Rename Crankycheck -> TemperCheck

Package crankycheck/ -> tempercheck/, CrankyVerdict -> TemperVerdict,
CRANKY_* env vars -> TEMPER_*, plus all UI copy, prompt branding, README
Space card, and docs. Behavior unchanged; tests pass.

Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>

CLAUDE.md

@@ -4,11 +4,11 @@ This file provides guidance to Claude Code (claude.ai/code) when working with co
 
 ## What this is
 
-**Crankycheck** — a Gradio app that takes an image or screenshot of a social-media profile and estimates how likely that person is to be cranky to deal with. Built for the Hugging Face **Build Small Hackathon** (`https://huggingface.co/build-small-hackathon/`).
+**TemperCheck** — a Gradio app that takes an image or screenshot of a social-media profile and estimates how short-tempered / difficult that person looks to deal with. Built for the Hugging Face **Build Small Hackathon** (`https://huggingface.co/build-small-hackathon/`).
 
 This is an image-in → score/verdict-out vision task. The pipeline is: **profile image → small Gemma 4 E4B vision model → structured JSON verdict → Gradio UI**.
 
-> **Status:** scaffolded and working end-to-end. `app.py` (Gradio UI) → `crankycheck/inference.py` (swappable backend) → `crankycheck/prompt.py` (prompt + defensive JSON parsing). Parsing tests pass; the Ollama backend has been smoke-tested against a real image.
+> **Status:** scaffolded; ZeroGPU Space deploy is the next step. `app.py` (Gradio UI) → `tempercheck/inference.py` (swappable backend) → `tempercheck/prompt.py` (prompt + defensive JSON parsing). Parsing tests pass. ⚠️ The vision path has NOT been validated yet — local Ollama vision is broken (see below), so the real read only works once deployed to the Space.
 
 ## Hard hackathon constraints (these gate eligibility — do not violate)
 
@@ -18,25 +18,25 @@ This is an image-in → score/verdict-out vision task. The pipeline is: **profil
 - **Deadline: June 15, 2026** (hack window June 5–15). This is a time-boxed hackathon project — prefer the simplest thing that works end-to-end over architectural polish.
 
 ### Optional bonus "merit badges" (worth steering toward when cheap)
-- **Off-Grid** — no cloud APIs; run the model locally / on-Space. Crankycheck should aim for this (local VLM inference) since it's a strong fit.
+- **Off-Grid** — no cloud APIs; run the model locally / on-Space. TemperCheck should aim for this (local VLM inference) since it's a strong fit.
 - **Well-Tuned** — fine-tuned model. **Custom UI** — bespoke Gradio styling. **Llama.cpp** — GGUF inference path. **Agent traces** shared. **Field notes** documentation.
 
 ## Environment & commands (this machine)
 
 Python is **`uv`-only** on this workstation (no system Python; `pip`/`conda` not installed). Use:
 - `uv run app.py` — run the Gradio app locally (NOT `python app.py`). Defaults to the **Ollama** backend on **port 7140**.
-- `CRANKY_BACKEND=transformers uv run app.py` — run the same backend the HF Space uses (`google/gemma-4-E4B-it` via `transformers`).
+- `TEMPER_BACKEND=transformers uv run app.py` — run the same backend the HF Space uses (`google/gemma-4-E4B-it` via `transformers`). Needs torch/transformers/spaces installed locally (not part of the default local setup).
 - `uv run pytest` — run the parsing tests. Single test: `uv run pytest tests/test_parsing.py::test_clean_json`.
 - `uv add <pkg>` / `uv add --dev <pkg>` — add a runtime / dev dependency.
 
 GPU: local **RTX 5090, 32 GB VRAM (sm_120 / Blackwell)** — the E4B model (~8B) fits trivially. The HF Space runs on smaller hardware; E4B is sized for that. Don't bump to a larger Gemma 4 (12B/26B/31B) without checking the target Space tier.
 
