Add routes for care scribe (#493)

ayushma/models/enums.py

@@ -18,6 +18,13 @@ class STTEngine(IntegerChoices):
     GOOGLE = 2
     SELF_HOSTED = 3
 
+    @classmethod
+    def get_id_from_name(cls, name):
+        for member in cls:
+            if member.name.lower() == name.lower():
+                return member.value
+        return None
+
 
 class TTSEngine(IntegerChoices):
     OPENAI = (1, "openai")

ayushma/utils/converse.py

@@ -47,7 +47,7 @@ def converse_api(
         if not open_ai_key:
             open_ai_key = (
                 request.headers.get("OpenAI-Key")
-                or (chat.project.open_ai_key)
+                or (chat.project and chat.project.openai_key)
                 or (user.allow_key and settings.OPENAI_API_KEY)
             )
         noonce = request.data.get("noonce")

ayushma/utils/openaiapi.py

@@ -311,7 +311,7 @@ def converse(
 
     prompt = chat.prompt or (chat.project and chat.project.prompt)
 
-    if documents or chat.project.model == ModelType.GPT_4_VISUAL:
+    if documents or (chat.project and chat.project.model == ModelType.GPT_4_VISUAL):
         prompt = "Image Capabilities: Enabled\n" + prompt
 
     # excluding the latest query since it is not a history
@@ -327,7 +327,7 @@ def converse(
         elif message.messageType == ChatMessageType.AYUSHMA:
             chat_history.append(AIMessage(content=f"Ayushma: {message.message}"))
 
-    tts_engine = chat.project.tts_engine
+    tts_engine = chat.project and chat.project.tts_engine
 
     if not stream:
         lang_chain_helper = LangChainHelper(

ayushma/views/orphan.py

@@ -1,27 +1,29 @@
-import time
-from types import SimpleNamespace
-
 import openai
 from django.conf import settings
-from django.http import StreamingHttpResponse
-from drf_spectacular.utils import extend_schema, extend_schema_view, inline_serializer
+from drf_spectacular.utils import extend_schema
 from rest_framework import permissions, status
 from rest_framework.decorators import action
 from rest_framework.exceptions import ValidationError
 from rest_framework.mixins import CreateModelMixin, ListModelMixin, RetrieveModelMixin
 from rest_framework.parsers import MultiPartParser
 from rest_framework.response import Response
-from rest_framework.serializers import CharField, IntegerField
 
-from ayushma.models import APIKey, Chat, ChatMessage, Project
+from ayushma.models import APIKey, Chat
+from ayushma.models.enums import STTEngine
 from ayushma.serializers import ChatDetailSerializer, ConverseSerializer
 from ayushma.utils.converse import converse_api
-from ayushma.utils.language_helpers import translate_text
-from ayushma.utils.openaiapi import converse
+from ayushma.utils.speech_to_text import speech_to_text
 from utils.views.base import BaseModelViewSet
 from utils.views.mixins import PartialUpdateModelMixin
 
-from .chat import ChatViewSet
+PREDEFINED_CONFIGS = {
+    "ai_form_fill": {
+        "model": "gpt-4-turbo-preview",
+        "response_format": "json_object",
+        "max_tokens": 4096,
+        "temperature": 0,
+    },
+}
 
 
 class Struct:
@@ -34,7 +36,7 @@ def has_permission(self, request, view):
         if request.headers.get("X-API-KEY"):
             api_key = request.headers.get("X-API-KEY")
             try:
-                key = APIKey.objects.get(key=api_key)
+                APIKey.objects.get(key=api_key)
                 return True
             except APIKey.DoesNotExist:
                 return False
@@ -82,6 +84,7 @@ def create(self, request, *args, **kwargs):
                 response = converse_api(
                     request=self.request,
                     chat=chat,
+                    is_thread=False,
                 )
                 return Response(
                     {
@@ -110,7 +113,57 @@ def converse(self, *args, **kwarg):
             response = converse_api(
                 request=self.request,
                 chat=chat,
+                is_thread=False,
             )
             return response
         except Exception as e:
             return Response({"error": str(e)}, status=status.HTTP_400_BAD_REQUEST)
+
+    @action(detail=False, methods=["post"])
+    def transcribe(self, request, *args, **kwargs):
+        language = request.data.get("language") or "en"
+        audio = request.data.get("audio")
+        engine = request.data.get("engine")
+        if not audio or not engine:
+            raise ValidationError("audio and engine are required")
+        try:
+            engine_id = STTEngine.get_id_from_name(engine)
+            transcript = speech_to_text(engine_id, audio, language + "-IN")
+            return Response({"transcript": transcript})
+        except Exception as e:
+            return Response({"error": str(e)}, status=status.HTTP_400_BAD_REQUEST)
+
+    @action(detail=False, methods=["post"])
+    def completion(self, request, *args, **kwargs):
+        task = request.data.get("task")
+
+        if not task:
+            raise ValidationError("task is required")
+
+        config = PREDEFINED_CONFIGS.get(task)
+
+        if not config:
+            raise ValidationError("Invalid task")
+
+        model = config.get("model")
+        response_format = config.get("response_format")
+        max_tokens = config.get("max_tokens")
+        temperature = config.get("temperature")
+
+        messages = request.data.get("messages")
+        if not messages or len(messages) == 0:
+            raise ValidationError("audio and engine are required")
+        try:
+            client = openai.OpenAI(api_key=settings.OPENAI_API_KEY)
+            completion = client.chat.completions.create(
+                model=model,
+                temperature=temperature,
+                response_format={"type": response_format},
+                messages=messages,
+                max_tokens=max_tokens,
+            )
+
+            ai_response = completion.choices[0].message.content
+            return Response({"response": ai_response})
+        except Exception as e:
+            return Response({"error": str(e)}, status=status.HTTP_400_BAD_REQUEST)