add nim inference

inference/nim_inference.py

@@ -0,0 +1,59 @@
+
+import requests, base64
+import os
+import json
+
+def vision_inference(image_name):
+  try:
+    invoke_url = "https://ai.api.nvidia.com/v1/gr/meta/llama-3.2-11b-vision-instruct/chat/completions"
+    stream = False
+
+    with open(image_name, "rb") as f:
+      image_b64 = base64.b64encode(f.read()).decode()
+
+    #assert len(image_b64) < 180_000, \
+    #  "To upload larger images, use the assets API (see docs)"
+      
+    api_key = os.environ["NIM_API_KEY"]
+
+    headers = {
+      "Authorization": f"Bearer {api_key}",
+      "Accept": "text/event-stream" if stream else "application/json"
+    }
+    payload = {
+      "model": 'meta/llama-3.2-11b-vision-instruct',
+      "messages": [
+        {
+          "role": "user",
+          "content": f'What is in this image? <img src="data:image/png;base64,{image_b64}" />'
+        }
+      ],
+      "max_tokens": 512,
+      "temperature": 1.00,
+      "top_p": 1.00,
+      "stream": stream
+    }
+
+    response = requests.post(invoke_url, headers=headers, json=payload)
+
+    if stream:
+        for line in response.iter_lines():
+            if line:
+                #print(line.decode("utf-8"))
+                data = line.decode("utf-8")
+                #content = json.loads(data)['choices'][0]['delta'].get('content', '') 
+    else:
+        #print(response.json())
+        data =  response.json()
+        content = data['choices'][0]['message']['content']
+
+        #print(content)
+        return content
+
+  except Exception as e:  # Added general exception handling
+        print(f"Error: {e}")
+        return None
+
+#image_name = "/home/gaganyatri/Pictures/hackathon/eat-health/fruit-stall-1.jpg"
+#content = vision_inference(image_name)
+#print(content)

recipes/engine.py

@@ -2,7 +2,7 @@ import pandas as pd
 import numpy as np
 import requests
 import json
-from .mistral_inference import text_llm
+from inference.mistral_inference import text_llm
 from django.core.files.storage import default_storage
 
 def execute_prompt(prompt, local=True):

recipes/urls.py

@@ -1,9 +1,10 @@
 from django.urls import path
 from .views import  recipe_generate_route, execute_prompt_route_get
-from .views import VisionLLMView
+from .views import VisionLLMView, NIMVisionLLMView
 
 urlpatterns = [
     path('execute_prompt_get/', execute_prompt_route_get, name='execute_prompt_get'),
     path('recipe_generate/', recipe_generate_route, name='recipe_generate'),
     path('vision_llm_url/', VisionLLMView.as_view()),
+     path('nim_vision_llm_url/', NIMVisionLLMView.as_view()),
 ]

recipes/views.py

@@ -8,6 +8,7 @@ from mistralai import Mistral
 import os
 import base64
 import json
+import requests
 
 class PromptSerializer(serializers.Serializer):
     prompt = serializers.CharField()
@@ -60,6 +61,7 @@ class VisionLLMView(APIView):
         #image_data = base64.b64decode(data['image'])
         #image_data = base64.b64decode(data['messages'][0]['image'][0])
         image_data = (data['messages'][0]['image'][0])
+        prompt =  data['messages'][0]['prompt']
 
         # Define the messages for the chat
         messages = [
@@ -68,7 +70,7 @@ class VisionLLMView(APIView):
                 "content": [
                     {
                         "type": "text",
-                        "text": data['messages'][0]['prompt']
+                        "text": prompt
                     },
                     {
                         "type": "image_url",
@@ -83,7 +85,56 @@ class VisionLLMView(APIView):
             model=model,
             messages=messages
         )
+
+        content = chat_response.choices[0].message.content
         #print(chat_response.choices[0].message.content)
         # Return the content of the response
-        return Response({"response": chat_response.choices[0].message.content})
+        return Response({"response": content})
+
 
+class NIMVisionLLMView(APIView):
+    def post(self, request, format=None):
+        try:
+            invoke_url = "https://ai.api.nvidia.com/v1/gr/meta/llama-3.2-11b-vision-instruct/chat/completions"
+            stream = False
+            api_key = os.environ["NIM_API_KEY"]
+            data = request.data
+            image_data = (data['messages'][0]['image'][0])
+            prompt =  data['messages'][0]['prompt']
+            headers = {
+            "Authorization": f"Bearer {api_key}",
+            "Accept": "text/event-stream" if stream else "application/json"
+            }
+            payload = {
+            "model": 'meta/llama-3.2-11b-vision-instruct',
+            "messages": [
+                {
+                "role": "user",
+                "content": f'{prompt} <img src="data:image/png;base64,{image_data}" />'
+                }
+            ],
+            "max_tokens": 512,
+            "temperature": 1.00,
+            "top_p": 1.00,
+            "stream": stream
+            }
+            response = requests.post(invoke_url, headers=headers, json=payload)
+
+            if stream:
+                for line in response.iter_lines():
+                    if line:
+                        #print(line.decode("utf-8"))
+                        data = line.decode("utf-8")
+                        #content = json.loads(data)['choices'][0]['delta'].get('content', '') 
+            else:
+                #print(response.json())
+                data =  response.json()
+                content = data['choices'][0]['message']['content']
+
+                #print(content)
+                return Response({"response": content})
+
+
+        except Exception as e:  # Added general exception handling
+            print(f"Error: {e}")
+        return None