Spaces:
Sleeping
Sleeping
Upload 8 files
Browse files- any_to_image.py +179 -0
- database_operations.py +39 -0
- gemini_image_to_json.py +41 -0
- graph_visualization.py +60 -0
- main.py +282 -0
- openai_image_to_json.py +95 -0
- requirements.txt +99 -0
- utils.py +42 -0
any_to_image.py
ADDED
@@ -0,0 +1,179 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
from PIL import Image
|
2 |
+
import fitz # PyMuPDF
|
3 |
+
import os
|
4 |
+
import pyheif
|
5 |
+
|
6 |
+
|
7 |
+
def pdf_to_images(pdf_path):
|
8 |
+
# Ensure the PDF file exists
|
9 |
+
if not os.path.exists(pdf_path):
|
10 |
+
print(f"The file {pdf_path} does not exist.")
|
11 |
+
return []
|
12 |
+
|
13 |
+
# Open the PDF file
|
14 |
+
pdf_document = fitz.open(pdf_path)
|
15 |
+
|
16 |
+
# List to store PIL images
|
17 |
+
images = []
|
18 |
+
|
19 |
+
# Process each page
|
20 |
+
for page_num in range(len(pdf_document)):
|
21 |
+
# Get the page
|
22 |
+
page = pdf_document.load_page(page_num)
|
23 |
+
|
24 |
+
# Convert the page to a PIL image
|
25 |
+
pix = page.get_pixmap()
|
26 |
+
img = Image.frombytes("RGB", [pix.width, pix.height], pix.samples)
|
27 |
+
|
28 |
+
# Append the image to the list
|
29 |
+
images.append(img)
|
30 |
+
|
31 |
+
# Close the PDF document
|
32 |
+
pdf_document.close()
|
33 |
+
|
34 |
+
return images
|
35 |
+
|
36 |
+
|
37 |
+
def heic_to_image(heic_path):
|
38 |
+
|
39 |
+
# Ensure the HEIC file exists
|
40 |
+
if not os.path.exists(heic_path):
|
41 |
+
print(f"The file {heic_path} does not exist.")
|
42 |
+
return []
|
43 |
+
|
44 |
+
if heic_path.endswith(".HEIC"):
|
45 |
+
# Create the new filename by replacing .HEIC with .heic
|
46 |
+
new_file_path = heic_path[:-5] + ".heic"
|
47 |
+
# Rename the file
|
48 |
+
os.rename(heic_path, new_file_path)
|
49 |
+
print(f"Renamed: {heic_path} to {new_file_path}")
|
50 |
+
heic_path = new_file_path
|
51 |
+
|
52 |
+
try:
|
53 |
+
# Open the HEIC file
|
54 |
+
heif_file = pyheif.read(heic_path)
|
55 |
+
|
56 |
+
# Convert to a PIL image
|
57 |
+
image = Image.frombytes(
|
58 |
+
heif_file.mode,
|
59 |
+
heif_file.size,
|
60 |
+
heif_file.data,
|
61 |
+
"raw",
|
62 |
+
heif_file.mode,
|
63 |
+
heif_file.stride,
|
64 |
+
)
|
65 |
+
except Exception as e:
|
66 |
+
print(f"An error occurred while processing the HEIC file: {e}")
|
67 |
+
return []
|
68 |
+
|
69 |
+
return image
|
70 |
+
|
71 |
+
|
72 |
+
def process_image(file_path):
|
73 |
+
|
74 |
+
if file_path.endswith(".pdf"):
|
75 |
+
|
76 |
+
images = pdf_to_images(file_path)
|
77 |
+
|
78 |
+
# Save the images so we can pass their path to the rest of the gradio functions
|
79 |
+
if images:
|
80 |
+
output_dir = "data_processed"
|
81 |
+
os.makedirs(output_dir, exist_ok=True)
|
82 |
+
image_paths = []
|
83 |
+
label_prefix = file_path.split(os.sep)[-1].split(".")[0]
|
84 |
+
|
85 |
+
for i, img in enumerate(images, start=1):
|
86 |
+
image_path = os.path.join(output_dir, f"{label_prefix}_page_{i}.png")
|
87 |
+
img.save(image_path, "PNG")
|
88 |
+
# saving it back to the same path assigned by gradio so that we can benefit from gradio's cache
|
89 |
+
# replace the .pdf with .png in the file_path
|
90 |
+
file_path = file_path.replace(".pdf", ".png")
|
91 |
+
# img.save(file_path, "PNG")
|
92 |
+
# image_paths.append(image_path)
|
93 |
+
image_paths.append(file_path)
|
94 |
+
print(f"Saved {image_path}")
|
95 |
+
print(f"Saved {file_path}")
