Spaces:
Running on Zero
Running on Zero
Fix Gradio FileData path extraction and remove ZeroGPU decorator from file ingestion
Browse files
app.py
CHANGED
|
@@ -104,7 +104,6 @@ def get_status_markdown() -> str:
|
|
| 104 |
"""
|
| 105 |
|
| 106 |
|
| 107 |
-
@gpu_decorator
|
| 108 |
def process_uploaded_files(files) -> str:
|
| 109 |
"""Processes newly uploaded files and adds them to the index."""
|
| 110 |
global indexed_filenames
|
|
@@ -112,26 +111,47 @@ def process_uploaded_files(files) -> str:
|
|
| 112 |
return get_status_markdown()
|
| 113 |
|
| 114 |
new_parsed = []
|
|
|
|
|
|
|
| 115 |
for file_obj in files:
|
| 116 |
-
|
| 117 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 118 |
try:
|
| 119 |
-
parsed = pipeline.ingest_file(file_path)
|
| 120 |
-
parsed.
|
| 121 |
-
|
| 122 |
-
|
| 123 |
-
|
| 124 |
-
|
| 125 |
-
|
| 126 |
except Exception as e:
|
| 127 |
-
|
| 128 |
|
| 129 |
if new_parsed:
|
| 130 |
existing = list(retriever.chunks)
|
| 131 |
combined = existing + [c for p in new_parsed for c in p.chunks]
|
| 132 |
retriever.index_documents(combined)
|
| 133 |
|
| 134 |
-
|
|
|
|
|
|
|
|
|
|
| 135 |
|
| 136 |
|
| 137 |
def reload_samples_action() -> str:
|
|
|
|
| 104 |
"""
|
| 105 |
|
| 106 |
|
|
|
|
| 107 |
def process_uploaded_files(files) -> str:
|
| 108 |
"""Processes newly uploaded files and adds them to the index."""
|
| 109 |
global indexed_filenames
|
|
|
|
| 111 |
return get_status_markdown()
|
| 112 |
|
| 113 |
new_parsed = []
|
| 114 |
+
error_messages = []
|
| 115 |
+
|
| 116 |
for file_obj in files:
|
| 117 |
+
# Robust path and filename extraction across Gradio FileData / wrappers
|
| 118 |
+
if hasattr(file_obj, "path"):
|
| 119 |
+
file_path = file_obj.path
|
| 120 |
+
orig_name = getattr(file_obj, "orig_name", None) or os.path.basename(file_path)
|
| 121 |
+
elif hasattr(file_obj, "name"):
|
| 122 |
+
file_path = file_obj.name
|
| 123 |
+
orig_name = getattr(file_obj, "orig_name", None) or os.path.basename(file_path)
|
| 124 |
+
elif isinstance(file_obj, dict):
|
| 125 |
+
file_path = file_obj.get("path") or file_obj.get("name")
|
| 126 |
+
orig_name = file_obj.get("orig_name") or os.path.basename(file_path)
|
| 127 |
+
else:
|
| 128 |
+
file_path = str(file_obj)
|
| 129 |
+
orig_name = os.path.basename(file_path)
|
| 130 |
+
|
| 131 |
+
if not file_path or not os.path.exists(file_path):
|
| 132 |
+
error_messages.append(f"File not found: {orig_name}")
|
| 133 |
+
continue
|
| 134 |
+
|
| 135 |
try:
|
| 136 |
+
parsed = pipeline.ingest_file(file_path, original_filename=orig_name)
|
| 137 |
+
if parsed.chunks:
|
| 138 |
+
new_parsed.append(parsed)
|
| 139 |
+
if orig_name not in indexed_filenames:
|
| 140 |
+
indexed_filenames.append(orig_name)
|
| 141 |
+
else:
|
| 142 |
+
error_messages.append(f"No extractable text found in: {orig_name}")
|
| 143 |
except Exception as e:
|
| 144 |
+
error_messages.append(f"Error parsing {orig_name}: {str(e)}")
|
| 145 |
|
| 146 |
if new_parsed:
|
| 147 |
existing = list(retriever.chunks)
|
| 148 |
combined = existing + [c for p in new_parsed for c in p.chunks]
|
| 149 |
retriever.index_documents(combined)
|
| 150 |
|
| 151 |
+
status_md = get_status_markdown()
|
| 152 |
+
if error_messages:
|
| 153 |
+
status_md += "\n\n**⚠️ Upload Warnings:**\n" + "\n".join([f"- {msg}" for msg in error_messages])
|
| 154 |
+
return status_md
|
| 155 |
|
| 156 |
|
| 157 |
def reload_samples_action() -> str:
|
ingest.py
CHANGED
|
@@ -138,30 +138,51 @@ class DocumentIngestionPipeline:
|
|
| 138 |
self.ocr_fallback = ocr_fallback
|
| 139 |
self.ocr_min_chars = ocr_min_chars
|
| 140 |
|
| 141 |
-
def ingest_file(
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 142 |
"""Main entry point to parse and chunk a document."""
