Knowledge capture & classification
Documents, standards and experience are onboarded into one library, classified by space, tag and item type; items can be created from templates, with version history retained for review and rollback.
Services & Products | Provenan Desktop Knowledge Base · KnoWeave Enterprise Internal Knowledge Base System · Database & Research Software · Research Consulting · Data Analysis & Publication — Natural Analysis Technology
Product One · Desktop Software
Provenan: literature, notes and knowledge entries in one searchable library
A local knowledge base desktop application for researchers. Compile it yourself, deploy it yourself; the entire import, index and search workflow needs no network.
Knowledge Management & Metadata
Multi-library isolation
Multiple independent namespaces within one application; each library is bound to a vault directory and they never interfere with each other.
Item model
40+ item types, structured authors (multiple, ordered and typed), star ratings and scores, categories, tags and saved searches.
Identifier resolution
Enter a DOI / ISBN / PMID to auto-complete the full metadata, equivalent to Zotero's "magic wand".
Import & merge
Reads
zotero.sqlite in place, or imports RIS / BibTeX / JSON; supports duplicate and author merging, plus a soft-delete trash.Wiki Semantics
Bidirectional links
[[Wikilink]] syntax with backlink and outlink panels, plus a global graph view.Template engine
{{templates}} support a MediaWiki-compatible subset and can run real Lua modules (Scribunto).Footnote references
<ref> renders as clickable numbered footnotes.Offline rendering
A complete wikitext → HTML rendering pipeline without running a MediaWiki server.
Bulk migration
Bulk import pages from a local MediaWiki or Wikipedia; templates and modules have their dependencies extracted recursively.
Search & Provenance
Hybrid recall
FTS5 BM25 full-text search + 384-dimension local vector search, merged with RRF (reciprocal rank fusion).
Incremental indexing
File watching triggers debounced incremental indexing; no manual rebuild after edits.
Result location
PDF to page number and bounding box with highlight jumping; Markdown / Wikitext to line number and heading path; audio/video to timestamp.
Embedding model
Local ONNX inference by default; automatically falls back to BM25 search when the model is unavailable, never failing silently.
Cross-modal & Integrations
PDF
Parsed block by block while preserving each block's bounding box; the built-in reader jumps from search results to the hit location.
Audio/video
SRT / VTT / ASS subtitle segmentation, with chunks anchored to timestamps. Supports annotations on PDFs and web pages.
Structured extraction
{{chembox}}, {{Infobox scientist / gene / journal}} are extracted into typed entries such as chemical substances, researchers, genes and journals.KB-Link extension
A Zotero Connector-like Chrome extension that recognizes COinS / Highwire / JSON-LD / Dublin-Core / Open Graph and saves entries, full-page snapshots and wiki sources.
axiomd daemon
Exposes the same library to AI assistants over HTTP + MCP. Listens on localhost only and requires authentication by default.
Product Two · Enterprise Intranet
KnoWeave: turn scattered enterprise knowledge into a searchable internal asset
An internal knowledge base system deployed on your corporate network. Documents, standards and experience scattered across personal computers, shared drives and chat logs are consolidated, classified and made searchable; permissions, classification labels and audit logs keep knowledge flowing safely inside the company. It can work alongside the desktop Provenan, or run standalone as your enterprise knowledge base.
Full-text search
Titles, bodies, attachments and metadata are indexed together with Chinese word segmentation; results point precisely to the source paragraph and page, and attachment contents are searchable too.
Permissions & confidentiality
Visibility is controlled by department, role and item, with classification labels and export restrictions; access to sensitive content and unauthorized attempts are alerted and logged.
Collaborative editing
Multiple editors, comments and @ mentions, with a draft–review–publish workflow; every change is traceable to a specific person and time.
Attachment management
Attachments are content-addressed by sha256 for deduplication, so identical files are stored only once library-wide; version management and quota control are included, and downloads and previews are written to the audit log.