|
96 |
+
# saving the first image of the pdf only to be processed in the gradio UI.
|
97 |
+
# TODO: Accomodate for multiple images
|
98 |
+
images[0].save(file_path, "PNG")
|
99 |
+
|
100 |
+
return file_path, ""
|
101 |
+
|
102 |
+
# return image_paths[0], ""
|
103 |
+
else:
|
104 |
+
return None, "No image uploaded or invalid file"
|
105 |
+
|
106 |
+
elif (
|
107 |
+
file_path.endswith(".png")
|
108 |
+
or file_path.endswith(".jpg")
|
109 |
+
or file_path.endswith(".jpeg")
|
110 |
+
):
|
111 |
+
print(
|
112 |
+
f"file_path from the image processing function for compatible images: {file_path}"
|
113 |
+
)
|
114 |
+
return file_path, ""
|
115 |
+
|
116 |
+
elif file_path.endswith(".JPEG"):
|
117 |
+
# Create the new filename by replacing .JPEG with .jpeg
|
118 |
+
new_file_path = file_path[:-5] + ".jpeg"
|
119 |
+
# Rename the file
|
120 |
+
os.rename(file_path, new_file_path)
|
121 |
+
print(f"Renamed: {file_path} to {new_file_path}")
|
122 |
+
file_path = new_file_path
|
123 |
+
return file_path, ""
|
124 |
+
|
125 |
+
elif file_path.endswith(".JPG"):
|
126 |
+
# Create the new filename by replacing .JPG with .jpg
|
127 |
+
new_file_path = file_path[:-4] + ".jpg"
|
128 |
+
# Rename the file
|
129 |
+
os.rename(file_path, new_file_path)
|
130 |
+
print(f"Renamed: {file_path} to {new_file_path}")
|
131 |
+
file_path = new_file_path
|
132 |
+
return file_path, ""
|
133 |
+
|
134 |
+
elif file_path.endswith(".PNG"):
|
135 |
+
# Create the new filename by replacing .PNG with .png
|
136 |
+
new_file_path = file_path[:-4] + ".png"
|
137 |
+
# Rename the file
|
138 |
+
os.rename(file_path, new_file_path)
|
139 |
+
print(f"Renamed: {file_path} to {new_file_path}")
|
140 |
+
file_path = new_file_path
|
141 |
+
return file_path, ""
|
142 |
+
|
143 |
+
elif file_path.endswith(".heic") or file_path.endswith(".HEIC"):
|
144 |
+
image = heic_to_image(file_path)
|
145 |
+
|
146 |
+
output_dir = "data_processed"
|
147 |
+
os.makedirs(output_dir, exist_ok=True)
|
148 |
+
|
149 |
+
image_path = os.path.join(
|
150 |
+
output_dir, f"{os.path.splitext(os.path.basename(file_path))[0]}.png"
|
151 |
+
)
|
152 |
+
image.save(image_path, "PNG")
|
153 |
+
# saving it back to the same path assigned by gradio so that we can benefit from gradio's cache
|
154 |
+
image.save(file_path, "PNG")
|
155 |
+
|
156 |
+
print(f"Saved {image_path}")
|
157 |
+
print(f"Saved {file_path}")
|
158 |
+
|
159 |
+
# return image_path, ""
|
160 |
+
return file_path, ""
|
161 |
+
|
162 |
+
else:
|
163 |
+
return None, "No image uploaded or invalid file"
|
164 |
+
|
165 |
+
|
166 |
+
# Example usage
|
167 |
+
if __name__ == "__main__":
|
168 |
+
pdf_path = "data/Augustin REMY Mindmap OOP .pdf"
|
169 |
+
images = pdf_to_images(pdf_path)
|
170 |
+
|
171 |
+
# Example: Save the images if you want to check them
|
172 |
+
if images:
|
173 |
+
output_dir = "data_processed"
|
174 |
+
os.makedirs(output_dir, exist_ok=True)
|
175 |
+
|
176 |
+
for i, img in enumerate(images, start=1):
|
177 |
+
image_path = os.path.join(output_dir, f"page_{i}.png")
|
178 |
+
img.save(image_path, "PNG")
|
179 |
+
print(f"Saved {image_path}")
|
database_operations.py
ADDED
@@ -0,0 +1,39 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
from neo4j import GraphDatabase
|
2 |
+
|
3 |
+
class Neo4jDatabase:
|
4 |
+
def __init__(self, uri, username, password):
|
5 |
+
self.driver = GraphDatabase.driver(uri, auth=(username, password))
|
6 |
+
|
7 |
+
def close(self):
|
8 |
+
self.driver.close()
|
9 |
+
|
10 |
+
def dump_to_neo4j(self, nodes, edges, label_prefix):
|
11 |
+
with self.driver.session() as session:
|
12 |
+
for node in nodes:
|
13 |
+
session.run(f"CREATE (n:{label_prefix}:Node {{id: $id, label: $label}})", id=node['id'], label=node['label'])
|
14 |
+
|
15 |
+
for edge in edges:
|
16 |
+
session.run(f"""
|
17 |
+
MATCH (a:{label_prefix}:Node {{id: $source}}), (b:{label_prefix}:Node {{id: $target}})
|
18 |
+
CREATE (a)-[r:RELATION {{type: $type}}]->(b)
|
19 |
+
""", source=edge['source'], target=edge['target'], type=edge['type'])
|
20 |
+
|
21 |
+
def check_existing_graph(self, label_prefix):
|
22 |
+
with self.driver.session() as session:
|
23 |
+
result = session.run(f"MATCH (n:{label_prefix}) RETURN count(n) as count")
|
24 |
+
count = result.single()["count"]
|
25 |
+
return count > 0
|
26 |
+
|
27 |
+
def get_graph_data(self, label_prefix):
|
28 |
+
with self.driver.session() as session:
|
29 |
+
nodes = session.run(f"MATCH (n:{label_prefix}) RETURN n.id AS id, n.label AS label")
|
30 |
+
