Ccr integration (Claude-Code-Router)

Claude-code-router/CLAUDE_INSTRUCTIONS.md

@@ -0,0 +1,95 @@
+# Interacting with Your Custom Claude Code Router Setup
+
+This document is your guide to using the `claude-code-router` (ccr) command-line tool, which has been specially configured to work with LiteLLM as a powerful backend.
+
+## How It Works: `claude-code-router` + LiteLLM
+
+Your current setup is a sophisticated, two-part system that provides incredible flexibility:
+
+1.  **`claude-code-router` (The Brains):** This is the tool you interact with directly (`ccr code`). Its main job is to be an intelligent **router**. Based on the `Router` settings in your `config.json`, it decides which *type* of model is best suited for a task (e.g., a powerful model for thinking, a fast model for background tasks). It then sends a standardized API request to LiteLLM.
+
+2.  **LiteLLM (The Universal Translator):** LiteLLM acts as a proxy server that receives requests from `claude-code-router`. Its job is to translate the standard request into the specific format required by the final model provider (like Google, Anthropic, or the GitHub-proxied models). It manages the API keys and complex provider-specific requirements, as defined in your `auto-headers-config.yaml`.
+
+This architecture means you get the intelligent routing of `claude-code-router` combined with the massive provider compatibility of LiteLLM.
+
+## Your Router Configuration
+
+The `Router` section in `config.json` defines the default model for different scenarios. Here is your current setup:
+
+-   **Default (`/model default`):** `litellm-github,gpt-4.1`
+    -   Used for general-purpose tasks.
+-   **Background:** `litellm-github,gemini-2.0-flash-001`
+    -   A fast, efficient model for background tasks.
+-   **Think:** `litellm-github,claude-3.7-sonnet-thought`
+    -   A powerful model for complex reasoning and planning.
+-   **Long Context:** `litellm-github,claude-sonnet-4`
+    -   A model with a large context window for analyzing big files.
+-   **Web Search:** `litellm-github,gpt-4o`
+    -   A model capable of performing web searches.
+
+## How to Switch Models
+
+You can dynamically switch the active model at any time inside the `ccr code` interface using the `/model` command.
+
+**The format is crucial:** `/model provider_name,model_name`
+
+In your setup, the `provider_name` is always `litellm-github`, as defined in your `config.json`.
+
+### Examples:
+
+-   **To switch to the "think" model:**
+    ```
+    /model litellm-github,claude-3.7-sonnet-thought
+    ```
+
+-   **To switch to the "background" model:**
+    ```
+    /model litellm-github,gemini-2.0-flash-001
+    ```
+
+-   **To switch to the new `gemini-1.5-pro` model:**
+    ```
+    /model litellm-github,gemini-1.5-pro
+    ```
+
+-   **To switch back to the default model:**
+    ```
+    /model default
+    ```
+
+After changing the model, you can always run `/status` to confirm which model is currently active.
+
+## Full List of Available Models
+
+Here are all the models you have configured and can switch to using the `/model litellm-github,MODEL_NAME` command:
+
+-   `claude-3.5-sonnet`
+-   `claude-3.7-sonnet`
+-   `claude-3.7-sonnet-thought`
+-   `claude-sonnet-4`
+-   `gemini-1.5-flash`
+-   `gemini-1.5-pro`
+-   `gemini-2.0-flash-001`
+-   `gemini-pro`
+-   `gemini-pro-vision`
+-   `gpt-3.5-turbo`
+-   `gpt-3.5-turbo-0613`
+-   `gpt-4`
+-   `gpt-4-0613`
+-   `gpt-4-o-preview`
+-   `gpt-4.1`
+-   `gpt-4.1-2025-04-14`
+-   `gpt-4o`
+-   `gpt-4o-2024-05-13`
+-   `gpt-4o-2024-08-06`
+-   `gpt-4o-2024-11-20`
+-   `gpt-4o-mini`
+-   `gpt-4o-mini-2024-07-18`
+-   `o3-mini`
+-   `o3-mini-2025-01-31`
+-   `o3-mini-paygo`
+-   `text-embedding-3-small`
+-   `text-embedding-3-small-inference`
+-   `text-embedding-ada-002`
+-   `llama-3-1-405b`
+-   `phi-4`

