init r1

computer_use_demo/gradio/agent/llm_utils/groqclient.py

@@ -0,0 +1,59 @@
+from groq import Groq
+import os
+from .utils import is_image_path
+
+def run_groq_interleaved(messages: list, system: str, llm: str, api_key: str, max_tokens=256, temperature=0.6):
+    """
+    Run a chat completion through Groq's API, ignoring any images in the messages.
+    """
+    api_key = api_key or os.environ.get("GROQ_API_KEY")
+    if not api_key:
+        raise ValueError("GROQ_API_KEY is not set")
+    
+    client = Groq(api_key=api_key)
+    # avoid using system messages for R1
+    final_messages = [{"role": "user", "content": system}]
+
+    if isinstance(messages, list):
+        for item in messages:
+            if isinstance(item, dict):
+                # For dict items, concatenate all text content, ignoring images
+                text_contents = []
+                for cnt in item["content"]:
+                    if isinstance(cnt, str):
+                        if not is_image_path(cnt):  # Skip image paths
+                            text_contents.append(cnt)
+                    else:
+                        text_contents.append(str(cnt))
+                
+                if text_contents:  # Only add if there's text content
+                    message = {"role": "user", "content": " ".join(text_contents)}
+                    final_messages.append(message)
+            else:  # str
+                message = {"role": "user", "content": item}
+                final_messages.append(message)
+    
+    elif isinstance(messages, str):
+        final_messages.append({"role": "user", "content": messages})
+
+    try:
+        completion = client.chat.completions.create(
+            model="deepseek-r1-distill-llama-70b",
+            messages=final_messages,
+            temperature=0.6,
+            max_completion_tokens=max_tokens,
+            top_p=0.95,
+            stream=False,
+            reasoning_format="raw"
+        )
+        
+        response = completion.choices[0].message.content
+        final_answer = response.split('</think>\n')[-1] if '</think>' in response else response
+        final_answer = final_answer.replace("<output>", "").replace("</output>", "")
+        token_usage = completion.usage.total_tokens
+        
+        return final_answer, token_usage
+    except Exception as e:
+        print(f"Error in interleaved Groq: {e}")
+
+        return str(e), 0

computer_use_demo/gradio/agent/llm_utils/oai.py

@@ -2,21 +2,9 @@ import os
 import logging
 import base64
 import requests
-
-def is_image_path(text):
-    image_extensions = (".jpg", ".jpeg", ".png", ".gif", ".bmp", ".tiff", ".tif")
-    if text.endswith(image_extensions):
-        return True
-    else:
-        return False
-
-def encode_image(image_path):
-    """Encode image file to base64."""
-    with open(image_path, "rb") as image_file:
-        return base64.b64encode(image_file.read()).decode("utf-8")
+from .utils import is_image_path, encode_image
 
 def run_oai_interleaved(messages: list, system: str, llm: str, api_key: str, max_tokens=256, temperature=0):
-
     api_key = api_key or os.environ.get("OPENAI_API_KEY")
     if not api_key:
         raise ValueError("OPENAI_API_KEY is not set")
@@ -54,8 +42,6 @@ def run_oai_interleaved(messages: list, system: str, llm: str, api_key: str, max
     elif isinstance(messages, str):
         final_messages = [{"role": "user", "content": messages}]
 
-    print("[oai] sending messages:", {"role": "user", "content": messages})
-
     payload = {
         "model": llm,
         "messages": final_messages,

computer_use_demo/gradio/agent/llm_utils/omniparserclient.py

@@ -2,7 +2,7 @@ import requests
 import base64
 from pathlib import Path
 from tools.screen_capture import get_screenshot
-from agent.llm_utils.oai import encode_image
+from agent.llm_utils.utils import encode_image
 
 OUTPUT_DIR = "./tmp/outputs"
 

computer_use_demo/gradio/agent/llm_utils/utils.py

@@ -0,0 +1,13 @@
+import base64
+
+def is_image_path(text):
+    image_extensions = (".jpg", ".jpeg", ".png", ".gif", ".bmp", ".tiff", ".tif")
+    if text.endswith(image_extensions):
+        return True
+    else:
+        return False
+
+def encode_image(image_path):
+    """Encode image file to base64."""
+    with open(image_path, "rb") as image_file:
+        return base64.b64encode(image_file.read()).decode("utf-8")