router-maestro 0.1.6__tar.gz → 0.1.7__tar.gz

{router_maestro-0.1.6 → router_maestro-0.1.7}/CLAUDE.md

@@ -51,8 +51,8 @@ Router-Maestro is a multi-model routing system that exposes both OpenAI-compatib
 
 **Server (`src/router_maestro/server/`)**
 - FastAPI application with two API flavors:
-  - OpenAI-compatible: `/v1/chat/completions`, `/v1/models`
-  - Anthropic-compatible: `/v1/messages`, `/api/anthropic/v1/messages`
+  - OpenAI-compatible: `/api/openai/v1/chat/completions`, `/api/openai/v1/models`
+  - Anthropic-compatible: `/api/anthropic/v1/messages`
 - `translation.py` - Converts between Anthropic and OpenAI request/response formats
 - `schemas/` - Pydantic models for both API formats
 

{router_maestro-0.1.6 → router_maestro-0.1.7}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: router-maestro
-Version: 0.1.6
+Version: 0.1.7
 Summary: Multi-model routing and load balancing system with OpenAI-compatible API
 Author-email: Kanwen Li <likanwen@icloud.com>
 License-Expression: MIT
@@ -26,6 +26,7 @@ Requires-Dist: pydantic>=2.5.0
 Requires-Dist: python-dotenv>=1.0.0
 Requires-Dist: rich>=13.7.0
 Requires-Dist: tiktoken>=0.5.0
+Requires-Dist: tomlkit>=0.12.0
 Requires-Dist: typer>=0.12.0
 Requires-Dist: uvicorn>=0.27.0
 Provides-Extra: dev
@@ -118,14 +119,33 @@ router-maestro auth login github-copilot
 #   3. Authorize "GitHub Copilot Chat"
 ```
 
-### 4. Configure Claude Code
+### 4. Configure Your CLI Tool
+
+#### Claude Code
 
 ```bash
 router-maestro config claude-code
 # Follow the wizard to select models
 ```
 
-**Done!** Now run `claude` and your requests will route through Router-Maestro.
+#### OpenAI Codex (CLI, Extension, App)
+
+```bash
+router-maestro config codex
+# Follow the wizard to select models
+```
+
+After configuration, set the API key environment variable:
+
+```bash
+# Get your API key
+router-maestro server show-key
+
+# Set the environment variable (add to your shell profile)
+export ROUTER_MAESTRO_API_KEY="your-api-key-here"
+```
+
+**Done!** Now run `claude` or `codex` and your requests will route through Router-Maestro.
 
 > **For production deployment**, see the [Deployment](#deployment) section.
 
@@ -258,6 +278,7 @@ router-maestro model list
 | Command              | Description                   |
 | -------------------- | ----------------------------- |
 | `config claude-code` | Generate Claude Code settings |
+| `config codex`       | Generate Codex config (CLI/Extension/App) |
 
 ## API Reference
 

{router_maestro-0.1.6 → router_maestro-0.1.7}/README.md

@@ -81,14 +81,33 @@ router-maestro auth login github-copilot
 #   3. Authorize "GitHub Copilot Chat"
 ```
 
-### 4. Configure Claude Code
+### 4. Configure Your CLI Tool
+
+#### Claude Code
 
 ```bash
 router-maestro config claude-code
 # Follow the wizard to select models
 ```
 
-**Done!** Now run `claude` and your requests will route through Router-Maestro.
+#### OpenAI Codex (CLI, Extension, App)
+
+```bash
+router-maestro config codex
+# Follow the wizard to select models
+```
+
+After configuration, set the API key environment variable:
+
+```bash
+# Get your API key
+router-maestro server show-key
+
+# Set the environment variable (add to your shell profile)
+export ROUTER_MAESTRO_API_KEY="your-api-key-here"
+```
+
+**Done!** Now run `claude` or `codex` and your requests will route through Router-Maestro.
 
 > **For production deployment**, see the [Deployment](#deployment) section.
 
@@ -221,6 +240,7 @@ router-maestro model list
 | Command              | Description                   |
 | -------------------- | ----------------------------- |
 | `config claude-code` | Generate Claude Code settings |
+| `config codex`       | Generate Codex config (CLI/Extension/App) |
 
 ## API Reference
 

{router_maestro-0.1.6 → router_maestro-0.1.7}/pyproject.toml

@@ -1,6 +1,6 @@
 [project]
 name = "router-maestro"
-version = "0.1.6"
+version = "0.1.7"
 description = "Multi-model routing and load balancing system with OpenAI-compatible API"
 readme = "README.md"
 license = "MIT"
@@ -49,6 +49,9 @@ dependencies = [
     # Utils
     "aiosqlite>=0.19.0",
     "python-dotenv>=1.0.0",
+
+    # TOML writing (for Codex config generation)
+    "tomlkit>=0.12.0",
 ]
 
 [project.optional-dependencies]

{router_maestro-0.1.6 → router_maestro-0.1.7}/src/router_maestro/__init__.py

@@ -1,3 +1,3 @@
 """Router-Maestro: Multi-model routing and load balancing system."""
 