|
| 143 |
if not os.path.exists(file_path):
|
| 144 |
raise FileNotFoundError(f"File not found: {file_path}")
|
| 145 |
|
| 146 |
-
file_name = os.path.basename(file_path)
|
| 147 |
ext = os.path.splitext(file_name)[1].lower()
|
| 148 |
if not doc_id:
|
| 149 |
doc_id = str(uuid.uuid4())[:8]
|
| 150 |
|
| 151 |
-
|
| 152 |
-
|
| 153 |
-
|
| 154 |
-
|
| 155 |
-
|
| 156 |
-
|
| 157 |
-
|
| 158 |
-
|
| 159 |
-
|
| 160 |
-
|
| 161 |
-
|
| 162 |
-
|
| 163 |
-
|
| 164 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 165 |
|
| 166 |
return ParsedDocument(
|
| 167 |
doc_id=doc_id,
|
|
|
|
| 138 |
self.ocr_fallback = ocr_fallback
|
| 139 |
self.ocr_min_chars = ocr_min_chars
|
| 140 |
|
| 141 |
+
def ingest_file(
|
| 142 |
+
self,
|
| 143 |
+
file_path: str,
|
| 144 |
+
doc_id: Optional[str] = None,
|
| 145 |
+
original_filename: Optional[str] = None
|
| 146 |
+
) -> ParsedDocument:
|
| 147 |
"""Main entry point to parse and chunk a document."""
|
| 148 |
if not os.path.exists(file_path):
|
| 149 |
raise FileNotFoundError(f"File not found: {file_path}")
|
| 150 |
|
| 151 |
+
file_name = original_filename or os.path.basename(file_path)
|
| 152 |
ext = os.path.splitext(file_name)[1].lower()
|
| 153 |
if not doc_id:
|
| 154 |
doc_id = str(uuid.uuid4())[:8]
|
| 155 |
|
| 156 |
+
# Ensure working path has proper extension for third-party parsers
|
| 157 |
+
actual_path = file_path
|
| 158 |
+
temp_target = None
|
| 159 |
+
if not file_path.lower().endswith(ext) and ext:
|
| 160 |
+
import shutil
|
| 161 |
+
temp_target = f"{file_path}_{uuid.uuid4().hex[:6]}{ext}"
|
| 162 |
+
shutil.copyfile(file_path, temp_target)
|
| 163 |
+
actual_path = temp_target
|
| 164 |
+
|
| 165 |
+
try:
|
| 166 |
+
if ext == ".pdf":
|
| 167 |
+
raw_elements, total_pages = self._parse_pdf(actual_path)
|
| 168 |
+
file_type = "pdf"
|
| 169 |
+
elif ext in [".docx", ".doc"]:
|
| 170 |
+
raw_elements, total_pages = self._parse_docx(actual_path)
|
| 171 |
+
file_type = "docx"
|
| 172 |
+
elif ext in [".txt", ".md"]:
|
| 173 |
+
raw_elements, total_pages = self._parse_txt(actual_path)
|
| 174 |
+
file_type = "txt"
|
| 175 |
+
else:
|
| 176 |
+
raise ValueError(f"Unsupported file type: {ext}. Supported: .pdf, .docx, .txt, .md")
|
| 177 |
+
|
| 178 |
+
# Chunk parsed elements with structure awareness
|
| 179 |
+
chunks = self._chunk_elements(raw_elements, doc_id, file_name)
|
| 180 |
+
finally:
|
| 181 |
+
if temp_target and os.path.exists(temp_target):
|
| 182 |
+
try:
|
| 183 |
+
os.remove(temp_target)
|
| 184 |
+
except Exception:
|
| 185 |
+
pass
|
| 186 |
|
| 187 |
return ParsedDocument(
|
| 188 |
doc_id=doc_id,
|