DorsalHub - The File Metadata Platform

The Metadata Platform for

📂 Unstructured Data.
✅ Open Validation.
🧠 AI Extraction.
🗃️ Data Onboarding.
🏷️ File Annotation.
🚀 ETL Pipelines.
🔍 Semantic Search.
📂 Unstructured Data.

Treat Your Filesystem
Like a Database.

Build a queryable metadata layer for your unstructured data with the open-source Dorsal CLI and Python API.

[1]:

from dorsal import LocalFile
lf = LocalFile("C:\work stuff\report.pdf")

[2]:

lf.pdf.title

'ACME Corp End of Year Report'

[3]:

lf.pdf.page_count

[4]:

lf.get_latest_annotation("open/classification").record.labels

[{'label': 'Confidential', 'score': 0.95}]

[5]:

lf.add_label("Reviewed")

INFO:dorsal.file.dorsal_file:Private tag (label=Reviewed) added to local record for 'C:\work stuff\report.pdf'. Call push() to sync record with DorsalHub.

LocalFile[ report.pdf ]

[6]:

resp = lf.push(private=True)

INFO:dorsal.file.dorsal_file:Successfully pushed file record for 'C:\work stuff\report.pdf' to DorsalHub. Total: 1, Success: 1, Error: 0

(venv) ~/work $ dorsal search "extension:pdf label:Reviewed"
🔎 Searching user scope for records matching: 'extension:pdf label:Reviewed'
                            User Search Results
┏━━━━━━━━━━━━┳━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┓
┃ Name       ┃    Size ┃ SHA256 Hash                                                      ┃
┡━━━━━━━━━━━━╇━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┩
│ report.pdf │  80 KiB │ 46b065c86856297dec7ec0004127a5cdebeeff668f99ea7a8169609cbc4a0da5 │
│ Draft.pdf  │ 107 KiB │ 51d72e1186ae0b1e82ba0a4e77e10051231d8c4367f973a002e7afce79a7c094 │
└────────────┴─────────┴──────────────────────────────────────────────────────────────────┘
Showing page 1 of 1 | Displaying records 1 - 2 of 2 total.

(venv) ~/work $ dorsal record get 46b065c86856297dec7ec0004127a5cdebeeff668f99ea7a8169609cbc4a0da5
🔎 Searching hash 46b065c86856297dec7ec0004127a5cdebeeff668f99ea7a8169609cbc4a0da5
╭────────────────────────────── File Record: report.pdf ─────────────────────────────╮
│                                                                                    │
│  Hashes                                                                            │
│       SHA-256:  46b065c86856297dec7ec0004127a5cdebeeff668f99ea7a8169609cbc4a0da5   │
│                                                                                    │
│  File Info                                                                         │
│          Name:  report.pdf                                                         │
│          Size:  80 KiB                                                             │
│    Media Type:  application/pdf                                                    │
│                                                                                    │
│  Tags                                                                              │
│          label:  Reviewed (private) [id: 692ecb06dd8211652d6ab45b]                 │
│                                                                                    │
│  Pdf Info                                                                          │
│            author:  Jane Doe                                                       │
│             title:  ACME Corp End of Year Report                                   │
│          producer:  PDFlib+PDI 9.1.2p1 (PHP5/Linux-x86_64)                         │
│           version:  1.4                                                            │
│        page_count:  11                                                             │
│     creation_date:  2025-11-24T08:49:46Z                                           │
│     modified_date:  2025-11-24T08:49:46Z                                           │
│                                                                                    │
╰────────────────────────────────────────────────────────────────────────────────────╯
(venv) ~/work $

A Contract for Your Metadata.

Unstructured data shouldn't mean unstructured metadata. DorsalHub supports Open Validation Schemas for annotation, ensuring downstream workflows, dashboards, and models receive the data structure they expect.

Bring Your Own Intelligence.

Dorsal's pluggable local extraction pipeline makes it easy to integrate your own custom extraction logic.
Whether you need to parse proprietary binary formats, run internal models, or fetch data from an ERP, you can integrate it seamlessly with Dorsal.

📦 Open Source.

⚡ Local First.

🔒 Private by Default.