Skip to content

Releases: rusmanadodi2598/pannelAI

Experimental release

Experimental release Pre-release
Pre-release

Choose a tag to compare

@rusmanadodi2598 rusmanadodi2598 released this 16 Sep 04:31

Release v0.1.0-experimental: Initial AI Gateway & Router

Welcome to the first experimental release of pannelAI! This version introduces a high-performance custom proxy, AI Gateway, and AI Router designed to unify, manage, and optimize requests across multiple AI providers (OpenAI, Anthropic, Gemini, etc.) through a single endpoint.

⚠️ DISCLAIMER: This is an Experimental Release. Features, configurations, and API structures are subject to breaking changes. Do not use this version in production environments.


Key Features

Smart AI Routing & Multi-Provider

  • Unified API Endpoint: Access multiple LLM providers using a single, standardized request/response format.
  • Dynamic Routing: Smart traffic direction based on cost, latency, model capabilities, or custom rules.

Reliability & Scale

  • Automatic Failover: Instant fallback to backup providers if the primary provider hits rate limits or goes downtime.
  • Load Balancing: Workload distribution across multiple API keys and vendors to maximize throughput.

Security & Proxy Layer

  • API Key Masking: Centralized token management to keep your production vendor keys hidden from client applications.
  • Internal Rate Limiting: Prevent abuse by setting custom request caps at the gateway level.

🛠️ Quick Start

  1. Pull the experimental Docker image:
    docker pull pannelAI/v0.0.1-experimental
    

Known Issues & Limitations

  • Streaming Responses (SSE): Token streaming is currently unstable on certain downstream providers.
  • Cost Tracking: Token cost estimation metrics might experience slight synchronization delays under heavy loads.

Feedback & Contributions

Your testing and feedback are crucial for making pannelAI stable. If you encounter any bugs or have performance suggestions:

Full Changelog: https://github.com/rusmanadodi2598/pannelAI/commits/v0.0.1