From 95dd5e971187429248ebad2f1daaa5b728146170 Mon Sep 17 00:00:00 2001 From: mathkartel Date: Thu, 19 Nov 2026 11:58:37 -0300 Subject: [PATCH] docs: recipe for OpenAI-compatible gateways (NotHoly example) --- README.md | 8 ++++++++ 1 file changed, 8 insertions(+) diff --git a/README.md b/README.md index 307a035..c198402 100644 --- a/README.md +++ b/README.md @@ -428,6 +428,14 @@ runtime supports it. `/chat/completions` (or `/responses` for `openai-responses`), so their base **does** end in `/v1` (for example `https://api.openai.com/v1`). +**OpenAI-compatible gateways**: any provider that serves an OpenAI-style +`/v1/chat/completions` endpoint works through the **OpenAI** format — no dedicated format +needed. Create a profile with format **OpenAI**, set **Base URL** to the gateway's `/v1` +root, paste the key, then click the **↻** button next to *Model* to load the provider's +catalog (`GET /v1/models`) and pick a model; models the provider adds later appear on the +next refresh. For example, [NotHoly](https://notholy.com) (offensive-security models) uses +base URL `https://api.notholy.com/v1` with an API key from its console. + **Concurrency**: work agents per task, in "System Settings" (default 3). **Untrusted-output boundary (on by default)**: tool results for the worker, planner, chat and main