Roadmap¶
This page says where Docbank is going. Today it stores, organizes, versions, verifies, and backs up documents. The CLI, web app, and terminal browser share one authenticated local daemon. Go applications can also own separate vaults and use the document-processing packages directly.
For detail on current features, see the capability guide. Instructions and limits live in the guides linked below. The changelog records published releases.
What can I use now?¶
| Task | What works today | Guide |
|---|---|---|
| Collect documents | Import files and trees, select files with glob patterns, replace existing content on request, and watch local inbox folders | Importing |
| Organize documents | Stable IDs, folders, tags, and atomic batch moves. The web app groups and colors tags | Organizing |
| Find documents | Ranked name and text search, filtering without a query, and saved queries and highlight sets in the web app and API | Searching |
| Review and export | Frozen queries, verified ZIP bundles, dated search reports, load-file packages, and Bates-stamped selected pages | Web application and export bundles |
| Read email and recordings | Mailbox import, navigation between messages and their attachments, stored email PDFs, imported recordings, and timed transcripts | Web application and processing |
| Organize photos | RAW/JPEG/XMP grouping, a web grid grouped by month or capture session, CLI and HTTP albums, and metadata and album filters | Photo assets |
| Keep earlier content | Immutable versions, revision checks, reversion, and history pruning on request | Editing and versions |
| Record origin and evidence | Append-only provenance and permanent audited directory scopes | Importing and audited history |
| Recover documents | Trash with restore, permanent deletion as a separate step, garbage collection, and pack reclamation | Trash and GC |
| Recover a vault | Incremental backups, verified restore into a separate target, and snapshot removal through the embedded API | Backup and restore |
| Add physical storage | Verified placement, repair, salvage, and evacuation across filesystem and S3-compatible stores | Multi-store storage |
| Integrate an application | Authenticated HTTP API, local MCP server, offline OpenAPI generation, and embedded Go vaults with their own roots | Agent integration, MCP, and Embed in Go |
Release archives cover Linux, macOS, and Windows on amd64 and arm64.
docbank update installs a published release and coordinates daemon restart.
What document processing is available?¶
Go applications can extract source metadata, generate visual previews, and use local or hosted providers to produce text, Markdown, and embeddings. An embedding represents content as numbers for comparison. The Go processing guide maps the provider packages and their requirements.
The vault stores each derived result with the source version it came from. It also stores processing profiles, disclosure consent, and independent embedding sets. Vector indexes can be rebuilt. Hybrid retrieval combines lexical and vector matches, and is limited to the source versions a search is authorized for. Optional internal components expand queries, rerank results, and retrieve from an operator-hosted QMD. See Document processing for the flow and trust boundaries.
The daemon extracts supported plain text automatically. Configured embedding workers can run stored jobs for the registered runtime contracts. Those jobs still need prepared input generations, matching profiles, and current consent.
You can preview and run configured processing profiles, inspect jobs, and read the stored sanitized renditions through the CLI, HTTP API, web app, and TUI. Processing search has lexical, semantic, hybrid, and auto modes, and searches only the source versions you have authorized.
New imports do not automatically run OCR or prepare semantic search. The default configuration has no processing profiles, and search without them uses names and locally extracted text. With a configured embedding binding, Auto search in the web app and TUI uses Hybrid. Saving a query with hybrid-search settings does not run it. Frozen query snapshots support lexical mode only.
What can the web app and terminal browser do?¶
In the web app you can browse and search the vault, upload files, download current or earlier content, manage tags, and trash or restore documents. You can also inspect versions, source history, audit evidence, background jobs, backup recovery points, and storage health.
The terminal browser has tree navigation, search, document and version details, a node audit timeline, and trash with restore. Its operations views show daemon jobs, storage inventory, and configured backup recovery points.
Storage placement is still done through the CLI or the API. The storage views in the web app and terminal browser are read-only. The web app and HTTP API manage saved queries and highlight sets. The terminal browser has no screen for managing them.
What is planned?¶
- Automatic PDF and Office extraction for new vault imports and broader processing workflows.
- Overlapping permanent audit scopes.
- A retention contract for external references to Docbank nodes. Embedded source references currently record origin; they do not prevent deletion.
- Broader metadata inspection and version comparison in the web app.
- Import progress and broader metadata and evidence in the terminal browser.
- Tree, timeline, comparison, evidence, and job components that other applications can reuse.
- Automatic replication targets, primary-store replacement, lifecycle policies, and web or terminal storage-placement controls.
- Broader backup retention policy, and hardening against representative document collections.
Rich document comparison belongs in the web app or external tools.
Deferred beyond v1¶
- Encryption for live storage and backup repositories.
- Importing attachments from msgvault.
- Multi-user access and sharing.