Update example to use files.py for base64 encode

Author: Replicate Staff

README.md

@@ -82,29 +82,23 @@ Some models, like [methexis-inc/img2prompt](https://replicate.com/methexis-inc/i
 To run a model that takes a file input, pass a URL to a publicly accessible file. Or, for smaller files (<10MB), you can convert file data into a base64-encoded data URI and pass that directly:
 
 ```
-# Run andreasjansson/blip-2:f677695e by providing a base64 encoded image from local file system
+# Run andreasjansson/blip-2:f677695e by providing an image from local file system which is base64 encoded by files.py
 import replicate
 import base64
 
-# Download `gg_bridge.jpeg` from https://replicate.delivery/pbxt/IJEPmgAlL2zNBNDoRRKFegTEcxnlRhoQxlNjPHSZEy0pSIKn/gg_bridge.jpeg
-# This reads example image `gg_bridge.jpeg` from current working directory into a buffer
+# This reads an example image `gg_bridge.jpeg` from current working directory into a buffer and into the prediction request 
 with open("gg_bridge.jpeg", "rb") as image_file:
-    # Then encodes the buffer into a base64 string
-    encoded_string = base64.b64encode(image_file.read()).decode("utf-8")
-    # Finally, sets the MIME type and adds the data URI scheme to the encoded string
-    encoded_string = f"data:image/jpeg;base64,{encoded_string}"
-
-output = replicate.run(
-    "andreasjansson/blip-2:f677695e5e89f8b236e52ecd1d3f01beb44c34606419bcc19345e046d8f786f9",
-    input={
-        "image": encoded_string,
-        "caption": False,
-        "question": "what body of water does this bridge cross?",
-        "temperature": 1,
-        "use_nucleus_sampling": False
-    }
-)
-print(output)
+    output = replicate.run(
+        "andreasjansson/blip-2:f677695e5e89f8b236e52ecd1d3f01beb44c34606419bcc19345e046d8f786f9",
+        input={
+            "image": image_file,
+            "caption": False,
+            "question": "what body of water does this bridge cross?",
+            "temperature": 1,
+            "use_nucleus_sampling": False
+        }
+    )
+    print(output)
 ```
 
 ## Run a model and stream its output