Skip to main content

​
TrustGate

TrustGate is an OpenAI-compatible API proxy that routes inference requests to the optimal LLM provider based on passport policies.

​
How It Works

Client → TrustGate (:4010) → Passport Matching → LLM Provider → Receipt
  1. Your app sends a standard OpenAI POST /v1/chat/completions request
  2. TrustGate matches the request against passport policies (model, cost, latency, capabilities)
  3. The best-fit provider is selected and the request is forwarded
  4. A cryptographic receipt is generated for every inference
  5. Receipts are batched into epochs and anchored on Solana

​
OpenAI Compatibility

TrustGate implements the OpenAI Chat Completions API. Any OpenAI SDK works — just change the base URL: 
import OpenAI from "openai";

const client = new OpenAI({
  apiKey: "your-lucid-api-key",
  baseURL: "https://api.lucid.foundation/v1",
});

const response = await client.chat.completions.create({
  model: "gpt-4o",  // or any passport ID
  messages: [{ role: "user", content: "Hello!" }],
});

​
Passport Routing

Instead of hardcoding a model, you can specify passport requirements and let TrustGate find the best match: 
const response = await lucid.inference.run({
  passportId: "your-passport-id",
  messages: [{ role: "user", content: "Analyze this data" }],
});
Passport matching considers:
  • Model capabilities — vision, function calling, code generation
  • Cost constraints — max price per million tokens
  • Latency requirements — time-to-first-token targets
  • Provider preferences — specific providers or exclusions

​
Features

FeatureDescription
Multi-providerOpenAI, Anthropic, Google, Mistral, open-source via LiteLLM
StreamingSSE streaming with receipt generation
Quota enforcementPer-org, per-key rate limits and token budgets
MeteringReal-time usage tracking via OpenMeter
Receipt signingEd25519 signatures on every inference
Epoch batchingReceipts batched and anchored on Solana

​
Architecture

TrustGate is part of lucid-plateform-core, running as a Fastify service on port 4010. It depends on:
  • LiteLLM — unified LLM provider interface
  • @raijinlabs/passport — passport matching engine
  • Supabase — receipt and quota storage
  • OpenMeter — usage metering