edges = session.run(f"MATCH (a:{label_prefix})-[r]->(b:{label_prefix}) RETURN a.id AS source, b.id AS target, type(r) AS type")
|
31 |
+
|
32 |
+
nodes = [{"id": record["id"], "label": record["label"]} for record in nodes]
|
33 |
+
edges = [{"source": record["source"], "target": record["target"], "type": record["type"]} for record in edges]
|
34 |
+
|
35 |
+
return {"nodes": nodes, "edges": edges}
|
36 |
+
|
37 |
+
def delete_graph(self, label_prefix):
|
38 |
+
with self.driver.session() as session:
|
39 |
+
session.run(f"MATCH (n:{label_prefix}) DETACH DELETE n")
|
gemini_image_to_json.py
ADDED
@@ -0,0 +1,41 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
import google.generativeai as genai
|
2 |
+
|
3 |
+
from dotenv import load_dotenv
|
4 |
+
import os
|
5 |
+
|
6 |
+
load_dotenv()
|
7 |
+
GOOGLE_API_KEY = os.getenv("GOOGLE_API_KEY")
|
8 |
+
|
9 |
+
genai.configure(api_key=GOOGLE_API_KEY)
|
10 |
+
|
11 |
+
# gemini-1.5-pro only gives 50 requests per day. check https://ai.google.dev/pricing for more details
|
12 |
+
# model = genai.GenerativeModel('gemini-1.5-pro',
|
13 |
+
model = genai.GenerativeModel(
|
14 |
+
"gemini-1.5-flash",
|
15 |
+
# Set the `response_mime_type` to output JSON
|
16 |
+
# Pass the schema object to the `response_schema` field
|
17 |
+
generation_config={
|
18 |
+
"response_mime_type": "application/json",
|
19 |
+
"temperature": 0.0,
|
20 |
+
},
|
21 |
+
)
|
22 |
+
# "response_schema": Recipe, 'max_output_tokens':4000})
|
23 |
+
|
24 |
+
PROMPT = """
|
25 |
+
You are responsible for extracting the entities (nodes) and relationships (edges) from the images of mind maps. The mind maps are for Object Oriented Programming.
|
26 |
+
Don't make up facts, just extracts them. Do not create new entity types that aren't mentioned in the image, and at the same time don't miss anything.
|
27 |
+
Give the output in JSON format with this schema:
|
28 |
+
{
|
29 |
+
"nodes": [{"id": "1", "label": string},{"id": "2", "label": string}],"edges": [{"source": SOURCE_ID, "target": TARGET_ID, "type": "->"},{"source": SOURCE_ID, "target": TARGET_ID, "type": "->"}]
|
30 |
+
}
|
31 |
+
|
32 |
+
Now extract the entities and relationships from this image:
|
33 |
+
"""
|
34 |
+
|
35 |
+
|
36 |
+
def fetch_gemini_response(mind_map_image):
|
37 |
+
|
38 |
+
print("fetching gemini response")
|
39 |
+
response = model.generate_content([PROMPT, mind_map_image], stream=False)
|
40 |
+
|
41 |
+
return response.text
|
graph_visualization.py
ADDED
@@ -0,0 +1,60 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
import json
|
2 |
+
from pyvis.network import Network
|
3 |
+
|
4 |
+
def create_graph(nodes, edges, physics_enabled=True):
|
5 |
+
net = Network(notebook=True, height='100vh', width='100vw', bgcolor='#222222', font_color='white', cdn_resources='remote')
|
6 |
+
|
7 |
+
for node in nodes:
|
8 |
+
net.add_node(
|
9 |
+
node['id'],
|
10 |
+
label=node['label'],
|
11 |
+
title=node['label'],
|
12 |
+
color='blue' if node['label'] == 'OOP' else 'green'
|
13 |
+
)
|
14 |
+
|
15 |
+
for edge in edges:
|
16 |
+
net.add_edge(edge['source'], edge['target'], title=edge['type'])
|
17 |
+
|
18 |
+
net.force_atlas_2based(
|
19 |
+
gravity=-50,
|
20 |
+
central_gravity=0.01,
|
21 |
+
spring_length=100,
|
22 |
+
spring_strength=0.08,
|
23 |
+
damping=0.4
|
24 |
+
)
|
25 |
+
|
26 |
+
options = {
|
27 |
+
"nodes": {
|
28 |
+
"physics": physics_enabled
|
29 |
+
},
|
30 |
+
"edges": {
|
31 |
+
"smooth": True
|
32 |
+
},
|
33 |
+
"interaction": {
|
34 |
+
"hover": True,
|
35 |
+
"zoomView": True
|
36 |
+
},
|
37 |
+
"physics": {
|
38 |
+
"enabled": physics_enabled,
|
39 |
+
"stabilization": {
|
40 |
+
"enabled": True,
|
41 |
+
"iterations": 200
|
42 |
+
}
|
43 |
+
}
|
44 |
+
}
|
45 |
+
|
46 |
+
net.set_options(json.dumps(options))
|
47 |
+
return net
|
48 |
+
|
49 |
+
def visualize_graph(json_data, physics_enabled=True):
|
50 |
+
if isinstance(json_data, str):
|
51 |
+
data = json.loads(json_data)
|
52 |
+
else:
|
53 |
+
data = json_data
|
54 |
+
nodes = data['nodes']
|
55 |
+
edges = data['edges']
|
56 |
+
net = create_graph(nodes, edges, physics_enabled)
|
57 |
+
html = net.generate_html()
|
58 |
+
html = html.replace("'", "\"")
|
59 |
+
html = html.replace('<div id="mynetwork"', '<div id="mynetwork" style="height: 100vh; width: 100%;"')
|
60 |
+
return f"""<iframe style="width: 100%; height: 100vh; border: none; margin: 0; padding: 0;" srcdoc='{html}'></iframe>"""
|
main.py
ADDED
@@ -0,0 +1,282 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
import gradio as gr
|
2 |
+
import json
|
3 |
+
import os
|
4 |