 ### Backend selection (the key seam)
-`CRANKY_BACKEND` switches the model path; the UI is identical either way. It
+`TEMPER_BACKEND` switches the model path; the UI is identical either way. It
 defaults to **`transformers` on a Space** (detected via the `SPACE_ID` env) and
 **`ollama` locally** — so no manual config is needed in either place.
 - **`transformers`** (the deployed Space — this is where verdicts are real) — loads `google/gemma-4-E4B-it` at module level onto `cuda` and runs generation inside a `@spaces.GPU` function (ZeroGPU). `requirements.txt` carries `spaces`/torch≥2.8/transformers; local dev does not install them.
-- **`ollama`** (local only) — POSTs to `127.0.0.1:11434`. ⚠️ **Local Gemma 4 vision is broken** in the current Ollama (0.30.7): the abliterated model hallucinates instead of reading images, and the official `gemma4:e4b` returns a blank. So the Ollama path is only good for **UI/plumbing work** — it cannot produce a real crankiness read. Real verdicts require the Space. (See `memory/abliterated-gemma4-vision-broken.md`.)
+- **`ollama`** (local only) — POSTs to `127.0.0.1:11434`. ⚠️ **Local Gemma 4 vision is broken** in the current Ollama (0.30.7): the abliterated model hallucinates instead of reading images, and the official `gemma4:e4b` returns a blank. So the Ollama path is only good for **UI/plumbing work** — it cannot produce a real temper read. Real verdicts require the Space. (See `memory/abliterated-gemma4-vision-broken.md`.)
 
 ### ZeroGPU specifics (the Space)
 - Hardware **ZeroGPU** is set in the Space *settings* (not the README). Default size `large` = 48 GB, ample for E4B.
@@ -44,11 +44,11 @@ defaults to **`transformers` on a Space** (detected via the `SPACE_ID` env) and
 - `google/gemma-4-E4B-it` is **gated** — the Space needs an `HF_TOKEN` secret whose account has accepted the Gemma license, or model download fails at boot.
 
 ### Port
-This app binds **7140** (Gradio default 7860 is triple-booked on this machine). Override with `CRANKY_PORT`. Already registered in the global port list.
+This app binds **7140** (Gradio default 7860 is triple-booked on this machine). Override with `TEMPER_PORT`. Already registered in the global port list.
 
 ## Architecture notes (load-bearing decisions)
 
-- **The VLM lives behind `crankycheck/inference.py`.** The rest of the app only calls `score_image(pil_image) -> CrankyVerdict` and never imports a backend directly. This is what lets the Ollama ↔ transformers swap (and a future llama.cpp/GGUF path for the Llama.cpp badge) happen without touching the UI.
+- **The VLM lives behind `tempercheck/inference.py`.** The rest of the app only calls `score_image(pil_image) -> TemperVerdict` and never imports a backend directly. This is what lets the Ollama ↔ transformers swap (and a future llama.cpp/GGUF path for the Llama.cpp badge) happen without touching the UI.
 - **The output contract is JSON, parsed defensively.** `prompt.py` asks for `{score, verdict, rationale, signals}`; `parse_verdict` extracts the first balanced JSON object and clamps/falls back on every field so a malformed small-VLM response never raises. If you change the JSON shape, update `parse_verdict`, the UI rendering in `app.py`, and `tests/test_parsing.py` together — that's the riskiest seam and the reason the tests exist (they run with no model).
 - **The transformers model loads once** at import (module-level, per the ZeroGPU rule), built by `_build_transformers_scorer()`; the returned `@spaces.GPU` scorer is reused per request.
 - **Gemma 4 multimodal expects the image before the text** in the message content (see `build_messages`).

README.md

@@ -1,5 +1,5 @@
 ---
-title: Crankycheck
+title: TemperCheck
 emoji: 😤
 colorFrom: yellow
 colorTo: red
@@ -7,20 +7,20 @@ sdk: gradio
 sdk_version: "6.18.0"
 app_file: app.py
 pinned: false
-short_description: How cranky does this social-media profile look to deal with?
+short_description: How short a temper does this social-media profile look to have?
 ---
 
-# 😤 Crankycheck
+# 😤 TemperCheck
 
-Upload a social-media profile (or a screenshot of one) and Crankycheck gives a
-playful read on **how cranky that person looks to deal with** — a 0–100 score, a
-punchy verdict, and the "signals" it picked up on.
+Upload a social-media profile (or a screenshot of one) and TemperCheck gives a
+playful read on **how short-tempered / difficult that person looks to deal with**
+— a 0–100 score, a punchy verdict, and the "signals" it picked up on.
 
