AlphaOne HyperRouter Engine

Local AI — Device or Cloud (Priority 1)

Select Local Device for 100% offline GGUF execution with CUDA, Vulkan, or CPU, or select Local Cloud to connect a user-owned Modal-managed GPU inference endpoint. Only the selected Local AI source is used before normal Direct Provider / cascade rollover.

Unified OpenAI-Compatible SSE & Responses Gateway

Translates Local Device AI, Local Cloud Modal, Google Gemini, Groq LPU, OpenAI, Anthropic, DeepSeek, Qwen, Kimi AI, Cloudflare Workers AI, and Agnes AI into standardized OpenAI-compatible SSE streaming and Responses API protocols. Compatible with VS Code, Visual Studio .NET, Android Studio, OpenAI Codex CLI & Codex IDE Extension, Cline, Cursor, Continue.dev, and Antigravity-compatible clients, while preserving active coding sessions across provider rollovers.

LAN Team Sharing & Centralized AI Gateway

Run one HyperRouter instance on a workstation or server and let multiple PCs and laptops connect over the local network through a configurable Server Address and Port. Share Local Device GGUF resources, Local Cloud configuration, cloud provider access, routing settings, and server compute centrally while each client keeps its own independent conversation and Responses session. The internal Local Device llama.cpp runtime remains isolated on loopback rather than being exposed directly to the LAN.

12-Tier Rapid Cascade & Auto RollOver Failover

When an API hits rate limits or quota bounds, HyperRouter automatically tries the next eligible provider or cascade tier. Once downstream output has been committed, it avoids unsafe mid-stream provider switching and preserves canonical continuity for subsequent agent turns.

RPD Auto-Cooldown & Instant Quota Bypass

Tracks daily request and token limits across all model tiers. Automatically quarantines depleted models and bypasses them locally without wasting network roundtrips.

Memory Context Load Limiter

Choose between 🟢 10 Messages (Eco Mode), 🟡 20 Messages (Balanced), or 🔵 Full Context (Default) in Settings to dramatically shrink payload sizes & speed up responses.

Live Status & Logs Dashboard

One-click visual status and request logs dashboards accessible from the Windows System Tray menu, showing real-time 🟢 ACTIVE vs 🔴 SUSPENDED badges, reset timers, and full token metrics.

ANSI Color Code Stripping

Automatically strips terminal ANSI color codes from user prompts to optimize token usage and ensure clean formatting in cascaded requests.

Tool Call Compatibility & Stateful Context Memory

Preserves tool call results across provider transitions (tool_call_id mapping) and persists Google interaction.id state per conversation. For OpenAI & Anthropic, tracks automatic server-side prompt caching (cachedInputTokens) to optimize input token costs on repeated contexts.

Direct OpenAI, Anthropic, DeepSeek, Qwen, Kimi AI & Cloudflare Workers AI Support

Bring official API credentials from OpenAI, Anthropic, DeepSeek, Qwen, Kimi AI, and Cloudflare Workers AI. Cloudflare uses a free-form Model ID entered by the user, so newly available Workers AI models can be selected without waiting for a HyperRouter dropdown update.

Token Usage & Quota Diagnostics

Logs input, cached input, output, thought, tool-use, and total tokens across Google, Cloudflare Workers AI, Agnes AI, Groq, OpenAI, Anthropic, DeepSeek, Qwen, and Kimi AI. Streaming quota errors are classified as RPM, TPM, or RPD with live retry details in the Logs dashboard.

Native Web Search & Real-Time Knowledge Protocol

Equipped with native OpenAI web_search translation and a permanent system-wide protocol across OpenAI, Anthropic, DeepSeek, Qwen, Kimi AI, Google, and Agnes AI for real-time market data, current news, and live API updates.

Multimodal Vision Image Upload Translator

Seamlessly parses image_url and Base64 image attachments from IDEs (Cline, Cursor, Continue) and translates them into native vision blocks for OpenAI, Anthropic Claude, Qwen, Kimi, Google Gemini, and Agnes AI.

Lossless Token Saver & Context Optimization

Provides optional lossless context optimization for Cloud AI requests by reducing unnecessary formatting overhead where a safe transformation can be deterministically verified. HyperRouter prioritizes context integrity over maximum token reduction and automatically leaves unsupported content unchanged. As compatibility may vary with certain tools or uncommon formats, Token Saver can be disabled at any time if unexpected behavior occurs.

