feat: add image query tool

tools.py

@@ -182,7 +182,46 @@ def aggregate_information(results: list[str], query: str) -> str:
         print_tool_response(output_text)
 
     return output_text
-    
+
+
+gemini = ChatGoogleGenerativeAI(model="gemini-1.5-flash")
+
+
+@tool
+def image_query_tool(image_path: str, question: str) -> str:
+    """
+    Uses Gemini Vision to answer a question about an image.
+    - image_path: file path to the image to analyze (.png)
+    - question: the query to ask about the image
+    """
+    try:
+        base64_img = encode_image_to_base64(image_path)
+    except OSError:
+        response = f"OSError: Invalid argument (invalid image path or file format): {image_path}. Please provide a valid PNG image."
+        print_tool_response(response)
+        return response
+
+    base64_img_str = f"data:image/png;base64,{base64_img}"
+    if CUSTOM_DEBUG:
+        print_tool_call(
+            image_query_tool,
+            tool_name='image_query_tool',
+            args={'base64_image': base64_img_str[:100], 'question': question},
+        )
+    msg = HumanMessage(content=[
+        {"type": "text", "text": question},
+        {"type": "image_url", "image_url": base64_img_str},
+    ])
+    try:
+        response = gemini.invoke([msg])
+    except ChatGoogleGenerativeAIError:
+        response = "ChatGoogleGenerativeAIError: Invalid argument provided to Gemini: 400 Provided image is not valid"
+        print_tool_response(response)
+        return response
+    if CUSTOM_DEBUG:
+        print_tool_response(response.content)
+    return response.content
+
 
 def extract_video_id(url: str) -> str:
     parsed = urlparse(url)