 Built for the [Build Small Hackathon](https://huggingface.co/build-small-hackathon/)
 using a small **Gemma 4 E4B** vision-language model (~8B params, well under the
 32B limit).
 
-> ⚠️ **It's a party game.** Crankycheck reads vibes in a picture for laughs. It is
+> ⚠️ **It's a party game.** TemperCheck reads vibes in a picture for laughs. It is
 > not a real personality test and makes no factual claim about any real person.
 
 ## How it runs
@@ -39,11 +39,11 @@ so the local path is for UI work — real verdicts come from the Space.
 
 | Var | Default | Purpose |
 |-----|---------|---------|
-| `CRANKY_BACKEND` | `transformers` on a Space, else `ollama` | force a backend |
-| `CRANKY_HF_MODEL` | `google/gemma-4-E4B-it` | transformers model id |
-| `CRANKY_OLLAMA_MODEL` | `huihui_ai/gemma-4-abliterated:e4b-q8_0` | local Ollama model id |
+| `TEMPER_BACKEND` | `transformers` on a Space, else `ollama` | force a backend |
+| `TEMPER_HF_MODEL` | `google/gemma-4-E4B-it` | transformers model id |
+| `TEMPER_OLLAMA_MODEL` | `huihui_ai/gemma-4-abliterated:e4b-q8_0` | local Ollama model id |
 | `OLLAMA_HOST` | `http://127.0.0.1:11434` | local Ollama server |
-| `CRANKY_PORT` | `7140` | local Gradio port |
+| `TEMPER_PORT` | `7140` | local Gradio port |
 
 ## Tests
 
@@ -54,6 +54,6 @@ uv run pytest
 ## Project layout
 
 - `app.py` — Gradio UI.
-- `crankycheck/inference.py` — backend abstraction (Ollama ↔ transformers).
-- `crankycheck/prompt.py` — system prompt + defensive JSON parsing.
+- `tempercheck/inference.py` — backend abstraction (Ollama ↔ transformers).
+- `tempercheck/prompt.py` — system prompt + defensive JSON parsing.
 - `tests/test_parsing.py` — output-parsing tests (no model needed).

app.py

@@ -1,11 +1,11 @@
-"""Crankycheck — Gradio app.
+"""TemperCheck — Gradio app.
 
-Upload a social-media profile screenshot; get a playful "crankiness" read from a
+Upload a social-media profile screenshot; get a playful "temper" read from a
 small Gemma 4 vision model. For the Build Small Hackathon (model <= 32B, Gradio
 on Hugging Face Spaces).
 
 Run locally (Ollama backend, default):   uv run app.py
-Run the Spaces backend locally:          CRANKY_BACKEND=transformers uv run app.py
+Run the Spaces backend locally:          TEMPER_BACKEND=transformers uv run app.py
 """
 
 from __future__ import annotations
@@ -14,9 +14,9 @@ import os
 
 import gradio as gr
 
-from crankycheck import get_backend_name, score_image
+from tempercheck import get_backend_name, score_image
 
-PORT = int(os.environ.get("CRANKY_PORT", "7140"))
+PORT = int(os.environ.get("TEMPER_PORT", "7140"))
 
 # Bind real Ctrl-V: Gradio's "clipboard" source only adds a click-to-paste
 # button, so we listen for the browser paste event ourselves and feed the
@@ -26,7 +26,7 @@ PORT = int(os.environ.get("CRANKY_PORT", "7140"))
 PASTE_JS = """
 () => {
   const inject = (blob) => {
-    const root = document.querySelector('#cranky_image');
+    const root = document.querySelector('#temper_image');
     if (!root) return;
     let input = root.querySelector("input[type='file']");
     if (!input) {
@@ -56,7 +56,7 @@ PASTE_JS = """
 """
 
