Gyldby OpSpot

Get started

Gyld docs

Run a niche AI employee in one line. Self-host the open weights for free, or call the OpenAI-compatible hosted API — change one line and the rest of your stack never notices.

Two ways to run it

Every Gyld model ships as open weights and on a hosted API. Both speak the OpenAI Chat Completions shape, so you can move between them without touching your code.

  • Self-host — free. Download the Apache-2.0 weights and run them with llama.cpp or vLLM. Your data never leaves the machine, and a lapsed plan never bricks a model you already hold.
  • Hosted API — subscription. No hardware to run. Mint a key, point any OpenAI SDK at us, and hit the niche model over a clean endpoint.

Quickstart

Mint a key on the Account page, then send a request. The model field is the niche slug (e.g. agency-ops or fitness):

shell
curl https://gyld.dev/api/v1/chat/completions \
  -H "Authorization: Bearer $GYLD_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "agency-ops",
    "messages": [{ "role": "user", "content": "Draft a discovery-call agenda." }]
  }'

It’s the same shape as OpenAI — point any SDK at https://gyld.dev/api/v1. Full walkthrough in the Quickstart guide.

Download the weights

Prefer to run it yourself? Every published model is free on Hugging Face and one command to pull.

Tip.Browse and download every model on the Open Source page, or pull them straight from the flywheel-ai org on Hugging Face. See Self-hosting for hardware sizing and serving with vLLM or llama.cpp.

Reference