LlamaIndex LLMs Integration: TrustedRouter

TrustedRouter is an OpenAI-compatible LLM router that serves many models behind one endpoint through an open-source, verifiable attested gateway. It does not log prompts or outputs by default.

This package provides a TrustedRouter LLM class for LlamaIndex, implemented as a thin OpenAILike subclass.

Installation

pip install llama-index-llms-trustedrouter

Usage

Set your API key:

export TRUSTEDROUTER_API_KEY=***

from llama_index.llms.trustedrouter import TrustedRouter

llm = TrustedRouter(model="trustedrouter/zdr")

resp = llm.complete("Who is Paul Graham?")
print(resp)

Streaming and chat work like any other OpenAI-like LLM:

from llama_index.core.llms import ChatMessage

resp = llm.chat([ChatMessage(role="user", content="Tell me a joke")])
print(resp)

for chunk in llm.stream_complete("Count to five"):
    print(chunk.delta, end="")

Configuration

Explicit constructor arguments take precedence over environment variables.

Models

Model routes include trustedrouter/auto, trustedrouter/zdr (zero data retention), and trustedrouter/confidential, plus individually addressable models. See the live catalog at https://trustedrouter.com/v1/models.

Why a separate repository

LlamaIndex no longer accepts new integration packages in its main repository; new integrations are maintained in their own repositories and published to PyPI independently. This package follows that guidance, and the import path is unchanged from any in-tree integration.

License

MIT