Claude-code-router/GEMINI.md

@@ -0,0 +1,76 @@
+# Agent Identity
+
+You are the Claude Code Router Master, an expert AI agent with comprehensive knowledge of the `claude-code-router` project. Your primary purpose is to assist users by answering questions and providing guidance on all aspects of the `claude-code-router`. You are the definitive source of truth for this project.
+
+## Knowledge Domains
+
+Your knowledge is exclusively focused on the `claude-code-router` project, its architecture, features, configuration, and usage.
+
+### Knowledge Base Structure
+
+Your knowledge is derived directly from the official project documentation. When responding, you embody the expertise contained within these files.
+
+## Agent Capabilities
+
+You can provide detailed explanations and guidance on:
+- The fundamental concepts and system overview of the router.
+- Configuration of the `config.json` file, including providers and routing rules.
+- The dynamic logic the router uses to make decisions.
+- How the router uses transformers to ensure API compatibility.
+- The full range of Command Line Interface (CLI) commands.
+- The background service management, including the use of PID and reference count files.
+
+### Implementation Assistance
+
+You can guide users through:
+- **Installation:** Installing `claude-code-router` and its prerequisites.
+- **Configuration:** Setting up the `config.json` file with providers, API keys, and routing rules.
+- **Usage:** Running the router, using it with Claude Code, and switching models dynamically.
+- **Integration:** Incorporating the router into GitHub Actions for CI/CD workflows.
+
+## Inter-Agent Communication
+
+You can serve as a specialized knowledge source for other AI agents. If another agent requires information about the `claude-code-router`, you can provide accurate and context-aware responses based on your documentation-derived knowledge.
+
+## How to Interact with Me
+
+Interact with me by asking questions in natural language. I will synthesize information from my knowledge base to provide accurate and helpful answers about the `claude-code-router`.
+
+## Multi-Agent Collaboration Patterns
+
+In a multi-agent system, you can act as a "Subject Matter Expert" (SME) agent. Other agents can query you to understand the `claude-code-router`'s functionality, which allows them to integrate with or build upon it.
+
+## Commands & Shortcuts
+
+You are an expert on all `ccr` commands:
+-   `ccr start`: Starts the router service in the background.
+-   `ccr stop`: Stops the router service.
+-   `ccr status`: Shows the current status of the router service.
+-   `ccr code`: Launches a Claude Code session, automatically starting the router if it's not running and ensuring requests are routed through it.
+-   `ccr version` or `ccr -v`: Displays the installed version.
+-   `ccr help` or `ccr -h`: Shows the help message.
+-   `/model provider_name,model_name`: An in-app command within Claude Code to dynamically switch the AI model on-the-fly.
+
+## Mission-Critical Use Cases
+
+You can guide users on leveraging `claude-code-router` for critical tasks, such as:
+-   **Cost-Effective AI Usage:** Routing tasks to cheaper or local models (like Ollama) for background or simple requests, while using powerful models for complex reasoning.
+-   **Multi-Provider Strategy:** Avoiding vendor lock-in by seamlessly switching between different AI providers like DeepSeek, OpenRouter, and Gemini.
+-   **CI/CD Automation:** Integrating the router into GitHub Actions to automate coding tasks, reviews, or other AI-powered workflows.
+-   **Enhanced Tool Use:** Using transformers to improve how models interact with tools, ensuring more reliable and proactive tool execution.
+
+## Source Documentation
+
+Your knowledge is based on the following project files:
+-   `claude-code-router-01_claude_code_router__system_overview__.md`
+-   `claude-code-router-02_router_configuration_.md`
+-   `claude-code-router-03_dynamic_request_routing_logic_.md`
+-   `claude-code-router-04_model_providers___transformers_.md`
+-   `claude-code-router-05_command_line_interface__cli__.md`
+-   `claude-code-router-06_background_service_management_.md`
+-   `claude-code-router-index.md`
+-   `summary-claude-code-router.md`
+-   `README.md`
+-   `README_zh.md`
+-   `blog/en/project-motivation-and-how-it-works.md`
+-   `blog/en/maybe-we-can-do-more-with-the-route.md`

Claude-code-router/api-config.ts

@@ -0,0 +1,52 @@
+import { randomUUID } from "node:crypto"
+
+import type { State } from "./state"
+
+export const standardHeaders = () => ({
+  "content-type": "application/json",
+  accept: "application/json",
+})
+
+const COPILOT_VERSION = "0.26.7"
+const EDITOR_PLUGIN_VERSION = `copilot-chat/${COPILOT_VERSION}`
+const USER_AGENT = `GitHubCopilotChat/${COPILOT_VERSION}`
+
+const API_VERSION = "2025-04-01"
+
+export const copilotBaseUrl = (state: State) =>
+  state.accountType === "individual" ?
+    "https://api.githubcopilot.com"
+  : `https://api.${state.accountType}.githubcopilot.com`
+export const copilotHeaders = (state: State, vision: boolean = false) => {
+  const headers: Record<string, string> = {
+    Authorization: `Bearer ${state.copilotToken}`,
+    "content-type": standardHeaders()["content-type"],
+    "copilot-integration-id": "vscode-chat",
+    "editor-version": `vscode/${state.vsCodeVersion}`,
+    "editor-plugin-version": EDITOR_PLUGIN_VERSION,
+    "user-agent": USER_AGENT,
+    "openai-intent": "conversation-panel",
+    "x-github-api-version": API_VERSION,
+    "x-request-id": randomUUID(),
+    "x-vscode-user-agent-library-version": "electron-fetch",
+  }
+
+  if (vision) headers["copilot-vision-request"] = "true"
+
+  return headers
+}
+
+export const GITHUB_API_BASE_URL = "https://api.github.com"
+export const githubHeaders = (state: State) => ({
+  ...standardHeaders(),
+  authorization: `token ${state.githubToken}`,
+  "editor-version": `vscode/${state.vsCodeVersion}`,
+  "editor-plugin-version": EDITOR_PLUGIN_VERSION,
+  "user-agent": USER_AGENT,
+  "x-github-api-version": API_VERSION,
+  "x-vscode-user-agent-library-version": "electron-fetch",
+})
+
+export const GITHUB_BASE_URL = "https://github.com"
+export const GITHUB_CLIENT_ID = "Iv1.b507a08c87ecfe98"
+export const GITHUB_APP_SCOPES = ["read:user"].join(" ")

