Spaces:
Sleeping
Sleeping
switch faster-whisper to CPU int8 (base) — no CUDA libs; free CPU Space
Browse files- README.md +6 -9
- app.py +15 -31
- requirements.txt +0 -2
README.md
CHANGED
|
@@ -23,12 +23,11 @@ so every quote is citable by the minute), a short **summary**, and a list of **a
|
|
| 23 |

|
| 24 |
|
| 25 |
## How it works
|
| 26 |
-
- **Transcribe** with **faster-whisper** (`
|
| 27 |
-
|
| 28 |
- **Summarize** with one **Claude** call using **forced tool-use**, so the summary + action items come
|
| 29 |
back as schema-valid JSON every time (no brittle parsing).
|
| 30 |
-
- **Two-stage pipeline:** transcription
|
| 31 |
-
the network call.
|
| 32 |
|
| 33 |
## Two tiers
|
| 34 |
- **Free (this demo)** — short clips → transcript + summary + action items.
|
|
@@ -39,14 +38,12 @@ so every quote is citable by the minute), a short **summary**, and a list of **a
|
|
| 39 |
```bash
|
| 40 |
pip install -r requirements.txt
|
| 41 |
export ANTHROPIC_API_KEY=sk-ant-... # for the summary + action items
|
| 42 |
-
python app.py # http://127.0.0.1:7860
|
| 43 |
```
|
| 44 |
-
The `spaces` import is shimmed to a no-op locally, so the UI builds without a GPU — but transcription
|
| 45 |
-
requires CUDA + faster-whisper.
|
| 46 |
|
| 47 |
## Deploy (Hugging Face Space)
|
| 48 |
-
1. Create a **Gradio** Space
|
| 49 |
-
2. Add
|
| 50 |
3. Push these files; point **notes.gritai.solutions** at the Space (CNAME → `hf.space`, DNS-only).
|
| 51 |
|
| 52 |
---
|
|
|
|
| 23 |

|
| 24 |
|
| 25 |
## How it works
|
| 26 |
+
- **Transcribe** with **faster-whisper** (`base`, int8) on **CPU** — reliable and quick for short clips,
|
| 27 |
+
no GPU/CUDA libraries required.
|
| 28 |
- **Summarize** with one **Claude** call using **forced tool-use**, so the summary + action items come
|
| 29 |
back as schema-valid JSON every time (no brittle parsing).
|
| 30 |
+
- **Two-stage pipeline:** transcription → summarization, each a clean step.
|
|
|
|
| 31 |
|
| 32 |
## Two tiers
|
| 33 |
- **Free (this demo)** — short clips → transcript + summary + action items.
|
|
|
|
| 38 |
```bash
|
| 39 |
pip install -r requirements.txt
|
| 40 |
export ANTHROPIC_API_KEY=sk-ant-... # for the summary + action items
|
| 41 |
+
python app.py # http://127.0.0.1:7860
|
| 42 |
```
|
|
|
|
|
|
|
| 43 |
|
| 44 |
## Deploy (Hugging Face Space)
|
| 45 |
+
1. Create a **Gradio** Space — **CPU basic** hardware is enough (no GPU needed).
|
| 46 |
+
2. Add secret **`ANTHROPIC_API_KEY`** for the summary + action items.
|
| 47 |
3. Push these files; point **notes.gritai.solutions** at the Space (CNAME → `hf.space`, DNS-only).
|
| 48 |
|
| 49 |
---
|
app.py
CHANGED
|
@@ -17,22 +17,9 @@ import os
|
|
| 17 |
|
| 18 |
import gradio as gr
|
| 19 |
|
| 20 |
-
|
| 21 |
-
|
| 22 |
-
|
| 23 |
-
except Exception:
|
| 24 |
-
_HAS_SPACES = False
|
| 25 |
-
|
| 26 |
-
class spaces: # local no-op shim so the UI builds without GPU
|
| 27 |
-
@staticmethod
|
| 28 |
-
def GPU(*a, **k):
|
| 29 |
-
def wrap(fn):
|
| 30 |
-
return fn
|
| 31 |
-
return wrap
|
| 32 |
-
|
| 33 |
-
|
| 34 |
-
WHISPER_SIZE = "small" # good accuracy/speed balance; float16 on ZeroGPU
|
| 35 |
-
WHISPER_REPO = "Systran/faster-whisper-small"
|
| 36 |
CLAUDE_MODEL = os.environ.get("ANTHROPIC_MODEL", "claude-haiku-4-5")
|
| 37 |
MAX_TRANSCRIPT_CHARS = 24000 # keep the summarize call comfortably in-context
|
| 38 |
|
|
@@ -40,23 +27,21 @@ _MODEL = None
|
|
| 40 |
|
| 41 |
|
| 42 |
def _get_model():
|
| 43 |
-
"""Construct the faster-whisper model on
|
| 44 |
-
only loads from the warm disk cache, not the network, inside the GPU budget."""
