What problem does it solve? Setting up actual.inc (Actual Computer) as an inference provider in Hermes involves two distinct modes — a hosted end-to-end-encrypted relay and a local on-device daemon — each with its own authentication, model discovery, and configuration steps, plus several cryptic failure modes that are hard to diagnose without guidance. ## Core Features & Use Cases - Relay Mode Setup: Configure Hermes to route through Actual's hosted API at api.actual.inc using an ac_ key, with model discovery via the /v1/models endpoint. - Local Daemon Mode: Point Hermes at the loopback daemon (http://127.0.0.1:8080) for fully on-device inference with automatic no-auth detection, including GGUF model search, download, and load commands. - Pitfall Diagnosis: Resolve known issues like reasoning_effort 400 errors, context-window overflow producing empty streams on small models, and download-id versus installed-name mismatches. - Use Case: A user with an ac_ key wants Hermes to run on their private Actual cluster; the skill walks through key placement in .env, provider selection, model discovery, and end-to-end verification with hermes chat. ## Quick Start Ask the agent to set up actual.inc as your Hermes inference provider using your ACTUAL_API_KEY, then verify with a test chat message.
