chore: auto-commit (5 archivos)

- CMakeLists.txt
- app.md
- appicon.ico
- backend/
- main.cpp

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>

app.md

@@ -0,0 +1,71 @@
+---
+name: image_to_3d_studio
+lang: cpp
+domain: gfx
+description: "Image to 3D studio: imagen origen + POST a backend Python (TripoSR/Hunyuan3D/Trellis) y guarda mesh GLB."
+tags: [imagegen, 3d, mesh, viewer, imgui]
+icon:
+  phosphor: "cube-transparent"
+  accent: "#0ea5e9"
+uses_functions:
+  - http_request_cpp_core       # POST /generate al backend Python
+  - gl_texture_load_cpp_gfx     # preview de la imagen origen en panel Input
+  - gltf_load_mesh_cpp_gfx      # parse GLB -> Mesh CPU para el viewer 3D
+  - mesh_gpu_cpp_gfx            # sube Mesh a GPU (VAO/VBO/EBO)
+  - mesh_viewer_cpp_viz         # render orbit + FBO depth del mesh
+  - orbit_camera_cpp_core       # camara orbital (drag rotar / wheel zoom)
+  - gl_framebuffer_cpp_gfx      # FBO con depth para el render offscreen
+uses_types: []
+framework: "imgui"
+entry_point: "main.cpp"
+dir_path: "projects/imagegen/apps/image_to_3d_studio"
+repo_url: "https://gitea.organic-machine.com/dataforge/image_to_3d_studio"
+---
+
+# image_to_3d_studio
+
+UI ImGui que envia una imagen al backend Python (FastAPI) y guarda el
+mesh GLB resultante en `local_files/cache/`. Despachador soporta varios
+modelos image-to-3D detras del mismo endpoint:
+
+- **TripoSR** (MIT, ~0.5 s, ~6 GB VRAM) — backend listo.
+- **Hunyuan3D-2** (Tencent Community, mejor textura) — stub.
+- **Trellis** (Microsoft MIT, mesh/3DGS/NeRF) — stub.
+
+Viewer GLB integrado: panel **Viewer 3D** carga el mesh con `gltf_load_mesh`,
+lo sube a GPU y lo renderiza con `mesh_viewer` (orbit camera + FBO con depth).
+Drag = rotar, rueda = zoom, "Reset cam" recentra, checkbox wireframe. Auto-carga
+el mesh al terminar Generate; tambien boton "View 3D" en el panel Output.
+
+## Arquitectura
+
+```
+image_to_3d_studio (C++ ImGui)
+  ── multipart POST /generate ──▶  backend FastAPI 127.0.0.1:8600
+                                     dispatcher → backends/<id>.py
+  ◀── bytes GLB                  ──
+  guarda en <exe>/local_files/cache/<model>_<hash>.glb
+```
+
+## Build
+
+```bash
+cd cpp && cmake --build build --target image_to_3d_studio -j
+```
+
+## Run
+
+```bash
+# 1) backend Python (otra terminal)
+projects/imagegen/apps/image_to_3d_studio/backend/run.sh
+
+# 2) app C++
+./cpp/build/image_to_3d_studio
+```
+
+## Estado
+
+- Notebook `projects/imagegen/analysis/spike_image_to_3d/notebooks/01_panorama_image_to_3d.ipynb` — panorama + tabla comparativa.
+- Notebook `02_smoke_triposr.ipynb` — smoke end-to-end.
+- Backend `backend/server.py` — dispatcher, TripoSR funcional, Hunyuan3D-2 + Trellis = stubs `501 Not Implemented`.
+- App C++ — Input/Models/Output paneles + POST + cache. Sin viewer GLB todavia.