-__version__ = "0.1.6"
+__version__ = "0.1.7"

{router_maestro-0.1.6 → router_maestro-0.1.7}/src/router_maestro/cli/config.py

@@ -3,9 +3,11 @@
 import asyncio
 import json
 import shutil
+import tomllib
 from datetime import datetime
 from pathlib import Path
 
+import tomlkit
 import typer
 from rich.console import Console
 from rich.panel import Panel
@@ -24,6 +26,10 @@ CLI_TOOLS = {
         "name": "Claude Code",
         "description": "Generate settings.json for Claude Code CLI",
     },
+    "codex": {
+        "name": "OpenAI Codex",
+        "description": "Generate config.toml for OpenAI Codex CLI",
+    },
 }
 
 
@@ -35,6 +41,14 @@ def get_claude_code_paths() -> dict[str, Path]:
     }
 
 
+def get_codex_paths() -> dict[str, Path]:
+    """Get Codex config paths."""
+    return {
+        "user": Path.home() / ".codex" / "config.toml",
+        "project": Path.cwd() / ".codex" / "config.toml",
+    }
+
+
 @app.callback(invoke_without_command=True)
 def config_callback(ctx: typer.Context) -> None:
     """Generate configuration for CLI tools (interactive selection if not specified)."""
@@ -60,6 +74,8 @@ def config_callback(ctx: typer.Context) -> None:
     # Dispatch to the appropriate command
     if tool_key == "claude-code":
         claude_code_config()
+    elif tool_key == "codex":
+        codex_config()
 
 
 @app.command(name="claude-code")
@@ -175,3 +191,113 @@ def claude_code_config() -> None:
             border_style="green",
         )
     )
+
+
+@app.command(name="codex")
+def codex_config() -> None:
+    """Generate OpenAI Codex CLI config.toml for router-maestro."""
+    # Step 1: Select level
+    console.print("\n[bold]Step 1: Select configuration level[/bold]")
+    console.print("  1. User-level (~/.codex/config.toml)")
+    console.print("  2. Project-level (./.codex/config.toml)")
+    choice = Prompt.ask("Select", choices=["1", "2"], default="1")
+
+    paths = get_codex_paths()
+    level = "user" if choice == "1" else "project"
+    config_path = paths[level]
+
+    # Step 2: Backup if exists
+    if config_path.exists():
+        console.print(f"\n[yellow]config.toml already exists at {config_path}[/yellow]")
+        if Confirm.ask("Backup existing file?", default=True):
+            backup_path = config_path.with_suffix(
+                f".toml.backup.{datetime.now().strftime('%Y%m%d_%H%M%S')}"
+            )
+            shutil.copy(config_path, backup_path)
+            console.print(f"[green]Backed up to {backup_path}[/green]")
+
+    # Step 3: Get models from server
+    try:
+        client = get_admin_client()
+        models = asyncio.run(client.list_models())
+    except ServerNotRunningError as e:
+        console.print(f"[red]{e}[/red]")
+        console.print("[dim]Tip: Start router-maestro server first.[/dim]")
+        raise typer.Exit(1)
+    except Exception as e:
+        console.print(f"[red]Error: {e}[/red]")
+        raise typer.Exit(1)
+
+    if not models:
+        console.print("[red]No models available. Please authenticate first.[/red]")
+        raise typer.Exit(1)
+
+    # Display models
+    console.print("\n[bold]Available models:[/bold]")
+    table = Table()
+    table.add_column("#", style="dim")
+    table.add_column("Model Key", style="green")
+    table.add_column("Name", style="white")
+    for i, model in enumerate(models, 1):
+        table.add_row(str(i), f"{model['provider']}/{model['id']}", model["name"])
+    console.print(table)
+
+    # Select model
+    console.print("\n[bold]Step 2: Select model[/bold]")
+    model_choice = Prompt.ask("Enter number (or 0 for auto-routing)", default="0")
+    selected_model = "router-maestro"
+    if model_choice != "0" and model_choice.isdigit():
+        idx = int(model_choice) - 1
+        if 0 <= idx < len(models):
+            m = models[idx]
+            selected_model = f"{m['provider']}/{m['id']}"
+
+    # Step 4: Generate config
+    client = get_admin_client()
+    base_url = (
+        client.endpoint.rstrip("/") if hasattr(client, "endpoint") else "http://localhost:8080"
+    )
+    openai_url = f"{base_url}/api/openai/v1"
+
+    # Load existing config to preserve other sections
+    existing_config: tomlkit.TOMLDocument = tomlkit.document()
+    if config_path.exists():
+        try:
+            with open(config_path, "rb") as f:
+                existing_config = tomlkit.load(f)
+        except (tomllib.TOMLDecodeError, OSError):
+            pass  # If file is corrupted, start fresh
+
+    # Update configuration
+    existing_config["model"] = selected_model
+    existing_config["model_provider"] = "router-maestro"
+
+    # Create or update model_providers section
+    if "model_providers" not in existing_config:
+        existing_config["model_providers"] = tomlkit.table()
+
+    provider_config = tomlkit.table()
+    provider_config["name"] = "Router Maestro"
+    provider_config["base_url"] = openai_url
+    provider_config["env_key"] = "ROUTER_MAESTRO_API_KEY"
+    provider_config["wire_api"] = "responses"
+    existing_config["model_providers"]["router-maestro"] = provider_config
+
+    # Write config
+    config_path.parent.mkdir(parents=True, exist_ok=True)
+    with open(config_path, "w", encoding="utf-8") as f:
+        f.write(tomlkit.dumps(existing_config))
+
+    console.print(
+        Panel(
+            f"[green]Created {config_path}[/green]\n\n"
+            f"Model: {selected_model}\n\n"
+            f"Endpoint: {openai_url}\n\n"
+            "[dim]Start router-maestro server before using Codex:[/dim]\n"
+            "  router-maestro server start\n\n"
+            "[dim]Set API key environment variable (optional):[/dim]\n"
+            "  export ROUTER_MAESTRO_API_KEY=your-key",
+            title="Success",
+            border_style="green",
+        )
+    )