+
from PIL import Image
|
5 |
+
from database_operations import Neo4jDatabase
|
6 |
+
from graph_visualization import visualize_graph
|
7 |
+
from utils import extract_label_prefix, strip_keys, format_json, validate_json
|
8 |
+
from models.gemini_image_to_json import fetch_gemini_response
|
9 |
+
from models.openai_image_to_json import openaiprocess_image_to_json
|
10 |
+
from any_to_image import pdf_to_images, process_image
|
11 |
+
|
12 |
+
# Initialize Neo4j database
|
13 |
+
db = Neo4jDatabase("bolt://localhost:7687", "neo4j", "password123")
|
14 |
+
|
15 |
+
def dump_to_neo4j_with_confirmation(json_content, file_path, history, previous_states):
|
16 |
+
if not file_path:
|
17 |
+
return "No image uploaded or invalid file", history, previous_states, None
|
18 |
+
|
19 |
+
try:
|
20 |
+
json_data = json.loads(json_content)
|
21 |
+
except json.JSONDecodeError:
|
22 |
+
return "Invalid JSON data. Please check your input.", history, previous_states, None
|
23 |
+
|
24 |
+
label_prefix = extract_label_prefix(file_path)
|
25 |
+
|
26 |
+
if db.check_existing_graph(label_prefix):
|
27 |
+
previous_state = db.get_graph_data(label_prefix)
|
28 |
+
return f"A graph with label prefix '{label_prefix}' already exists in the database. Do you want to overwrite it?", history, previous_states, label_prefix
|
29 |
+
else:
|
30 |
+
json_data = strip_keys(json_data)
|
31 |
+
db.dump_to_neo4j(json_data['nodes'], json_data['edges'], label_prefix)
|
32 |
+
result = f"Data successfully dumped into the database with label prefix '{label_prefix}'."
|
33 |
+
new_history = f"{history}\n[NEW ENTRY] {result}" if history else f"[NEW ENTRY] {result}"
|
34 |
+
previous_states[label_prefix] = []
|
35 |
+
return result, new_history, previous_states, None
|
36 |
+
|
37 |
+
def confirm_overwrite(confirmation, gradio_state, json_content, file_path, history, previous_states):
|
38 |
+
if confirmation.lower() == 'yes':
|
39 |
+
try:
|
40 |
+
label_prefix = extract_label_prefix(file_path)
|
41 |
+
previous_state = db.get_graph_data(label_prefix)
|
42 |
+
# print(f'previous_state from the confirm_overwrite function: {previous_state}')
|
43 |
+
# print(f'label_prefix from the confirm_overwrite function: {label_prefix}')
|
44 |
+
# print(f'previouse_states from the confirm_overwrite function: {previous_states}')
|
45 |
+
|
46 |
+
if label_prefix not in previous_states:
|
47 |
+
previous_states[label_prefix] = []
|
48 |
+
previous_states[label_prefix].append(previous_state)
|
49 |
+
else:
|
50 |
+
previous_states[label_prefix].append(previous_state)
|
51 |
+
|
52 |
+
if len(previous_states[label_prefix]) > 3:
|
53 |
+
previous_states[label_prefix] = previous_states[label_prefix][-3:]
|
54 |
+
|
55 |
+
db.delete_graph(label_prefix)
|
56 |
+
|
57 |
+
json_data = json.loads(json_content)
|
58 |
+
json_data = strip_keys(json_data)
|
59 |
+
db.dump_to_neo4j(json_data['nodes'], json_data['edges'], label_prefix)
|
60 |
+
result = f"Data successfully overwritten in the database with label prefix '{label_prefix}'."
|
61 |
+
new_history = f"{history}\n[OVERWRITE] {result}" if history else f"[OVERWRITE] {result}"
|
62 |
+
return result, new_history, previous_states, ""
|
63 |
+
except json.JSONDecodeError:
|
64 |
+
return "Invalid JSON data. Please check your input.", history, previous_states, ""
|
65 |
+
else:
|
66 |
+
return "Operation cancelled. The existing graph was not overwritten.", history, previous_states, ""
|
67 |
+
|
68 |
+
def revert_last_action(history, previous_states):
|
69 |
+
if not history:
|
70 |
+
return "No actions to revert.", history, previous_states
|
71 |
+
|
72 |
+
last_action = history.split('\n')[-1]
|
73 |
+
label_prefix = last_action.split("'")[1]
|
74 |
+
|
75 |
+
if label_prefix in previous_states and previous_states[label_prefix]:
|
76 |
+
db.delete_graph(label_prefix)
|
77 |
+
db.dump_to_neo4j(previous_states[label_prefix][-1]['nodes'], previous_states[label_prefix][-1]['edges'], label_prefix)
|
78 |
+
new_history = history + f"\n[REVERT] Reverted overwrite of graph with label prefix '{label_prefix}'"
|
79 |
+
previous_states[label_prefix].pop()
|
80 |
+
return f"Reverted last action: {last_action}", new_history, previous_states
|
81 |
+
elif label_prefix in previous_states and not previous_states[label_prefix]:
|
82 |
+
db.delete_graph(label_prefix)
|
83 |
+
new_history = history + f"\n[REVERT] Deleted newly added graph with label prefix '{label_prefix}'"
|
84 |
+
del previous_states[label_prefix]
|
85 |
+
return f"Reverted last action: {last_action}", new_history, previous_states
