Converts various document formats into a unified structured representation. Extracts text, formatting, and metadata from files and returns them as document blocks with inline elements.
Supported formats:
- Documents: PDF, DOCX, plain text, Markdown, HTML
- Images: PNG, JPG, JPEG, WebP, GIF (OCR-capable)
- Legacy formats: DOC, RTF, ODT, XLS, XLSX, PPT, PPTX (basic support)
Input methods:
- Upload files directly
- Parse from local file paths (restricted to
allowed_local_base_dir; paths with..are always rejected)
Full API documentation: http://127.0.0.1:8087/docs
allowed_local_base_dir is required. The module will refuse to start if it is missing or the path cannot be resolved.
modules:
file-parser:
config:
max_file_size_mb: 100
allowed_local_base_dir: /data/documentsOnly files under this directory (after symlink resolution) are accessible via the parse-local endpoints. Paths containing .. are always rejected regardless of where they point.
curl -s http://127.0.0.1:8087/file-parser/v1/info | python3 -m json.toolOutput:
{
"supported_extensions": {
"plain_text": ["txt", "log", "md"],
"html": ["html", "htm"],
"pdf": ["pdf"],
"docx": ["docx"],
"image": ["png", "jpg", "jpeg", "webp", "gif"],
"generic_stub": ["doc", "rtf", "odt", "xls", "xlsx", "ppt", "pptx"]
}
}echo "Hello, Cyber Ware!" > /tmp/test.txt
curl -s -X POST "http://127.0.0.1:8087/file-parser/v1/upload?filename=test.txt" \
-H "Content-Type: application/octet-stream" \
--data-binary @/tmp/test.txt | python3 -m json.toolOutput:
{
"document": {
"id": "019bc231-fcfd-7df3-a49c-82174973ec44",
"title": "test.txt",
"meta": {
"source": {"type": "uploaded", "original_name": "test.txt"},
"content_type": "text/plain"
},
"blocks": [
{
"type": "paragraph",
"inlines": [{"type": "text", "text": "Hello, Cyber Ware!", "style": {}}]
}
]
}
}Assumes allowed_local_base_dir is set to /data/documents and the file exists there.
curl -s -X POST "http://127.0.0.1:8087/file-parser/v1/parse-local?render_markdown=true" \
-H "Content-Type: application/json" \
-d '{"file_path": "/data/documents/report.txt"}' | python3 -m json.toolOutput:
{
"document": {
"id": "019bc232-abcd-7890-b123-456789abcdef",
"title": "report.txt",
"meta": {
"source": {"type": "local_path", "path": "/data/documents/report.txt"},
"content_type": "text/plain"
},
"blocks": [
{
"type": "paragraph",
"inlines": [{"type": "text", "text": "Report content here.", "style": {}}]
}
]
},
"markdown": "Report content here.\n"
}Path with .. component — always rejected before any filesystem access:
curl -s -X POST "http://127.0.0.1:8087/file-parser/v1/parse-local" \
-H "Content-Type: application/json" \
-d '{"file_path": "/data/documents/../etc/passwd"}'Response: 403 Forbidden
{
"status": 403,
"title": "Path Traversal Blocked",
"detail": "Access denied: path '/data/documents/../etc/passwd' contains '..' traversal component"
}Path outside allowed_local_base_dir — rejected after canonicalization:
curl -s -X POST "http://127.0.0.1:8087/file-parser/v1/parse-local" \
-H "Content-Type: application/json" \
-d '{"file_path": "/etc/hostname"}'Response: 403 Forbidden
{
"status": 403,
"title": "Path Traversal Blocked",
"detail": "Access denied: '/etc/hostname' is outside the allowed base directory"
}For additional endpoints, see http://127.0.0.1:8087/docs.