chutes-ai

Chutes.ai for Hermes

Use this skill when the user asks about:

• Chutes.ai setup or usage
• open-source model inference through Chutes
• model routing / failover / latency / throughput
• TEE / confidential compute on Chutes
• billing, balances, discounts, or quotas on Chutes
• using Chutes as a Hermes model backend

Source of truth for live model metadata

When current model inventory, pricing, supported features, TTFT/TPS, or TEE status matter, use:

https://llm.chutes.ai/v1/models

Do not treat static model snapshots as authoritative.

Base URLs

• Management API: https://api.chutes.ai
• Inference API: https://llm.chutes.ai/v1

Authentication

Use a cpk_... API key. Store it as CHUTES_API_KEY in ~/.hermes/.env or in this repo's keychain-backed credential manager; never paste the raw key into config examples or chat.

Chutes auth was re-verified in the shared Chutes skills on 2026-06-11 and again 2026-10-07: use standard bearer authorization semantics (Authorization: Bearer plus the cpk_... value) for configured providers. GET /v1/models is public and should not be used as proof that a specific auth header works. Older April notes about X-API-Key are superseded for Hermes-facing setup.

Hermes setup pattern

Typical Hermes config:

providers:
  chutes:
    name: Chutes
    base_url: https://llm.chutes.ai/v1
    key_env: CHUTES_API_KEY
    transport: chat_completions
    default_model: default:latency
    discover_models: true
    models:
      default: {}
      "default:latency": {}
      "default:throughput": {}

model:
  provider: custom:chutes
  default: default:latency

Legacy Hermes configs may use custom_providers: with equivalent name, base_url, key_env, api_mode, and model fields.

The former research opt-in endpoint (research-data-opt-in-proxy.chutes.ai) is defunct — verified 404 and no longer advertised as of 2026-10-07. Do not configure it.

Routing guidance

Chutes supports:

• default
• default:latency
• default:throughput
• inline routing strings like modelA,modelB,modelC:latency