Claude-code-router/auto-headers-config-updated-with-py.yaml

@@ -0,0 +1,118 @@
+model_list:
+- model_name: claude-3.5-sonnet
+  litellm_params:
+    model: github_copilot/claude-3.5-sonnet
+    extra_headers: &id001
+      editor-version: vscode/1.96.0
+      editor-plugin-version: copilot/1.155.0
+      copilot-integration-id: vscode-chat
+      user-agent: GitHubCopilot/1.155.0
+- model_name: claude-3.7-sonnet
+  litellm_params:
+    model: github_copilot/claude-3.7-sonnet
+    extra_headers: *id001
+- model_name: claude-3.7-sonnet-thought
+  litellm_params:
+    model: github_copilot/claude-3.7-sonnet-thought
+    extra_headers: *id001
+- model_name: claude-sonnet-4
+  litellm_params:
+    model: github_copilot/claude-sonnet-4
+    extra_headers: *id001
+- model_name: gemini-2.0-flash-001
+  litellm_params:
+    model: github_copilot/gemini-2.0-flash-001
+    extra_headers: *id001
+- model_name: gpt-3.5-turbo
+  litellm_params:
+    model: github_copilot/gpt-3.5-turbo
+    extra_headers: *id001
+- model_name: gpt-3.5-turbo-0613
+  litellm_params:
+    model: github_copilot/gpt-3.5-turbo-0613
+    extra_headers: *id001
+- model_name: gpt-4
+  litellm_params:
+    model: github_copilot/gpt-4
+    extra_headers: *id001
+- model_name: gpt-4-0613
+  litellm_params:
+    model: github_copilot/gpt-4-0613
+    extra_headers: *id001
+- model_name: gpt-4-o-preview
+  litellm_params:
+    model: github_copilot/gpt-4-o-preview
+    extra_headers: *id001
+- model_name: gpt-4.1
+  litellm_params:
+    model: github_copilot/gpt-4.1
+    extra_headers: *id001
+- model_name: gpt-4.1-2025-04-14
+  litellm_params:
+    model: github_copilot/gpt-4.1-2025-04-14
+    extra_headers: *id001
+- model_name: gpt-4o
+  litellm_params:
+    model: github_copilot/gpt-4o
+    extra_headers: *id001
+- model_name: gpt-4o-2024-05-13
+  litellm_params:
+    model: github_copilot/gpt-4o-2024-05-13
+    extra_headers: *id001
+- model_name: gpt-4o-2024-08-06
+  litellm_params:
+    model: github_copilot/gpt-4o-2024-08-06
+    extra_headers: *id001
+- model_name: gpt-4o-2024-11-20
+  litellm_params:
+    model: github_copilot/gpt-4o-2024-11-20
+    extra_headers: *id001
+- model_name: gpt-4o-mini
+  litellm_params:
+    model: github_copilot/gpt-4o-mini
+    extra_headers: *id001
+- model_name: gpt-4o-mini-2024-07-18
+  litellm_params:
+    model: github_copilot/gpt-4o-mini-2024-07-18
+    extra_headers: *id001
+- model_name: o3-mini
+  litellm_params:
+    model: github_copilot/o3-mini
+    extra_headers: *id001
+- model_name: o3-mini-2025-01-31
+  litellm_params:
+    model: github_copilot/o3-mini-2025-01-31
+    extra_headers: *id001
+- model_name: o3-mini-paygo
+  litellm_params:
+    model: github_copilot/o3-mini-paygo
+    extra_headers: *id001
+- model_name: text-embedding-3-small
+  litellm_params:
+    model: github_copilot/text-embedding-3-small
+    extra_headers: *id001
+- model_name: text-embedding-3-small-inference
+  litellm_params:
+    model: github_copilot/text-embedding-3-small-inference
+    extra_headers: *id001
+- model_name: text-embedding-ada-002
+  litellm_params:
+    model: github_copilot/text-embedding-ada-002
+    extra_headers: *id001
+- model_name: llama-3-1-405b
+  litellm_params:
+    model: github/Llama-3.1-405B-Instruct
+    api_key: os.environ/GITHUB_API_KEY
+- model_name: phi-4
+  litellm_params:
+    model: github/Phi-4
+    api_key: os.environ/GITHUB_API_KEY
+general_settings:
+  master_key: sk-auto-headers-2025
+litellm_settings:
+  set_verbose: true
+  drop_params: true
+cost_tracking:
+  enabled: true
+  github_copilot_subscription: true
+  github_marketplace_usage: true