 DISCLAIMER = (
-    "🎈 *Crankycheck is a party-game novelty. It reads vibes in a picture for "
+    "🎈 *TemperCheck is a party-game novelty. It reads vibes in a picture for "
     "laughs — it is **not** a real personality test and says nothing factual "
     "about any actual person.*"
 )
@@ -67,7 +67,7 @@ def _render(verdict) -> tuple[str, str]:
     signals = "  ".join(f"`{s}`" for s in verdict.signals) if verdict.signals else "—"
     md = (
         f"## {verdict.band}\n\n"
-        f"**{verdict.verdict}** — crankiness **{verdict.score}/100**\n\n"
+        f"**{verdict.verdict}** — temper **{verdict.score}/100**\n\n"
         f"`{bar}`\n\n"
         f"{verdict.rationale}\n\n"
         f"**Signals:** {signals}"
@@ -87,8 +87,8 @@ def analyze(image):
         )
 
 
-with gr.Blocks(title="Crankycheck") as demo:
-    gr.Markdown("# 😤 Crankycheck\n### How cranky does this profile look to deal with?")
+with gr.Blocks(title="TemperCheck") as demo:
+    gr.Markdown("# 😤 TemperCheck\n### How short a temper does this profile look to have?")
     gr.Markdown(DISCLAIMER)
     with gr.Row():
         with gr.Column():
@@ -96,9 +96,9 @@ with gr.Blocks(title="Crankycheck") as demo:
                 type="pil",
                 label="Profile image / screenshot",
                 sources=["upload", "clipboard"],
-                elem_id="cranky_image",
+                elem_id="temper_image",
             )
-            go = gr.Button("Check the crank ☕", variant="primary")
+            go = gr.Button("Check the temper ☕", variant="primary")
         with gr.Column():
             result = gr.Markdown(label="Verdict")
             with gr.Accordion("Raw model output (agent trace)", open=False):

pyproject.toml

@@ -1,7 +1,7 @@
 [project]
-name = "crankycheck"
+name = "tempercheck"
 version = "0.1.0"
-description = "Estimate how cranky a social-media profile looks, from an image (Gemma 4 E4B)."
+description = "Estimate how short-tempered a social-media profile looks, from an image (Gemma 4 E4B)."
 readme = "README.md"
 requires-python = ">=3.12"
 dependencies = [

requirements.txt

@@ -1,5 +1,5 @@
 # Dependencies for the Hugging Face Space (ZeroGPU, transformers backend).
-# On a Space, CRANKY_BACKEND defaults to "transformers" automatically (SPACE_ID
+# On a Space, TEMPER_BACKEND defaults to "transformers" automatically (SPACE_ID
 # is set), so no env var is required. The Space also needs an HF_TOKEN secret
 # with access to the gated google/gemma-4-E4B-it repo.
 spaces

{crankycheck → tempercheck}/__init__.py

@@ -1,10 +1,10 @@
-"""Crankycheck — estimate how cranky a social-media profile looks, from an image."""
+"""TemperCheck — estimate how short-tempered a social-media profile looks, from an image."""
 
-from .prompt import CrankyVerdict, build_messages, parse_verdict
+from .prompt import TemperVerdict, build_messages, parse_verdict
 from .inference import score_image, get_backend_name
 
 __all__ = [
-    "CrankyVerdict",
+    "TemperVerdict",
     "build_messages",
     "parse_verdict",
     "score_image",

{crankycheck → tempercheck}/inference.py

@@ -1,6 +1,6 @@
-"""Inference layer for Crankycheck.
+"""Inference layer for TemperCheck.
 
-Two interchangeable backends, selected by the CRANKY_BACKEND env var (which
+Two interchangeable backends, selected by the TEMPER_BACKEND env var (which
 defaults to "transformers" on a Hugging Face Space, "ollama" elsewhere):
 
   - "transformers" (the Hugging Face Space / ZeroGPU) — loads google/gemma-4-E4B-it
@@ -26,7 +26,7 @@ from PIL import Image
 from .prompt import (
     SYSTEM_PROMPT,
     USER_INSTRUCTION,
-    CrankyVerdict,
+    TemperVerdict,
     build_messages,
     parse_verdict,
 )
@@ -34,7 +34,7 @@ from .prompt import (
 # On a Space, default to the transformers backend; locally, default to Ollama.
 _ON_SPACE = bool(os.environ.get("SPACE_ID"))
 BACKEND = os.environ.get(
-    "CRANKY_BACKEND", "transformers" if _ON_SPACE else "ollama"
+    "TEMPER_BACKEND", "transformers" if _ON_SPACE else "ollama"
 ).lower()
 
