Spaces:
Sleeping
Sleeping
Update app.py
Browse files
app.py
CHANGED
|
@@ -69,14 +69,37 @@ def smoldocling_readimage(image: Image.Image, prompt_text: str) -> str:
|
|
| 69 |
# -----------------------------
|
| 70 |
# Wrapper for MCP schema compatibility
|
| 71 |
# -----------------------------
|
| 72 |
-
def smoldocling_entry(image:
|
| 73 |
"""
|
| 74 |
-
Entry point for MCP tool.
|
| 75 |
-
|
| 76 |
-
|
| 77 |
-
|
| 78 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 79 |
"""
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 80 |
pil_image = convert_to_pil(image)
|
| 81 |
return smoldocling_readimage(pil_image, prompt_text)
|
| 82 |
|
|
|
|
| 69 |
# -----------------------------
|
| 70 |
# Wrapper for MCP schema compatibility
|
| 71 |
# -----------------------------
|
| 72 |
+
def smoldocling_entry(image: str, prompt_text: str) -> str:
|
| 73 |
"""
|
| 74 |
+
Entry point for the SmolDocling MCP tool.
|
| 75 |
+
|
| 76 |
+
Expected input formats:
|
| 77 |
+
- **Base64 string**: "data:image/png;base64,...."
|
| 78 |
+
- **Object** (Perplexity/Claude style): {"type": "image", "data": "data:image/png;base64,..."}
|
| 79 |
+
- **Local file path** (for internal testing)
|
| 80 |
+
|
| 81 |
+
Parameters
|
| 82 |
+
----------
|
| 83 |
+
image : str
|
| 84 |
+
A base64-encoded image string (with or without data: prefix) OR
|
| 85 |
+
a JSON-encoded object containing image data.
|
| 86 |
+
prompt_text : str
|
| 87 |
+
Instruction text for how to process the document (e.g., "Convert this page to docling.")
|
| 88 |
+
|
| 89 |
+
Returns
|
| 90 |
+
-------
|
| 91 |
+
str
|
| 92 |
+
Structured or textual content extracted from the image.
|
| 93 |
"""
|
| 94 |
+
# Handle Perplexity-style dicts encoded as JSON strings
|
| 95 |
+
try:
|
| 96 |
+
import json
|
| 97 |
+
maybe_json = json.loads(image)
|
| 98 |
+
if isinstance(maybe_json, dict) and "data" in maybe_json:
|
| 99 |
+
image = maybe_json
|
| 100 |
+
except Exception:
|
| 101 |
+
pass
|
| 102 |
+
|
| 103 |
pil_image = convert_to_pil(image)
|
| 104 |
return smoldocling_readimage(pil_image, prompt_text)
|
| 105 |
|