From f87d4e324485d67b8f0dfef6775a87bdfb586e35 Mon Sep 17 00:00:00 2001
From: Ishan Mishra <ishan2198@hotmail.com>
Date: Tue, 19 Sep 2023 21:58:03 +0530
Subject: [PATCH] self hosted model added (#327)

* self hosted model added

* url changed

* self hosted endpoint added
---
 .../0044_alter_project_stt_engine.py          | 18 +++++++++++++
 ayushma/models/enums.py                       |  1 +
 ayushma/utils/speech_to_text.py               | 26 +++++++++++++++++++
 core/settings/base.py                         |  3 +++
 4 files changed, 48 insertions(+)
 create mode 100644 ayushma/migrations/0044_alter_project_stt_engine.py

diff --git a/ayushma/migrations/0044_alter_project_stt_engine.py b/ayushma/migrations/0044_alter_project_stt_engine.py
new file mode 100644
index 00000000..0d1a9a9d
--- /dev/null
+++ b/ayushma/migrations/0044_alter_project_stt_engine.py
@@ -0,0 +1,18 @@
+# Generated by Django 4.2.5 on 2023-09-17 12:20
+
+from django.db import migrations, models
+
+
+class Migration(migrations.Migration):
+
+    dependencies = [
+        ('ayushma', '0043_merge_20230905_1530'),
+    ]
+
+    operations = [
+        migrations.AlterField(
+            model_name='project',
+            name='stt_engine',
+            field=models.IntegerField(choices=[(1, 'Whisper'), (2, 'Google'), (3, 'Self Hosted')], default=1),
+        ),
+    ]
diff --git a/ayushma/models/enums.py b/ayushma/models/enums.py
index 6e687e13..24a73400 100644
--- a/ayushma/models/enums.py
+++ b/ayushma/models/enums.py
@@ -16,6 +16,7 @@ class DocumentType(IntegerChoices):
 class STTEngine(IntegerChoices):
     WHISPER = 1
     GOOGLE = 2
+    SELF_HOSTED = 3
 
 
 class FeedBackRating(IntegerChoices):
diff --git a/ayushma/utils/speech_to_text.py b/ayushma/utils/speech_to_text.py
index 0b24c4e7..e81f97f0 100644
--- a/ayushma/utils/speech_to_text.py
+++ b/ayushma/utils/speech_to_text.py
@@ -1,6 +1,9 @@
+import json
 import os
 
 import openai
+import requests
+from django.conf import settings
 from google.cloud import speech
 
 from ayushma.models.enums import STTEngine
@@ -48,11 +51,34 @@ def recognize(self, audio):
         if not response.results:
             return ""
         return response.results[0].alternatives[0].transcript
+    
+class SelfHostedEngine:
+    def __init__(self, api_key, language_code):
+        self.language_code = language_code
+
+    def recognize(self, audio):
+
+        response = requests.post(
+            settings.SELF_HOSTED_ENDPOINT,
+            files={"audio": audio},
+            data={
+                # change this model to get faster results see: https://github.com/coronasafe/care-whisper
+                "model": "small",
+                "language": self.language_code.replace("-IN", ""),
+            },
+        )
+
+        if not response.ok:
+            print("Failed to recognize speech with self hosted engine")
+            return ""
+        response = response.json()
+        return response["data"]["transcription"].strip()
 
 
 engines = {
     "whisper": WhisperEngine,
     "google": GoogleEngine,
+    "self_hosted": SelfHostedEngine,
     # Add new engines here
 }
 
diff --git a/core/settings/base.py b/core/settings/base.py
index 0dbb645f..e500b50b 100644
--- a/core/settings/base.py
+++ b/core/settings/base.py
@@ -368,3 +368,6 @@
 
 AI_NAME = env("AI_NAME", default="Ayushma")
 GOOGLE_RECAPTCHA_SECRET_KEY = env("GOOGLE_RECAPTCHA_SECRET_KEY", default=None)
+
+# url for self hosted speech to text
+SELF_HOSTED_ENDPOINT = env("SELF_HOSTED_ENDPOINT", default=None)