 # Use 127.0.0.1, not "localhost": on Windows the latter resolves to IPv6 ::1
@@ -42,9 +42,9 @@ BACKEND = os.environ.get(
 # was over half the total latency).
 OLLAMA_HOST = os.environ.get("OLLAMA_HOST", "http://127.0.0.1:11434")
 OLLAMA_MODEL = os.environ.get(
-    "CRANKY_OLLAMA_MODEL", "huihui_ai/gemma-4-abliterated:e4b-q8_0"
+    "TEMPER_OLLAMA_MODEL", "huihui_ai/gemma-4-abliterated:e4b-q8_0"
 )
-HF_MODEL = os.environ.get("CRANKY_HF_MODEL", "google/gemma-4-E4B-it")
+HF_MODEL = os.environ.get("TEMPER_HF_MODEL", "google/gemma-4-E4B-it")
 
 # The verdict JSON is ~80 tokens; cap generation so the model can't ramble.
 # Headroom over that keeps the JSON from ever truncating (which would break
@@ -162,7 +162,7 @@ else:
 # --- Public API -------------------------------------------------------------
 
 
-def score_image(image: Image.Image) -> CrankyVerdict:
+def score_image(image: Image.Image) -> TemperVerdict:
     """Run the configured backend on a PIL image and return a parsed verdict."""
     if image is None:
         raise ValueError("No image provided.")

{crankycheck → tempercheck}/prompt.py

@@ -1,4 +1,4 @@
-"""Prompt construction and structured-output parsing for Crankycheck.
+"""Prompt construction and structured-output parsing for TemperCheck.
 
 The model is asked to return a single JSON object. Small VLMs drift from
 requested formats, so `parse_verdict` is deliberately defensive: it extracts the
@@ -15,10 +15,10 @@ from typing import Any
 # The whole app hangs off this contract. If you change the shape, update
 # parse_verdict, the Gradio output rendering, and tests/test_parsing.py together.
 SYSTEM_PROMPT = """\
-You are Crankycheck, a playful but sharp-eyed party-game judge. Given a photo or
+You are TemperCheck, a playful but sharp-eyed party-game judge. Given a photo or
 screenshot of a social-media profile, you estimate — for entertainment only — how
-"cranky to deal with" the person seems. This is a whimsical novelty read, NOT a
-real personality assessment or a factual claim about anyone.
+short-tempered / "cranky to deal with" the person seems. This is a whimsical
+novelty read, NOT a real personality assessment or a factual claim about anyone.
 
 Judge mainly by what the person CHOSE TO PUT in their profile (bio text, display
 name, handle, captions, stated attitude); treat facial expression and overall
@@ -28,7 +28,7 @@ vibe as a lighter, secondary signal. Read the profile text carefully and look fo
    tone; insults; mocking others; "I don't care what you think" energy.
 2. PRIDE IN CONFLICT — does the profile brag about upsetting people, starting
    fights, blocking/roasting/"destroying" others, or being "brutally honest"?
-   Treating confrontation as a personality is a strong crankiness signal.
+   Treating confrontation as a personality is a strong temper signal.
 3. EXPLICIT WARNING SIGNS stated right in the profile — slurs, hateful or
    demeaning language, open anger, ALL-CAPS ranting, relentless negativity or
    doom, or harsh criticism aimed at other people.
@@ -46,7 +46,7 @@ Score guide (0 = delightful, 100 = run):
 
 Respond with ONLY a single JSON object, no prose before or after, of the form:
 {
-  "score": <integer 0-100, higher = crankier>,
+  "score": <integer 0-100, higher = shorter-tempered / crankier>,
   "verdict": "<3-6 word punchy label>",
   "rationale": "<1-2 sentences naming the specific signals you actually saw>",
   "signals": ["<short flag>", "<short flag>", "<short flag>"]
@@ -58,7 +58,7 @@ mocking how someone looks."""
 