|
| 45 |
global _MODEL
|
| 46 |
if _MODEL is None:
|
| 47 |
from faster_whisper import WhisperModel
|
| 48 |
-
_MODEL = WhisperModel(WHISPER_SIZE, device="
|
| 49 |
return _MODEL
|
| 50 |
|
| 51 |
|
| 52 |
-
#
|
| 53 |
-
if
|
| 54 |
try:
|
| 55 |
-
|
| 56 |
-
|
| 57 |
-
print("[startup] faster-whisper weights cached; GPU attaches per request", flush=True)
|
| 58 |
except Exception as _e:
|
| 59 |
-
print(f"[startup] whisper
|
| 60 |
|
| 61 |
|
| 62 |
def _mmss(seconds):
|
|
@@ -64,15 +49,14 @@ def _mmss(seconds):
|
|
| 64 |
return f"{seconds // 60:02d}:{seconds % 60:02d}"
|
| 65 |
|
| 66 |
|
| 67 |
-
@spaces.GPU(duration=120)
|
| 68 |
def _transcribe(audio_path):
|
| 69 |
-
|
| 70 |
-
|
|
|
|
| 71 |
raise gr.Error(
|
| 72 |
-
"🎙️ Local UI preview — transcription runs on the
|
| 73 |
-
"The controls and Studio waitlist
|
| 74 |
)
|
| 75 |
-
model = _get_model()
|
| 76 |
segments, info = model.transcribe(audio_path, vad_filter=True, beam_size=5)
|
| 77 |
rows = [(seg.start, seg.end, seg.text.strip()) for seg in segments]
|
| 78 |
return rows, getattr(info, "language", "?")
|
|
|
|
| 17 |
|
| 18 |
import gradio as gr
|
| 19 |
|
| 20 |
+
ON_SPACE = bool(os.environ.get("SPACE_ID")) # True when running on a Hugging Face Space
|
| 21 |
+
|
| 22 |
+
WHISPER_SIZE = "base" # int8 on CPU — fast + reliable for short clips, no CUDA libs needed
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 23 |
CLAUDE_MODEL = os.environ.get("ANTHROPIC_MODEL", "claude-haiku-4-5")
|
| 24 |
MAX_TRANSCRIPT_CHARS = 24000 # keep the summarize call comfortably in-context
|
| 25 |
|
|
|
|
| 27 |
|
| 28 |
|
| 29 |
def _get_model():
|
| 30 |
+
"""Construct the faster-whisper model on CPU (int8), cached. Runs anywhere — no GPU / CUDA libs."""
|
|
|
|
| 31 |
global _MODEL
|
| 32 |
if _MODEL is None:
|
| 33 |
from faster_whisper import WhisperModel
|
| 34 |
+
_MODEL = WhisperModel(WHISPER_SIZE, device="cpu", compute_type="int8")
|
| 35 |
return _MODEL
|
| 36 |
|
| 37 |
|
| 38 |
+
# Warm the model at startup on the Space so the first request isn't slow.
|
| 39 |
+
if ON_SPACE:
|
| 40 |
try:
|
| 41 |
+
_get_model()
|
| 42 |
+
print("[startup] faster-whisper (base, cpu/int8) loaded", flush=True)
|
|
|
|
| 43 |
except Exception as _e:
|
| 44 |
+
print(f"[startup] whisper load failed (will retry on first call): {_e}", flush=True)
|
| 45 |
|
| 46 |
|
| 47 |
def _mmss(seconds):
|
|
|
|
| 49 |
return f"{seconds // 60:02d}:{seconds % 60:02d}"
|
| 50 |
|
| 51 |
|
|
|
|
| 52 |
def _transcribe(audio_path):
|
| 53 |
+
try:
|
| 54 |
+
model = _get_model()
|
| 55 |
+
except ImportError:
|
| 56 |
raise gr.Error(
|
| 57 |
+
"🎙️ Local UI preview — transcription runs on the Space (faster-whisper isn't installed here). "
|
| 58 |
+
"The controls and Studio waitlist are live."
|
| 59 |
)
|
|
|
|
| 60 |
segments, info = model.transcribe(audio_path, vad_filter=True, beam_size=5)
|
| 61 |
rows = [(seg.start, seg.end, seg.text.strip()) for seg in segments]
|
| 62 |
return rows, getattr(info, "language", "?")
|
requirements.txt
CHANGED
|
@@ -1,5 +1,3 @@
|
|
| 1 |
# gradio comes from the Space (sdk_version in README) — don't pin it here.
|
| 2 |
-
spaces>=0.30
|
| 3 |
-
torch
|
| 4 |
faster-whisper>=1.0
|
| 5 |
anthropic>=0.40
|
|
|
|
| 1 |
# gradio comes from the Space (sdk_version in README) — don't pin it here.
|
|
|
|
|
|
|
| 2 |
faster-whisper>=1.0
|
| 3 |
anthropic>=0.40
|