|
86 |
+
else:
|
87 |
+
return "Unable to revert the last action.", history, previous_states
|
88 |
+
|
89 |
+
def update_graph_from_edited_json(json_content, physics_enabled):
|
90 |
+
try:
|
91 |
+
json_data = json.loads(json_content)
|
92 |
+
json_data = strip_keys(json_data)
|
93 |
+
validate_json(json_data)
|
94 |
+
return visualize_graph(json_data, physics_enabled), ""
|
95 |
+
except json.JSONDecodeError as e:
|
96 |
+
return None, f"Invalid JSON format: {str(e)}"
|
97 |
+
except ValueError as e:
|
98 |
+
return None, f"Invalid graph structure: {str(e)}"
|
99 |
+
except Exception as e:
|
100 |
+
return None, f"An unexpected error occurred: {str(e)}"
|
101 |
+
|
102 |
+
def fetch_kg(image_file_path, model_choice_state):
|
103 |
+
if image_file_path:
|
104 |
+
mind_map_image = Image.open(image_file_path)
|
105 |
+
|
106 |
+
if model_choice_state == 'Gemini':
|
107 |
+
print(f'model choice is gemini')
|
108 |
+
kg_json_text = fetch_gemini_response(mind_map_image)
|
109 |
+
elif model_choice_state == 'OpenAI':
|
110 |
+
print(f'model choice is openai')
|
111 |
+
kg_json_text = openaiprocess_image_to_json(mind_map_image)
|
112 |
+
|
113 |
+
json_data = json.loads(kg_json_text)
|
114 |
+
return format_json(json_data), ""
|
115 |
+
return "", "No image uploaded or invalid file"
|
116 |
+
|
117 |
+
def input_file_handler(file_path):
|
118 |
+
if file_path:
|
119 |
+
image_path, error = process_image(file_path)
|
120 |
+
return image_path, error
|
121 |
+
|
122 |
+
return "", "No image uploaded or invalid file"
|
123 |
+
|
124 |
+
# Gradio interface
|
125 |
+
with gr.Blocks() as demo:
|
126 |
+
gr.Markdown("## Image to Knowledge Graph Transformation")
|
127 |
+
|
128 |
+
with gr.Row():
|
129 |
+
file_input = gr.File(label="Upload File", file_count="single",
|
130 |
+
type="filepath",
|
131 |
+
file_types=[".pdf", ".png", ".jpeg", ".jpg", ".heic"])
|
132 |
+
image_file = gr.Image(label="Input Image", type="filepath", visible=False)
|
133 |
+
json_editor = gr.Textbox(label="Edit JSON", lines=15, placeholder="JSON data will appear here after image upload")
|
134 |
+
|
135 |
+
with gr.Row():
|
136 |
+
with gr.Column():
|
137 |
+
with gr.Row():
|
138 |
+
CCW_rotate_button = gr.Button('Rotate Image Counter-Clockwise')
|
139 |
+
CW_rotate_button = gr.Button('Rotate Image Clockwise')
|
140 |
+
with gr.Column():
|
141 |
+
model_call = gr.Button('Transform Image into KG representation', scale=2)
|
142 |
+
with gr.Row():
|
143 |
+
physics_button = gr.Checkbox(value=True, label="Enable Graph Physics")
|
144 |
+
model_choice = gr.Radio(label="Select Model", choices=["OpenAI", "Gemini"], value="Gemini", interactive=True)
|
145 |
+
|
146 |
+
graph_output = gr.HTML(label="Graph Output")
|
147 |
+
error_output = gr.Textbox(label="Error Messages", interactive=False)
|
148 |
+
|
149 |
+
update_button = gr.Button("Update Graph")
|
150 |
+
dump_button = gr.Button("Dump to Neo4j")
|
151 |
+
revert_button = gr.Button("Revert Last Action")
|
152 |
+
|
153 |
+
history_block = gr.Textbox(label="History", placeholder="Graphs pushed to the Database", interactive=False, lines=5, max_lines=50)
|
154 |
+
history_state = gr.State("")
|
155 |
+
previous_states = gr.State({})
|
156 |
+
|
157 |
+
confirmation_output = gr.Textbox(label="Confirmation Message", visible=False, interactive=False)
|
158 |
+
confirmation_input = gr.Textbox(label="Type 'yes' to confirm overwrite", visible=False, interactive=True)
|
159 |
+
confirm_button = gr.Button("Confirm Overwrite", visible=False)
|
160 |
+
|
161 |
+
file_input.upload(
|
162 |
+
fn=input_file_handler,
|
163 |
+
inputs=[file_input],
|
164 |
+
outputs=[image_file, error_output]
|
165 |
+
).then(
|
166 |
+
lambda image_file: (
|
167 |
+
gr.Image(value=image_file, visible=True),
|
168 |
+
gr.File(visible=False)
|
169 |
+
),
|
170 |
+
inputs=[image_file],
|
171 |
+
outputs=[image_file, file_input]
|
172 |
+
)
|
173 |
+
|
174 |
+
image_file.clear(
|
175 |
+
lambda file_input, image_file: (
|
176 |
+
gr.File(visible=True),
|
177 |
+
gr.Image(visible=False)
|
178 |
+
),
|
179 |
+
inputs=[file_input, image_file],
|
180 |
+
outputs=[file_input, image_file]
|
181 |
+
)
|
182 |
+
|
183 |
+
def rotate_image_to_left(image_path):
|
184 |
+
if image_path:
|
185 |
+
image = Image.open(image_path)
|
186 |
+
image = image.rotate(-90, expand=True)
|
187 |
+
image.save(image_path)
|
188 |
+
return image_path
|
189 |
+
|
190 |
+
CW_rotate_button.click(
|
191 |
+
fn=rotate_image_to_left,
|
192 |
+
inputs=[image_file],
|
193 |
+
outputs=[image_file]
|
194 |
+
)
|
195 |
+
|
