TCloud: A TypeScript Toolkit for Decentralized AI Inference

TCloud is the official TypeScript toolkit for Tangle AI Cloud, a decentralized LLM inference network. It provides SDK, CLI, Agent, and Relayer components, supporting decentralized LLM inference, operator routing, reputation-based node selection, and anonymous payment (ShieldedCredits). Key features include OpenAI API compatibility, privacy protection, anti-censorship, and ease of use, addressing issues like privacy risks and single points of failure in centralized AI services.

Centralized AI inference services face privacy risks, censorship concerns, single-point failures, and uncontrollable costs. Decentralized AI inference networks offer an alternative. Tangle AI Cloud is such a network, and TCloud is its official TypeScript toolkit, enabling developers to use distributed computing power while maintaining privacy.

TCloud is an open-source TypeScript toolkit (Apache-2.0 license) supporting Node.js and browser environments. It consists of four core components:

1. tcloud (SDK + CLI): OpenAI-compatible client (zero-dependency fetch), CLI tools (chat, models, operators, etc.), and ShieldedClient for anonymous payments.
2. tcloud-agent: PrivateAgent (manages conversation state with privacy in mind) and PrivateRouter (supports routing strategies like round-robin, min-exposure, etc.), plus Pi extension for edge device integration.
3. tcloud-relayer: Gas relay (covers gas fees for privacy transactions), privacy proxy (strips identifying HTTP headers), and streaming proxy (supports SSE for privacy). Its core idea is to balance privacy/anti-censorship benefits with good development experience and API compatibility.

TCloud's privacy features are built on ShieldedCredits and operator rotation:

• ShieldedCredits: Anonymous payment system using zero-knowledge proofs and EIP-712 SpendAuth, decoupling inference requests from user identity.
• Operator Rotation: Strategies like min-exposure (least interacted operators), geo-location (specific jurisdictions), and latency-aware (balance privacy and performance) to reduce exposure risk.

Development Experience:

• Easy installation: npm install tcloud; CLI commands like npx tcloud chat, tcloud auth set-key.
• SDK examples: OpenAI-compatible code (e.g., client.ask, askStream), privacy mode (TCloud.shielded()).
• PrivateAgent for advanced privacy: supports routing strategies like min-exposure.

Use Cases:

• Privacy-sensitive apps (medical, legal, mental health).
• Anti-censorship communication (distributed operators).
• Enterprise data sovereignty (geo-specific operators).
• Edge computing integration (Pi extension for local sensitive processing).

Technical Details:

• OpenAI compatibility: Seamless migration for existing OpenAI SDK users (same request/response format, parameters, streaming, function calls).
• Reputation system: Operators with good service records get more requests, ensuring network reliability.
• Payment system: Blockchain-based, abstracted via Relayer; users use ShieldedCredits for anonymous payments, operators get rewards.

Comparison:

Feature	TCloud	Centralized API	Self-hosted
Privacy Protection	Strong (anonymous payment + operator rotation)	Weak	Medium (depends on operator)
Anti-censorship	Strong	Weak	Medium
Usability	High (OpenAI compatible)	High	Low
Cost Predictability	Medium	High	Low (hardware cost)
Service Reliability	Medium (network scale-dependent)	High	Depends on self-hosting
Data Sovereignty	Controllable	Weak	Strong

TCloud provides a practical toolkit for decentralized AI inference, allowing developers to experience decentralized services without theoretical barriers. Its OpenAI compatibility lowers migration costs, making it ideal for privacy-focused and anti-censorship applications. Future expectations include more operators, models, and privacy features. For teams wanting to reduce reliance on centralized services, now is a good time to explore TCloud.