v6: bump to Gradio 5 (fixes Starlette 0.40+ TemplateResponse signature)

README.md

@@ -4,12 +4,12 @@ emoji: 🌀
 colorFrom: indigo
 colorTo: blue
 sdk: gradio
-sdk_version: 4.44.1
+sdk_version: 5.5.0
 app_file: app.py
 pinned: false
 license: mit
 python_version: "3.12"
-short_description: Live llama.cpp + Hadamard rotation visualizer (TurboQuant)
+short_description: Live llama.cpp + Hadamard rotation demo (TurboQuant)
 ---
 
 # turbocpp — llama.cpp + TurboQuant
@@ -27,8 +27,9 @@ Two tabs:
 
 ## Build details
 
-- **Python 3.12** pinned (3.13 dropped stdlib `audioop` which Gradio's
-  pydub dep needs).
+- **Gradio 5** + **Python 3.12** — Gradio 4 + new Starlette is broken in
+  ways that don't resolve cleanly with version pins (TemplateResponse
+  signature change, pydantic schema change), so we just upgrade.
 - **llama-cpp-python** installed from a **prebuilt wheel** at
   [AIencoder/llama-cpp-wheels](https://huggingface.co/datasets/AIencoder/llama-cpp-wheels)
   (variant `0.3.16+basic_avx2_fma_f16c-cp312`). HF Spaces don't reliably

requirements.txt

@@ -1,21 +1,17 @@
-# Pin tested-known-good Gradio + gradio_client + Pydantic combo. Pydantic
-# 2.10 changed schema generation to emit `additionalProperties: True/False`
-# (bool) — gradio_client 1.3 / 1.4 raises `'bool' is not iterable` when
-# walking that schema. <2.10 keeps the pre-change behaviour.
-gradio==4.44.1
-gradio_client==1.3.0
-pydantic>=2.0,<2.10
+# Gradio 5 — escapes the 4.x dep-pinning rabbit hole.
+# 4.44.1 + new Starlette (0.40+) is broken: 4.x calls TemplateResponse with
+# the pre-0.40 positional-args signature, so jinja2's cache lookup gets a
+# dict as a key → TypeError. Gradio 5 uses the new Starlette API.
+gradio>=5.5,<6.0
 
 matplotlib>=3.7
 numpy>=1.24
 torch>=2.0
 pillow>=10.0
 
-# Gradio 4.x imports HfFolder which was removed in huggingface_hub 1.0.
-huggingface_hub>=0.24,<1.0
+# huggingface_hub stays <2.0 (pinning broad — Gradio 5 handles 1.x+).
+huggingface_hub>=0.24,<2.0
 
-# llama-cpp-python: prebuilt wheel from AIencoder/llama-cpp-wheels.
-# CPU-only, AVX2 + FMA + F16C. Avoids HF-Spaces source-build failures.
+# Prebuilt llama-cpp-python wheel from AIencoder/llama-cpp-wheels.
+# CPU-only, AVX2 + FMA + F16C — works on every HF Space x86_64 host.
 https://huggingface.co/datasets/AIencoder/llama-cpp-wheels/resolve/main/llama_cpp_python-0.3.16%2Bbasic_avx2_fma_f16c-cp312-cp312-manylinux_2_31_x86_64.whl
-
-audioop-lts; python_version >= "3.13"