196 |
+
def rotate_image_to_right(image_path):
|
197 |
+
if image_path:
|
198 |
+
image = Image.open(image_path)
|
199 |
+
image = image.rotate(90, expand=True)
|
200 |
+
image.save(image_path)
|
201 |
+
return image_path
|
202 |
+
|
203 |
+
CCW_rotate_button.click(
|
204 |
+
fn=rotate_image_to_right,
|
205 |
+
inputs=[image_file],
|
206 |
+
outputs=[image_file]
|
207 |
+
)
|
208 |
+
|
209 |
+
dump_button.click(
|
210 |
+
dump_to_neo4j_with_confirmation,
|
211 |
+
inputs=[json_editor, image_file, history_state, previous_states],
|
212 |
+
outputs=[confirmation_output, history_state, previous_states, gr.State()]
|
213 |
+
).then(
|
214 |
+
lambda message, history, previous_states, label_prefix: (
|
215 |
+
gr.Textbox(value=message, visible=True),
|
216 |
+
gr.Textbox(visible=True),
|
217 |
+
gr.Button(visible=True),
|
218 |
+
history,
|
219 |
+
previous_states,
|
220 |
+
label_prefix
|
221 |
+
),
|
222 |
+
inputs=[confirmation_output, history_state, previous_states, gr.State()],
|
223 |
+
outputs=[confirmation_output, confirmation_input, confirm_button, history_state, previous_states, gr.State()]
|
224 |
+
).then(
|
225 |
+
lambda history: history,
|
226 |
+
inputs=[history_state],
|
227 |
+
outputs=[history_block]
|
228 |
+
)
|
229 |
+
|
230 |
+
gr.on(
|
231 |
+
triggers=[confirm_button.click, confirmation_input.submit],
|
232 |
+
fn=confirm_overwrite,
|
233 |
+
inputs=[confirmation_input, gr.State(), json_editor, image_file, history_state, previous_states],
|
234 |
+
outputs=[confirmation_output, history_state, previous_states, confirmation_input]
|
235 |
+
).then(
|
236 |
+
lambda confirmation_output, confirmation_input: (
|
237 |
+
gr.Textbox(value=confirmation_output, visible=True),
|
238 |
+
gr.Textbox(value='', visible=False),
|
239 |
+
gr.Button(visible=False)
|
240 |
+
),
|
241 |
+
inputs=[confirmation_output, confirmation_input],
|
242 |
+
outputs=[confirmation_output, confirmation_input, confirm_button]
|
243 |
+
).then(
|
244 |
+
lambda history: history,
|
245 |
+
inputs=[history_state],
|
246 |
+
outputs=[history_block]
|
247 |
+
)
|
248 |
+
|
249 |
+
revert_button.click(
|
250 |
+
revert_last_action,
|
251 |
+
inputs=[history_state, previous_states],
|
252 |
+
outputs=[confirmation_output, history_state, previous_states]
|
253 |
+
).then(
|
254 |
+
lambda confirmation_output: gr.Textbox(value=confirmation_output, visible=True),
|
255 |
+
inputs=[confirmation_output],
|
256 |
+
outputs=[confirmation_output]
|
257 |
+
).then(
|
258 |
+
lambda history: history,
|
259 |
+
inputs=[history_state],
|
260 |
+
outputs=[history_block]
|
261 |
+
)
|
262 |
+
|
263 |
+
update_button.click(
|
264 |
+
update_graph_from_edited_json,
|
265 |
+
inputs=[json_editor, physics_button],
|
266 |
+
outputs=[graph_output, error_output]
|
267 |
+
)
|
268 |
+
|
269 |
+
physics_button.change(
|
270 |
+
update_graph_from_edited_json,
|
271 |
+
inputs=[json_editor, physics_button],
|
272 |
+
outputs=[graph_output, error_output]
|
273 |
+
)
|
274 |
+
|
275 |
+
model_call.click(
|
276 |
+
fn=fetch_kg,
|
277 |
+
inputs=[image_file, model_choice],
|
278 |
+
outputs=[json_editor, error_output]
|
279 |
+
)
|
280 |
+
|
281 |
+
if __name__ == "__main__":
|
282 |
+
demo.launch()
|
openai_image_to_json.py
ADDED
@@ -0,0 +1,95 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
import base64
|
2 |
+
import requests
|
3 |
+
from io import BytesIO
|
4 |
+
from PIL import Image
|
5 |
+
import os
|
6 |
+
from dotenv import load_dotenv
|
7 |
+
import json
|
8 |
+
|
9 |
+
# Load the .env file
|
10 |
+
load_dotenv()
|
11 |
+
|
12 |
+
# Get the API key from the environment
|
13 |
+
api_key = os.getenv('OPENAI_API_KEY')
|
14 |
+
|
15 |
+
# Function to encode the image
|
16 |
+
def encode_image(image):
|
17 |
+
# Convert the image to RGB if it has an alpha channel
|
18 |
+
if image.mode == 'RGBA':
|
19 |
+
image = image.convert('RGB')
|
20 |
+
|
21 |
+
buffered = BytesIO()
|
22 |
+
image.save(buffered, format="JPEG")
|
23 |
+
return base64.b64encode(buffered.getvalue()).decode('utf-8')
|
24 |
+
|
25 |
+
def openaiprocess_image_to_json(image):
|
26 |
+
|
27 |
+
print(f'fetching openai response')
|
28 |
+
|
29 |
+
# Encode the image
|
30 |
+
base64_image = encode_image(image)
|
31 |
+
|
32 |
+
headers = {
|
33 |
+
"Content-Type": "application/json",
|
34 |
+
"Authorization": f"Bearer {api_key}"
|
35 |
+
}
|
36 |
+
|
37 |
+
PROMPT = '''
|
38 |
+
You are responsible for extracting the entities (nodes) and relationships (edges) from the images of mind maps. The mind maps are for Object Oriented Programming.
