Update model configurations and enhance LLM runtime logic

- Revised model URLs and descriptions in `llamacpp-models.example.yaml` and `ollama-models.example.yaml` to reflect new recommendations and vision capabilities.
- Updated the LLM runtime logic to support vision projectors and improved model resolution handling.
- Enhanced the installation script to conditionally include vision projectors when available.
- Added tests to validate the inclusion of vision projectors in model presets and ensure proper URL remapping for deprecated models.
- Improved documentation to clarify model usage and configuration options.

src/gpu_rent/remote/install_llamacpp.sh

@@ -100,15 +100,27 @@ print(prefer[0] if prefer else "")
   rm -rf "$TMP"
 fi
 
-# Pick first GGUF if present; otherwise unit starts but API may idle without model.
+# Prefer a weights GGUF (skip mmproj), then attach --mmproj if present.
 MODEL_ARG=""
-FIRST_GGUF="$(find "$MODELS_DIR" -type f \( -name '*.gguf' -o -name '*.GGUF' \) | head -n1 || true)"
+MMPROJ_ARG=""
+FIRST_GGUF="$(
+  find "$MODELS_DIR" -type f \( -name '*.gguf' -o -name '*.GGUF' \) \
+    ! -iname '*mmproj*' 2>/dev/null | head -n1 || true
+)"
+MMPROJ_GGUF="$(
+  find "$MODELS_DIR" -type f \( -iname '*mmproj*.gguf' -o -iname '*mmproj*.GGUF' \) \
+    2>/dev/null | head -n1 || true
+)"
 if [[ -n "$FIRST_GGUF" ]]; then
   MODEL_ARG="-m ${FIRST_GGUF}"
   log "модель ${FIRST_GGUF}"
 else
   log "нет GGUF в ${MODELS_DIR} — положи файл вручную и systemctl restart ${UNIT}"
 fi
+if [[ -n "$MMPROJ_GGUF" ]]; then
+  MMPROJ_ARG="--mmproj ${MMPROJ_GGUF}"
+  log "mmproj ${MMPROJ_GGUF}"
+fi
 
 # GPU layers: share card with Swarm — full offload on mid+, leave headroom on low.
 NGL=99
@@ -144,7 +156,7 @@ Type=simple
 User=${SWARM_USER}
 Group=${SWARM_USER}
 WorkingDirectory=${LLAMA_ROOT}
-ExecStart=${SERVER_BIN} ${MODEL_ARG} --host 127.0.0.1 --port 8080 -ngl ${NGL} -c ${CTX}
+ExecStart=${SERVER_BIN} ${MODEL_ARG} ${MMPROJ_ARG} --host 127.0.0.1 --port 8080 -ngl ${NGL} -c ${CTX}
 Restart=on-failure
 RestartSec=8