KEY \
   --base-url https://api.anthropic.com \
   --main-model claude-sonnet-4 \
@@ -68,8 +83,26 @@ codewiki config set \
   --aws-region us-east-1 \
   --main-model anthropic.claude-sonnet-4-v2:0 \
   --cluster-model anthropic.claude-sonnet-4-v2:0
+
+# Subscription mode (Claude Code) — uses your existing Claude OAuth login.
+# Install the Claude Code CLI and run `claude login` first.
+codewiki config set \
+  --provider claude-code \
+  --main-model claude-sonnet-4-6 \
+  --cluster-model claude-sonnet-4-6
+
+# Subscription mode (Codex) — uses your existing Codex CLI login.
+# Install the Codex CLI and run `codex login` first.
+codewiki config set \
+  --provider codex \
+  --main-model gpt-5.4 \
+  --cluster-model gpt-5.5
 ```
 
+**Subscription mode** routes every LLM call through the local `claude` / `codex` CLI binary (via the [`caw`](https://github.com/zzjas/caw) library), so you can run CodeWiki on a Claude Pro/Max or Codex subscription instead of paying per-token API usage. Claude Code's built-in `Write`/`Edit`/`Bash` tools are disabled inside CodeWiki's agent loop so documentation writes still go through CodeWiki's Mermaid-validating editor.
+
+> **Note on model names.** In subscription mode the model string is forwarded directly to `claude --model` / `codex --model`, so use the bare CLI model name (e.g. `gpt-5.4`, `claude-sonnet-4-6`) — **not** the litellm-style `openai/…` or `anthropic/…` prefix used by `openai-compatible`. If you previously ran with `openai-compatible`, re-run `config set` for **both** `--main-model` and `--cluster-model` to clear any stale prefixes; `config set` only updates the keys you pass.
+
 ### 3. Generate Documentation
 
 ```bash
@@ -289,6 +322,8 @@ Generated documentation includes both **textual descriptions** and **visual arti
 └── index.html               # Interactive viewer (with --github-pages)
 ```
 
+> **See it in action:** This repository's own docs are checked in under [`./docs/`](./docs/) — open [`./docs/index.html`](./docs/index.html) in a browser for the interactive viewer, or start from [`./docs/overview.md`](./docs/overview.md).
+
 ---
 
 ## Experimental Results
@@ -359,6 +394,7 @@ CodeWiki employs a three-stage process for comprehensive documentation generatio
 ## Additional Resources
 
 ### Documentation & Guides
+- **[This Repo's Generated Docs](./docs/index.html)** - Interactive documentation for CodeWiki itself, produced by CodeWiki (start at [`docs/overview.md`](./docs/overview.md))
 - **[MCP Server](codewiki/mcp/)** - Model Context Protocol server for IDE integrations
 - **[Docker Deployment](docker/DOCKER_README.md)** - Containerized deployment instructions
 - **[Development Guide](DEVELOPMENT.md)** - Project structure, architecture, and contributing guidelines

codewiki/cli/adapters/doc_generator.py

@@ -210,12 +210,12 @@ async def _run_backend_generation(self, backend_config: BackendConfig):
         from codewiki.src.be.cluster_modules import cluster_modules
         from codewiki.src.utils import file_manager
         from codewiki.src.config import FIRST_MODULE_TREE_FILENAME, MODULE_TREE_FILENAME
-        
+
         working_dir = str(self.output_dir.absolute())
         file_manager.ensure_directory(working_dir)
         first_module_tree_path = os.path.join(working_dir, FIRST_MODULE_TREE_FILENAME)
         module_tree_path = os.path.join(working_dir, MODULE_TREE_FILENAME)
-        
+
         try:
             if os.path.exists(first_module_tree_path):
                 module_tree = file_manager.load_json(first_module_tree_path)
@@ -224,7 +224,13 @@ async def _run_backend_generation(self, backend_config: BackendConfig):
             else:
                 if self.verbose:
                     self.progress_tracker.update_stage(0.3, f"Clustering {len(leaf_nodes)} leaf nodes with LLM...")
-                module_tree = cluster_modules(leaf_nodes, components, backend_config)
+                cluster_model = backend_config.cluster_model or None
+                module_tree = cluster_modules(
+                    leaf_nodes,
+                    components,
+                    backend_config,
+                    completer=lambda p: doc_generator.backend.complete(p, model=cluster_model),
+                )
                 file_manager.save_json(module_tree, first_module_tree_path)
 
             file_manager.save_json(module_tree, module_tree_path)