1. Core Configuration: Merge this HyperRouter provider configuration into your existing %USERPROFILE%\.codex\config.toml (Windows) or ~/.codex/config.toml (macOS/Linux):

   model = "alphaone-hyperrouter"
   model_provider = "hyperrouter"
   
   [model_providers.hyperrouter]
   name = "AlphaOne HyperRouter"
   base_url = "http://127.0.0.1:11436/v1"
   wire_api = "responses"
   requires_openai_auth = false

2. Recommended Model Catalog (Optional): To allow Codex to cleanly recognize model capabilities (tools, 128k context) without fallback notices, create %USERPROFILE%\.codex\hyperrouter-models.json:

   model_catalog_json = "C:/Users/YOUR_USERNAME/.codex/hyperrouter-models.json"

   {
     "models": [
       {
         "slug": "alphaone-hyperrouter",
         "display_name": "AlphaOne HyperRouter",
         "description": "Single HyperRouter model; High/Medium/Low/Custom is controlled from the HyperRouter Dashboard",
         "default_reasoning_level": null,
         "supported_reasoning_levels": [],
         "shell_type": "shell_command",
         "visibility": "list",
         "supported_in_api": true,
         "priority": 1,
         "upgrade": null,
         "base_instructions": "You are a coding agent operating through AlphaOne HyperRouter. Use the available tools to inspect, edit, run, and test the current workspace as required.",
         "support_verbosity": false,
         "default_verbosity": null,
         "truncation_policy": {
           "mode": "tokens",
           "limit": 10000
         },
         "supports_parallel_tool_calls": true,
         "supports_image_detail_original": false,
         "context_window": 128000,
         "max_context_window": 128000,
         "experimental_supported_tools": []
       }
     ]
   }

3. Codex CLI & Extension: Launch codex in terminal or open Codex Extension in your IDE. Both share this configuration automatically.

1. Open GitHub Copilot Chat (Ctrl + Alt + C) → Model Selector → Manage Models.
2. Click Add Model Provider → Provider Type: Ollama.
3. Base URL: http://127.0.0.1:11436.
4. Click Add Model and enter alphaone-hyperrouter as the model name/ID.
5. Select alphaone-hyperrouter in Copilot Chat. The active High / Medium / Low / Custom profile is controlled centrally from the HyperRouter Dashboard.

1. Press Ctrl + Shift + P → Chat: Manage Language Models → + Add Models.
2. Select Custom Endpoint | Name: AlphaOne AI Auto | Key: sk-alphaone.
3. Click Chat Completions and insert JSON payload:

[
  {
    "name": "AlphaOne AI",
    "vendor": "customendpoint",
    "apiKey": "${input:chat.lm.secret.-601a523f}",
    "apiType": "chat-completions",
    "models": [
      {
        "id": "alphaone-hyperrouter",
        "name": "AlphaOne HyperRouter",
        "url": "http://127.0.0.1:11436/v1/chat/completions",
        "toolCalling": true,
        "vision": true,
        "maxInputTokens": 1048576,
        "maxOutputTokens": 64000
      }
    ]
  }
]

1. Select Manage Models.
2. Click the + (Add) button.
3. Select Local Provider.
4. Description: HyperRouter | Port: 11436.
5. Click Refresh.
6. Select or enter alphaone-hyperrouter. The Dashboard controls the active High / Medium / Low / Custom profile.

• API Provider: OpenAI Compatible
• Base URL: http://127.0.0.1:11436/v1
• Model ID: alphaone-hyperrouter
• API Key: sk-alphaone (Any text)

models:
  - name: AlphaOne HyperRouter Engine
    provider: openai
    model: alphaone-hyperrouter
    apiKey: sk-alphaone
    apiBase: http://127.0.0.1:11436/v1

1 Month

Full 12-Tier Access + Free Upgrades during active period.

3 Months

Full 12-Tier Access + Free Upgrades during active period.

6 Months

Full 12-Tier Access + Free Upgrades during active period.

12 Months

Full 12-Tier Access + Free Upgrades during active period.