compose: replace open-webui with AnythingLLM on swap-stack endpoint; drop IQ4 GGUFs

AnythingLLM (port 3001, profile webui) points at llama-swap-stack:8080
via generic-openai, default model ornith-35b-duo (131K ctx). Deleted the
two >RAM IQ4 GGUFs (36.5GB — thrash-only until 64GB RAM upgrade) and
their swap-stack entries; re-download paths noted in config comment.

Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>

swap-stack/config.yaml

@@ -102,17 +102,6 @@ models:
       --cont-batching --parallel 1
     ttl: 300
 
-  "qwen36-35b":
-    name: "Qwen3.6-35B-A3B UD-IQ4_XS"
-    description: "35B/3B-active MoE, 17.7GB mmap > RAM. Stop heavy containers first"
-    cmd: |
-      ${server-base} ${moe-offload} ${q8-kv}
-      --model /models/Qwen3.6-35B-A3B-UD-IQ4_XS.gguf
-      --ctx-size 32768
-      --batch-size 1024 --ubatch-size 512
-      --cont-batching --parallel 1
-    ttl: 300
-
   "qwen36-35b-q2":
     name: "Qwen3.6-35B-A3B UD-Q2_K_XL"
     description: "DAILY DRIVER 35B: 23.0 tg / 36 pp. ds4 asymmetric recipe (dense high-bit, experts 2-bit), 12.3GB page-cache resident, last 3 expert layers in VRAM. Gates pass"
@@ -137,16 +126,9 @@ models:
       --cont-batching --parallel 1
     ttl: 300
 
-  "ornith-35b-iq4":
-    name: "Ornith-1.0-35B IQ4_XS"
-    description: "Quality-first variant, 18.8GB mmap > RAM = ~2.8 t/s thrash. Batch jobs only (or post-RAM-upgrade)"
-    cmd: |
-      ${server-base} ${moe-offload} ${q8-kv}
-      --model /models/deepreinforce-ai_Ornith-1.0-35B-IQ4_XS.gguf
-      --ctx-size 32768
-      --batch-size 1024 --ubatch-size 512
-      --cont-batching --parallel 1
-    ttl: 300
+  # (IQ4 variants of ornith-35b/qwen36-35b deleted 2026-07-10 — 36.5GB disk,
+  #  both were >RAM thrash-only until the 64GB upgrade. Re-download if needed:
+  #  bartowski Ornith-1.0-35B IQ4_XS, unsloth Qwen3.6-35B-A3B UD-IQ4_XS.)
 
   # ── Dense 9B (RAM-bandwidth-bound, ~4.4 t/s) ───────────────────────────────
   # (ik_llama / upstream-master A/B variants removed 2026-07-10 — questions