Audit log
Logins, searches, exports, downloads and permission changes are fully recorded, traceable by user, item and time range, and exportable for archival.
Governance & Security
Permission model
Department, role and item-level permissions; classification and visibility are configurable; access is revoked immediately when staff leave.
Content compliance
Sensitive-word and classification checks, with export approval; unauthorized access is denied and alerts are recorded.
Authentication
Passwords stored with bcrypt; JWT short-lived access tokens plus rotatable, revocable refresh tokens; authentication endpoints are rate-limited by IP.
Files & paths
Extension whitelist, sha256 verification to prevent path traversal, and content-addressed deduplication.
Private Deployment
Deployment form
A single Rust binary with SQLite storage, deployable entirely on the corporate intranet and able to run offline; in production a reverse proxy terminates TLS and forwards to a local port.
Data boundary
Data stays inside the corporate network, with offline installation and intranet upgrades; attachments are content-addressed with configurable quotas and retention.
Backup & recovery
Full-library or per-space scheduled backups, with backup files kept offline or archived off-site.
Provenan integration
Point at a local corpus directory for incremental sync, and export internal knowledge base items back to the desktop app for further editing.
Services
Three service lines, each with its own deliverables and prerequisites
We also spell out "what we need from you" — most friction in research services comes from this step not being discussed clearly.
Service Line One
Database & Research Software
Delivered on top of our self-developed Provenan and KnoWeave, or as corpus governance and onboarding alone without software customization.
Deliverables
- A running discipline-specific knowledge base instance and vault directory
- Cleaned, deduplicated and normalized literature metadata
- Custom item types and template definitions
- Private deployment plan and operations documentation
What we need from you
- Corpus scope and sources (database exports, PDF collections, internal documents)
- Target discipline and entry structure
- Whether the data can leave your local environment
Service Line Two
Research Consulting
Clarify the technical route before experiments begin. Suited to the pre-proposal or pre-submission stage, or when a plan keeps stalling.
Deliverables
- Project feasibility and risk assessment
- Optimized experimental plan and control setup
- Justification of an interdisciplinary technical route
- Recommendations on analysis criteria and statistical methods
What we need from you
- Research question and existing preliminary data
- Target journal or grant application requirements
- Available platforms, samples and time window
Service Line Three
Data Analysis & Publication
From raw data to submission-ready figures and text. Delivered to the format requirements of your target journal.
Deliverables
- Personalized bioinformatics analysis results with reproduction notes
- Editable vector figures and standardized captions
- Polished manuscript and revision records
- Point-by-point response file to reviewer comments
What we need from you
- Raw data and sequencing platform information
- Grouping design and sample metadata
- Target journal format requirements and submission stage
Specific pricing and turnaround depend on project scope; we do not publish a flat price list. Please describe your needs through the channels below and we will first give a feasibility assessment and a quote.
FAQ
About Working With Us
Can we just use the free software first and consider services later?
Absolutely. Provenan can be downloaded and used directly, and KnoWeave can be self-deployed. Coming back to us for project-based work only when professional judgment is needed (corpus governance, bioinformatics analysis, publication stage) is the more cost-effective order.
Our data cannot leave the internal network. What then?
Go with private deployment: the whole toolset is installed on your own server or workstation, we provide deployment and tuning, and day-to-day operation needs no external network. Consulting can also proceed with only a high-level view of the data.
How can the delivered analysis be reproduced?
Deliverables include a description of the analysis pipeline and key parameters. Software deliverables stay in your own environment and do not depend on our service staying online, so there is no "vendor disappears and the results won't open" problem.
Can you help respond to reviewer comments?
Yes. This includes drafting the point-by-point response file, supplementary analyses, figure adjustments and language revision. You will need to provide the original reviewer comments and the current manuscript.
Can you help migrate our internal Wiki to Provenan?
Yes. Provenan natively supports wikitext, templates and Lua modules, and can bulk import from a local MediaWiki and Wikipedia, extracting template dependencies recursively during migration. This is exactly the typical scenario for service line one.