 USER_INSTRUCTION = (
     "Read this profile — its bio/handle text first, then the overall vibe — and "
-    "return the crankiness JSON. Be playful, not mean."
+    "return the temper JSON. Be playful, not mean."
 )
 
 # 0-100 score bucket -> emoji label used by the UI.
@@ -72,7 +72,7 @@ SCORE_BANDS = [
 
 
 @dataclass
-class CrankyVerdict:
+class TemperVerdict:
     score: int
     verdict: str
     rationale: str
@@ -120,8 +120,8 @@ def _first_json_object(text: str) -> dict | None:
     return None
 
 
-def parse_verdict(text: str) -> CrankyVerdict:
-    """Parse model output into a CrankyVerdict, never raising on bad output."""
+def parse_verdict(text: str) -> TemperVerdict:
+    """Parse model output into a TemperVerdict, never raising on bad output."""
     data = _first_json_object(text) or {}
 
     # score: accept int/float/str, clamp to 0-100, default mid on failure.
@@ -138,7 +138,7 @@ def parse_verdict(text: str) -> CrankyVerdict:
         signals = [str(signals)]
     signals = [str(s).strip() for s in signals if str(s).strip()][:5]
 
-    return CrankyVerdict(
+    return TemperVerdict(
         score=score,
         verdict=str(data.get("verdict", "Inscrutable")).strip() or "Inscrutable",
         rationale=str(data.get("rationale", "The model kept its cards close."))

tests/test_parsing.py

@@ -4,7 +4,7 @@ small VLM, so it's the part worth testing without a model in the loop.
 Run: uv run pytest          (single test: uv run pytest tests/test_parsing.py::test_clean_json)
 """
 
-from crankycheck.prompt import parse_verdict
+from tempercheck.prompt import parse_verdict
 
 
 def test_clean_json():

uv.lock

@@ -241,31 +241,6 @@ wheels = [
     { url = "https://files.pythonhosted.org/packages/d1/d6/3965ed04c63042e047cb6a3e6ed1a63a35087b6a609aa3a15ed8ac56c221/colorama-0.4.6-py2.py3-none-any.whl", hash = "sha256:4f1d9991f5acc0ca119f9d443620b77f9d6b33703e51011c16baf57afb285fc6", size = 25335, upload-time = "2022-10-25T02:36:20.889Z" },
 ]
 
-[[package]]
-name = "crankycheck"
-version = "0.1.0"
-source = { virtual = "." }
-dependencies = [
-    { name = "gradio" },
-    { name = "pillow" },
-    { name = "requests" },
-]
-
-[package.dev-dependencies]
-dev = [
-    { name = "pytest" },
-]
-
-[package.metadata]
-requires-dist = [
-    { name = "gradio", specifier = ">=6.18.0" },
-    { name = "pillow", specifier = ">=12.2.0" },
-    { name = "requests", specifier = ">=2.34.2" },
-]
-
-[package.metadata.requires-dev]
-dev = [{ name = "pytest", specifier = ">=9.1.0" }]
-
 [[package]]
 name = "fastapi"
 version = "0.137.1"
@@ -1114,6 +1089,31 @@ wheels = [
     { url = "https://files.pythonhosted.org/packages/ec/bb/2799cc2ede3ed41131f8975621e7213dfc7ef4acbbaadfa440f32500c370/starlette-1.3.1-py3-none-any.whl", hash = "sha256:c7372aae11c3c3f26a42df7bd626cec2f47d03483d261d369516a615a53714c6", size = 73632, upload-time = "2026-06-12T09:23:10.017Z" },
 ]
 
+[[package]]
+name = "tempercheck"
+version = "0.1.0"
+source = { virtual = "." }
+dependencies = [
+    { name = "gradio" },
+    { name = "pillow" },
+    { name = "requests" },
+]
+
+[package.dev-dependencies]
+dev = [
+    { name = "pytest" },
+]
+
+[package.metadata]
+requires-dist = [
+    { name = "gradio", specifier = ">=6.18.0" },
+    { name = "pillow", specifier = ">=12.2.0" },
+    { name = "requests", specifier = ">=2.34.2" },
+]
+
+[package.metadata.requires-dev]
+dev = [{ name = "pytest", specifier = ">=9.1.0" }]
+
 [[package]]
 name = "tomlkit"
 version = "0.14.0"