Update scripts/load_documents.py
Browse files- scripts/load_documents.py +31 -27
scripts/load_documents.py
CHANGED
@@ -7,31 +7,35 @@ from langchain_community.document_loaders import (
|
|
7 |
)
|
8 |
import pickle
|
9 |
|
10 |
-
|
11 |
-
|
12 |
-
|
13 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
14 |
|
15 |
-
|
16 |
-
|
17 |
-
loaders = {
|
18 |
-
".pdf": PyPDFLoader,
|
19 |
-
".txt": lambda path: TextLoader(path, encoding="utf-8"),
|
20 |
-
".py": PythonLoader,
|
21 |
-
".ipynb": NotebookLoader,
|
22 |
-
}
|
23 |
-
|
24 |
-
documents = []
|
25 |
-
for file in DATA_DIR.rglob("*"):
|
26 |
-
loader_class = loaders.get(file.suffix.lower())
|
27 |
-
if loader_class:
|
28 |
-
try:
|
29 |
-
docs = loader_class(str(file)).load()
|
30 |
-
documents.extend(docs)
|
31 |
-
print(f"[✓] Loaded: {file.name}")
|
32 |
-
except Exception as e:
|
33 |
-
print(f"[!] Failed to load {file.name}: {e}")
|
34 |
-
|
35 |
-
with open(OUTPUT_PATH, "wb") as f:
|
36 |
-
pickle.dump(documents, f)
|
37 |
-
print(f"📦 Saved {len(documents)} documents to {OUTPUT_PATH}")
|
|
|
7 |
)
|
8 |
import pickle
|
9 |
|
10 |
+
def main():
|
11 |
+
BASE_DIR = Path(__file__).resolve().parent.parent
|
12 |
+
DATA_DIR = BASE_DIR / "data"
|
13 |
+
OUTPUT_DIR = BASE_DIR / "output"
|
14 |
+
OUTPUT_PATH = OUTPUT_DIR / "all_docs.pkl"
|
15 |
+
|
16 |
+
OUTPUT_DIR.mkdir(parents=True, exist_ok=True)
|
17 |
+
|
18 |
+
loaders = {
|
19 |
+
".pdf": PyPDFLoader,
|
20 |
+
".txt": lambda path: TextLoader(path, encoding="utf-8"),
|
21 |
+
".py": PythonLoader,
|
22 |
+
".ipynb": NotebookLoader,
|
23 |
+
}
|
24 |
+
|
25 |
+
documents = []
|
26 |
+
for file in DATA_DIR.rglob("*"):
|
27 |
+
loader_class = loaders.get(file.suffix.lower())
|
28 |
+
if loader_class:
|
29 |
+
try:
|
30 |
+
docs = loader_class(str(file)).load()
|
31 |
+
documents.extend(docs)
|
32 |
+
print(f"[✓] Loaded: {file.name}")
|
33 |
+
except Exception as e:
|
34 |
+
print(f"[!] Failed to load {file.name}: {e}")
|
35 |
+
|
36 |
+
with open(OUTPUT_PATH, "wb") as f:
|
37 |
+
pickle.dump(documents, f)
|
38 |
+
print(f"📦 Saved {len(documents)} documents to {OUTPUT_PATH}")
|
39 |
|
40 |
+
if __name__ == "__main__":
|
41 |
+
main()
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|