{router_maestro-0.1.6 → router_maestro-0.1.7}/src/router_maestro/providers/__init__.py

@@ -9,6 +9,10 @@ from router_maestro.providers.base import (
     Message,
     ModelInfo,
     ProviderError,
+    ResponsesRequest,
+    ResponsesResponse,
+    ResponsesStreamChunk,
+    ResponsesToolCall,
 )
 from router_maestro.providers.copilot import CopilotProvider
 from router_maestro.providers.openai import OpenAIProvider
@@ -23,6 +27,10 @@ __all__ = [
     "ChatResponse",
     "ChatStreamChunk",
     "ModelInfo",
+    "ResponsesRequest",
+    "ResponsesResponse",
+    "ResponsesStreamChunk",
+    "ResponsesToolCall",
     # Providers
     "CopilotProvider",
     "OpenAIProvider",

{router_maestro-0.1.6 → router_maestro-0.1.7}/src/router_maestro/providers/base.py

@@ -59,6 +59,53 @@ class ModelInfo:
     provider: str
 
 
+@dataclass
+class ResponsesToolCall:
+    """A tool/function call from the Responses API."""
+
+    call_id: str
+    name: str
+    arguments: str
+
+
+@dataclass
+class ResponsesRequest:
+    """Request for the Responses API (used by Codex models)."""
+
+    model: str
+    input: str | list  # Can be string or list of message dicts
+    stream: bool = False
+    instructions: str | None = None
+    temperature: float = 1.0
+    max_output_tokens: int | None = None
+    # Tool support
+    tools: list[dict] | None = None
+    tool_choice: str | dict | None = None
+    parallel_tool_calls: bool | None = None
+
+
+@dataclass
+class ResponsesResponse:
+    """Response from the Responses API."""
+
+    content: str
+    model: str
+    usage: dict | None = None
+    tool_calls: list[ResponsesToolCall] | None = None
+
+
+@dataclass
+class ResponsesStreamChunk:
+    """A chunk from streaming Responses API completion."""
+
+    content: str
+    finish_reason: str | None = None
+    usage: dict | None = None
+    # Tool call support
+    tool_call: ResponsesToolCall | None = None  # A complete tool call
+    tool_call_delta: dict | None = None  # Partial tool call for streaming
+
+
 class ProviderError(Exception):
     """Error from a provider."""
 
@@ -121,3 +168,36 @@ class BaseProvider(ABC):
         Override this for providers that need token refresh.
         """
         pass
+
+    async def responses_completion(self, request: ResponsesRequest) -> ResponsesResponse:
+        """Generate a Responses API completion (for Codex models).
+
+        Args:
+            request: Responses completion request
+
+        Returns:
+            Responses completion response
+
+        Raises:
+            NotImplementedError: If provider does not support Responses API
+        """
+        raise NotImplementedError("Provider does not support Responses API")
+
+    async def responses_completion_stream(
+        self, request: ResponsesRequest
+    ) -> AsyncIterator[ResponsesStreamChunk]:
+        """Generate a streaming Responses API completion (for Codex models).
+
+        Args:
+            request: Responses completion request
+
+        Yields:
+            Responses completion chunks
+
+        Raises:
+            NotImplementedError: If provider does not support Responses API
+        """
+        raise NotImplementedError("Provider does not support Responses API")
+        # Make this a generator (required for type checking)
+        if False:
+            yield ResponsesStreamChunk(content="")