|
39 |
+
Don't make up facts, just extracts them. Do not create new entity types that aren't mentioned in the image, and at the same time don't miss anything.
|
40 |
+
Give the output in JSON format as follows:
|
41 |
+
{
|
42 |
+
"nodes": [
|
43 |
+
{"id": "1", "label": string},
|
44 |
+
{"id": "2", "label": string},...
|
45 |
+
],
|
46 |
+
"edges": [
|
47 |
+
{"source": SOURCE_ID, "target": TARGET_ID, "type": "->"},
|
48 |
+
{"source": SOURCE_ID, "target": TARGET_ID, "type": "->"},...
|
49 |
+
]
|
50 |
+
}
|
51 |
+
Only return valid python dictionary, dont include (line jump)n in it, dont include spaces, only a dictionary. Do not include any other text outside the Dictionary structure. Make sure that i will get a valid Python dictionary.
|
52 |
+
make sure that what you return as json_string i can use it in python in this function: json.loads(json_string)
|
53 |
+
Now extract the entities and relationships from this image:
|
54 |
+
'''
|
55 |
+
|
56 |
+
payload = {
|
57 |
+
"model": "gpt-4o",
|
58 |
+
"messages": [
|
59 |
+
{
|
60 |
+
"role": "user",
|
61 |
+
"content": [
|
62 |
+
{
|
63 |
+
"type": "text",
|
64 |
+
"text": PROMPT
|
65 |
+
},
|
66 |
+
{
|
67 |
+
"type": "image_url",
|
68 |
+
"image_url": {
|
69 |
+
"url": f"data:image/jpeg;base64,{base64_image}"
|
70 |
+
}
|
71 |
+
}
|
72 |
+
]
|
73 |
+
}
|
74 |
+
]
|
75 |
+
}
|
76 |
+
|
77 |
+
# Send the request to the OpenAI API
|
78 |
+
response = requests.post("https://api.openai.com/v1/chat/completions", headers=headers, json=payload)
|
79 |
+
|
80 |
+
# Parse the response
|
81 |
+
response_data = response.json()
|
82 |
+
print(response_data)
|
83 |
+
|
84 |
+
# Extract the JSON graph data from the response
|
85 |
+
if "choices" in response_data and response_data["choices"]:
|
86 |
+
content = response_data["choices"][0]["message"]["content"]
|
87 |
+
try:
|
88 |
+
graph_data = content
|
89 |
+
except json.JSONDecodeError as e:
|
90 |
+
print("Failed:", e)
|
91 |
+
graph_data = None
|
92 |
+
else:
|
93 |
+
raise ValueError("No valid response from OpenAI API")
|
94 |
+
|
95 |
+
return graph_data
|
requirements.txt
ADDED
@@ -0,0 +1,99 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
aiofiles==23.2.1
|
2 |
+
annotated-types==0.7.0
|
3 |
+
anyio==4.4.0
|
4 |
+
asttokens==2.4.1
|
5 |
+
cachetools==5.5.0
|
6 |
+
certifi==2024.8.30
|
7 |
+
cffi==1.17.1
|
8 |
+
charset-normalizer==3.3.2
|
9 |
+
click==8.1.7
|
10 |
+
contourpy==1.3.0
|
11 |
+
cycler==0.12.1
|
12 |
+
decorator==5.1.1
|
13 |
+
exceptiongroup==1.2.2
|
14 |
+
executing==2.1.0
|
15 |
+
fastapi==0.114.1
|
16 |
+
ffmpy==0.4.0
|
17 |
+
filelock==3.16.0
|
18 |
+
fonttools==4.53.1
|
19 |
+
fsspec==2024.9.0
|
20 |
+
google-ai-generativelanguage==0.6.9
|
21 |
+
google-api-core==2.19.2
|
22 |
+
google-api-python-client==2.145.0
|
23 |
+
google-auth==2.34.0
|
24 |
+
google-auth-httplib2==0.2.0
|
25 |
+
google-generativeai==0.8.1
|
26 |
+
googleapis-common-protos==1.65.0
|
27 |
+
gradio==4.44.0
|
28 |
+
gradio_client==1.3.0
|
29 |
+
grpcio==1.66.1
|
30 |
+
grpcio-status==1.66.1
|
31 |
+
h11==0.14.0
|
32 |
+
httpcore==1.0.5
|
33 |
+
httplib2==0.22.0
|
34 |
+
httpx==0.27.2
|
35 |
+
huggingface-hub==0.24.7
|
36 |
+
idna==3.8
|
37 |
+
importlib_resources==6.4.5
|
38 |
+
ipython==8.27.0
|
39 |
+
jedi==0.19.1
|
40 |
+
Jinja2==3.1.4
|
41 |
+
jsonpickle==3.3.0
|
42 |
+
kiwisolver==1.4.7
|
43 |
+
