quick flask api

10 files changed, 155 insertions, 0 deletions

diff --git a/.gitignore b/.gitignore
new file mode 100644
index 0000000..39bbdaa
--- /dev/null
+++ b/.gitignore
@@ -0,0 +1,5 @@
+__pycache__/
+.vscode/
+audio/
+models/
+venv/
diff --git a/app.py b/app.py
new file mode 100644
index 0000000..f2b470a
--- /dev/null
+++ b/app.py
@@ -0,0 +1 @@
+from dysphoria import app
diff --git a/config.py b/config.py
new file mode 100644
index 0000000..9475d5e
--- /dev/null
+++ b/config.py
@@ -0,0 +1,21 @@
+import os
+import secrets
+from os.path import join, dirname
+from dotenv import load_dotenv
+
+basedir = os.path.abspath(os.path.dirname(__file__))
+dotenv_path = join(dirname(__file__), ".env")
+load_dotenv(dotenv_path)
+
+
+class Config(object):
+    IP = os.environ.get("DYSPHORIA_IP") or "0.0.0.0"
+    PORT = os.environ.get("DYSPHORIA_PORT") or 5000
+    SECRET_KEY = os.environ.get("DYSPHORIA_SECRET_KEY") or "".join(
+        secrets.token_hex(16)
+    )
+    API_TOKEN = os.environ.get("DYSPHORIA_API_TOKEN") or "".join(secrets.token_hex(16))
+    WHISPER_MODEL_PATH = (
+        os.environ.get("DYSPHORIA_WHISPER_MODEL_PATH") or "models/whisper-large-v2-ct2/"
+    )
+    WHISPER_DEVICE = os.environ.get("DYSPHORIA_WHISPER_DEVICE") or "cpu"
diff --git a/dysphoria/__init__.py b/dysphoria/__init__.py
new file mode 100644
index 0000000..f63fc2f
--- /dev/null
+++ b/dysphoria/__init__.py
@@ -0,0 +1,9 @@
+from flask import Flask
+from config import Config
+
+app = Flask(__name__)
+app.config.from_object(Config)
+
+print(f"api token: {app.config['API_TOKEN']}")
+
+from dysphoria import routes, errors
diff --git a/dysphoria/errors.py b/dysphoria/errors.py
new file mode 100644
index 0000000..f0de77d
--- /dev/null
+++ b/dysphoria/errors.py
@@ -0,0 +1,12 @@
+from dysphoria import app
+from flask import request, jsonify
+
+
+@app.errorhandler(400)
+def _error_400(error):
+    return jsonify({"error": "bad request", "details": str(error)}), 400
+
+
+@app.errorhandler(404)
+def _error_404(error):
+    return jsonify({"error": "not found"}), 404
diff --git a/dysphoria/routes.py b/dysphoria/routes.py
new file mode 100644
index 0000000..a070497
--- /dev/null
+++ b/dysphoria/routes.py
@@ -0,0 +1,53 @@
+from dysphoria import app, whisper
+from flask import jsonify, request
+
+
+@app.route("/api/v1/health", methods=["GET"])
+def _get_health_v1():
+    return jsonify({"status": "ok"}), 200
+
+
+@app.route("/api/v1/transcribe", methods=["POST"])
+def _post_transcribe_v1():
+    # get the api token from the auth header
+    if (
+        not request.headers.get("Authorization")
+        or request.headers.get("Authorization") != f"Bearer {app.config['API_TOKEN']}"
+    ):
+        return jsonify({"error": "unauthorized"}), 401
+
+    if request.form.get("prompt"):
+        prompt = request.form.get("prompt")
+    else:
+        prompt = None
+
+    try:
+        segments, info = whisper.whisper.transcribe(
+            request.files["file"], language="en", initial_prompt=prompt
+        )
+    except:
+        return jsonify({"error": "bad request"}), 400
+
+    segments = [
+        {
+            "id": i,
+            "start": segment.start,
+            "end": segment.end,
+            "text": segment.text.strip(),
+        }
+        for i, segment in enumerate(segments)
+    ]
+    text = " ".join([segment["text"] for segment in segments])
+
+    # return the transcription
+    return (
+        jsonify(
+            {
+                "text": text,
+                "language": info.language,
+                "duration": info.duration,
+                "segments": segments,
+            }
+        ),
+        200,
+    )
diff --git a/dysphoria/whisper.py b/dysphoria/whisper.py
new file mode 100644
index 0000000..c95916f
--- /dev/null
+++ b/dysphoria/whisper.py
@@ -0,0 +1,9 @@
+from faster_whisper import WhisperModel
+from dysphoria import app
+
+whisper = WhisperModel(
+    app.config["WHISPER_MODEL_PATH"],
+    device=app.config["WHISPER_DEVICE"],
+    compute_type="default",
+    num_workers=4,
+)
diff --git a/requirements.txt b/requirements.txt
new file mode 100644
index 0000000..ebfa436
--- /dev/null
+++ b/requirements.txt
@@ -0,0 +1,5 @@
+faster-whisper[conversion]
+
+flask
+gunicorn
+python-dotenv
diff --git a/transcribe.py b/transcribe.py
new file mode 100644
index 0000000..f7e1e9d
--- /dev/null
+++ b/transcribe.py
@@ -0,0 +1,32 @@
+import time
+from faster_whisper import WhisperModel
+
+model_path = "models/whisper-tiny-en-ct2/"
+audio_files = [
+    "1670546132-2421.m4a",
+    "1670362801-1654.m4a",
+    "1675533069-1077.m4a",
+    "1675540057-3070.m4a",
+    "1677179024-2421.m4a",
+    "1677179036-2421.m4a",
+    "1677289865-3070.m4a",
+    "1677289881-3070.m4a",
+    "us.ny.monroe-1654-1678728684.m4a",
+    "us.ny.monroe-1704-1677730882.m4a",
+]
+
+t0 = time.perf_counter()
+model = WhisperModel(model_path, device="cpu", compute_type="default", num_workers=4)
+t1 = time.perf_counter()
+print(f"loaded model in {t1 - t0:0.2f} seconds")
+
+for file in audio_files:
+    t1 = time.perf_counter()
+    segments, info = model.transcribe(f"audio/{file}", language="en")
+
+    for segment in segments:
+        print("[%.2fs -> %.2fs] %s" % (segment.start, segment.end, segment.text))
+        pass
+
+    t2 = time.perf_counter()
+    print(f"transcription finished in {t2 - t1:0.2f} seconds")
diff --git a/wsgi.py b/wsgi.py
new file mode 100644
index 0000000..42482be
--- /dev/null
+++ b/wsgi.py
@@ -0,0 +1,8 @@
+"""
+Primary entry point for the app
+"""
+
+from dysphoria import app
+
+if __name__ == "__main__":
+    app.run(host=app.config["IP"], port=int(app.config["PORT"]))