From 85d216fce38ea8f9caab6aebdaf51e474fa77a15 Mon Sep 17 00:00:00 2001
From: "K. Isom" <kyle@imap.cc>
Date: Wed, 16 Sep 2026 05:56:27 -0700
Subject: [PATCH] straylight: add Ornith 1.5 35B-A3B MoE to llama-server

Official ornith-ai Q4_K_M GGUF (~22 GB). Multimodal 35B MoE with ~3B
active; llama-server fetches weights and mmproj on first load.
---
 hw/straylight/default.nix | 6 ++++++
 1 file changed, 6 insertions(+)

diff --git a/hw/straylight/default.nix b/hw/straylight/default.nix
index 1d7dee6..3e8c3ab 100644
--- a/hw/straylight/default.nix
+++ b/hw/straylight/default.nix
@@ -56,6 +56,12 @@ let
     hf-repo = mradermacher/Ornith-1.5-9B-uncensored-GGUF:Q4_K_M
     dedup-cache-models = true
 
+    ; Ornith 1.5 35B-A3B MoE (official GGUF). ~3B active, multimodal; mmproj
+    ; comes with --hf-repo. Q4_K_M is ~22 GB.
+    [ornith-1.5-35b-a3b]
+    hf-repo = ornith-ai/Ornith-1.5-35B-A3B-GGUF:Q4_K_M
+    dedup-cache-models = true
+
     ; Qwen 3.8 27B dense, refusal-ablated. Preferred uncensored coding agent.
     ; orcarouter/Qwen3.8-27B-Uncensored-GGUF is gated; huihui Q4_K_L keeps
     ; ablation-target tensors at Q8_0. llama-server fetches on first load.