markdown-it-py==3.0.0
|
44 |
+
MarkupSafe==2.1.5
|
45 |
+
matplotlib==3.9.2
|
46 |
+
matplotlib-inline==0.1.7
|
47 |
+
mdurl==0.1.2
|
48 |
+
neo4j==5.24.0
|
49 |
+
networkx==3.3
|
50 |
+
numpy==2.1.1
|
51 |
+
orjson==3.10.7
|
52 |
+
packaging==24.1
|
53 |
+
pandas==2.2.2
|
54 |
+
parso==0.8.4
|
55 |
+
pexpect==4.9.0
|
56 |
+
pillow==10.4.0
|
57 |
+
prompt_toolkit==3.0.47
|
58 |
+
proto-plus==1.24.0
|
59 |
+
protobuf==5.28.1
|
60 |
+
ptyprocess==0.7.0
|
61 |
+
pure_eval==0.2.3
|
62 |
+
pyasn1==0.6.1
|
63 |
+
pyasn1_modules==0.4.1
|
64 |
+
pycparser==2.22
|
65 |
+
pydantic==2.9.1
|
66 |
+
pydantic_core==2.23.3
|
67 |
+
pydub==0.25.1
|
68 |
+
Pygments==2.18.0
|
69 |
+
pyheif==0.8.0
|
70 |
+
PyMuPDF==1.24.10
|
71 |
+
PyMuPDFb==1.24.10
|
72 |
+
pyparsing==3.1.4
|
73 |
+
python-dateutil==2.9.0.post0
|
74 |
+
python-dotenv==1.0.1
|
75 |
+
python-multipart==0.0.9
|
76 |
+
pytz==2024.2
|
77 |
+
pyvis==0.3.2
|
78 |
+
PyYAML==6.0.2
|
79 |
+
requests==2.32.3
|
80 |
+
rich==13.8.1
|
81 |
+
rsa==4.9
|
82 |
+
ruff==0.6.4
|
83 |
+
semantic-version==2.10.0
|
84 |
+
shellingham==1.5.4
|
85 |
+
six==1.16.0
|
86 |
+
sniffio==1.3.1
|
87 |
+
stack-data==0.6.3
|
88 |
+
starlette==0.38.5
|
89 |
+
tomlkit==0.12.0
|
90 |
+
tqdm==4.66.5
|
91 |
+
traitlets==5.14.3
|
92 |
+
typer==0.12.5
|
93 |
+
typing_extensions==4.12.2
|
94 |
+
tzdata==2024.1
|
95 |
+
uritemplate==4.1.1
|
96 |
+
urllib3==2.2.3
|
97 |
+
uvicorn==0.30.6
|
98 |
+
wcwidth==0.2.13
|
99 |
+
websockets==12.0
|
utils.py
ADDED
@@ -0,0 +1,42 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
import os
|
2 |
+
import json
|
3 |
+
|
4 |
+
def extract_label_prefix(file_name):
|
5 |
+
label_prefix = os.path.splitext(os.path.basename(file_name))[0]
|
6 |
+
return label_prefix.replace(" ", "_").replace("-", "_")
|
7 |
+
|
8 |
+
def strip_keys(d):
|
9 |
+
if isinstance(d, dict):
|
10 |
+
return {k.strip(): strip_keys(v) for k, v in d.items()}
|
11 |
+
elif isinstance(d, list):
|
12 |
+
return [strip_keys(i) for i in d]
|
13 |
+
else:
|
14 |
+
return d
|
15 |
+
|
16 |
+
def format_json(json_data):
|
17 |
+
formatted_json = "{\n \"nodes\": [\n"
|
18 |
+
for node in json_data['nodes']:
|
19 |
+
formatted_json += f" {json.dumps(node)},\n"
|
20 |
+
formatted_json = formatted_json.rstrip(',\n') + "\n ],\n \"edges\": [\n"
|
21 |
+
for edge in json_data['edges']:
|
22 |
+
formatted_json += f" {json.dumps(edge)},\n"
|
23 |
+
formatted_json = formatted_json.rstrip(',\n') + "\n ]\n}"
|
24 |
+
return formatted_json
|
25 |
+
|
26 |
+
def validate_json(json_data):
|
27 |
+
if not isinstance(json_data, dict) or 'nodes' not in json_data or 'edges' not in json_data:
|
28 |
+
raise ValueError("JSON must contain 'nodes' and 'edges' keys")
|
29 |
+
|
30 |
+
if not isinstance(json_data['nodes'], list) or not isinstance(json_data['edges'], list):
|
31 |
+
raise ValueError("'nodes' and 'edges' must be lists")
|
32 |
+
|
33 |
+
for node in json_data['nodes']:
|
34 |
+
if 'id' not in node or 'label' not in node:
|
35 |
+
raise ValueError("Each node must have 'id' and 'label' properties")
|
36 |
+
|
37 |
+
for edge in json_data['edges']:
|
38 |
+
if 'source' not in edge or 'target' not in edge or 'type' not in edge:
|
39 |
+
raise ValueError("Each edge must have 'source', 'target', and 'type' properties")
|
40 |
+
|
41 |
+
if edge['type'] != "->":
|
42 |
+
raise ValueError("Edge type must be '->' strictly")
|