Introduction
Project Name: SiaMind
Name of the organisation or individual submitting the proposal: Jules Lai (Fabstir)
Describe your project.
SiaMind is a browser-based knowledge workspace — think Obsidian, but with every note, document, and link stored on Sia via Enhanced S5.js, encrypted and owned entirely by the user. To our knowledge, no application of this kind exists on Sia today.
Sia’s consumer application portfolio is growing — DecaNotes delivered basic Markdown note-taking, Chi-voice is collecting audio datasets — but there is no Obsidian-style knowledge workspace: a tool where users write and link notes, upload and reference documents, navigate a personal knowledge graph, and search across everything, all backed by decentralised, encrypted storage. SiaMind fills this gap.
Users write Markdown notes, upload documents (PDF, text, images), and link them together in a personal knowledge graph powered by FabstirDB v2, a graph database rewritten from scratch to run natively on Enhanced S5.js. No data ever touches a centralised server — everything is stored on Sia, encrypted with user-controlled keys.
As an optional extra, users can enable AI-powered features via Platformless AI — chatting with their knowledge base, summarising documents, and discovering connections across notes — all through decentralised, end-to-end encrypted inference. But the core application is fully functional without AI: a private, sovereign knowledge workspace on Sia.
Core features:
- A polished, browser-based note editor with Markdown support (block-based editing, formatting toolbar, keyboard shortcuts)
- Document upload and management (PDF, text, images) stored on Sia via Enhanced S5.js
- Bidirectional linking between notes and documents — [[wiki-style links]] powered by FabstirDB v2’s graph layer, creating a navigable knowledge graph
- Graph visualisation showing how notes and documents connect
- Full-text search across all notes and uploaded documents
- Installable as a Progressive Web App (PWA) for a native app experience on desktop and mobile
- All data encrypted with user-controlled keys; works offline with local cache and syncs when connected
Optional AI features (via Platformless AI):
- AI chat over the knowledge base — RAG-powered question answering with source citations
- Document summarisation and key concept extraction
- Connection discovery across notes and documents
Why browser-based via Enhanced S5.js rather than mobile-native or SaaS:
Enhanced S5.js — completed under a previous Sia Foundation grant — provides a path-based filesystem, HAMT sharding, XChaCha20-Poly1305 encryption, CBOR serialisation, and deterministic identity. It runs in the browser, allowing SiaMind to store data on Sia’s decentralised network without any server-side component. This is the same approach approved for the Chi-voice grant, which also builds on Enhanced S5.js.
A SaaS model would require routing user data through centralised servers, defeating the purpose. Native mobile apps would require maintaining separate iOS/Android codebases and app store approvals for what is fundamentally a document editor. A PWA gives mobile-quality UX while keeping the entire data path decentralised.
Optional AI integration via Platformless AI:
SiaMind is designed to be fully functional without AI — the knowledge workspace stands on its own. However, for users who want AI-powered features, SiaMind integrates with Platformless AI as an optional extra.
Platformless AI already supports project-based document upload for RAG — users can organise documents into a hierarchical database and folder structure (not just a flat file list like centralised AI platforms), upload them to decentralised hosts, and chat with their content via encrypted inference sessions. SiaMind enhances this by giving the AI awareness of the knowledge graph: instead of searching a flat collection of files, the RAG pipeline can traverse bidirectional links between notes and documents, surfacing connections that no flat-file system can find. This is particularly valuable for researchers working with large, interrelated bodies of material.
Because the AI layer is optional, SiaMind works for users who simply want a private, Sia-backed alternative to Obsidian — and it works even better for users who also want NotebookLM-style AI capabilities without sending their data to Google.
Who benefits from your project?
- Privacy-conscious knowledge workers — anyone currently using Notion, Obsidian Sync, or Google Keep who wants the same functionality without corporate servers holding their data. SiaMind gives them a fully-featured knowledge workspace where every note and document lives on Sia, encrypted and sovereign.
- Researchers and academics — manage papers, notes, and citations in a private knowledge graph with bidirectional links. Optionally enable AI to ask questions across their entire corpus without sending data to Google or OpenAI.
- Journalists and investigators — link sources, documents, and notes with bidirectional relationships. The graph visualisation surfaces connections across large volumes of material, all encrypted.
- The Sia ecosystem — SiaMind fills a gap in Sia’s consumer application portfolio, demonstrating Sia as infrastructure for everyday productivity tools and joining Chi-voice and DecaNotes in building a portfolio of consumer applications.
- The Platformless AI ecosystem (optional) — for users who enable AI features, SiaMind provides a compelling consumer frontend for decentralised RAG capabilities.
How does the project serve the Foundation’s mission of user-owned data?
SiaMind is a direct embodiment of user-owned data. Today, anyone using Notion, Obsidian Sync, Google Keep, or Apple Notes is storing their most private thoughts, research, and intellectual work on corporate servers. Even “local-first” tools like Obsidian require a paid sync service to work across devices, routing data through centralised infrastructure.
SiaMind replaces this with a model where:
- Notes live on Sia via Enhanced S5.js — the user’s data is distributed across Sia’s decentralised storage network, not sitting on any company’s servers
- Encryption is default — all content is encrypted with user-controlled keys before leaving the device. No one — not Sia hosts, not the application developer, not anyone — can read the user’s notes
- Data follows the user — because data is addressed by content hash and tied to the user’s identity, it’s accessible from any device with the user’s credentials. No vendor lock-in, no export/import headaches
- AI is private (optional) — for users who enable AI features, queries and embeddings flow through Platformless AI’s end-to-end encrypted channels. No centralised API provider sees the user’s documents or questions. But the core workspace is fully functional without AI.
This project demonstrates Sia as infrastructure for everyday productivity tools, not just developer-facing storage. It shows non-technical users what “user-owned data” feels like in practice: their notes, their keys, their data — accessible everywhere, controlled by no one else.
We cannot provide grants to residents of jurisdictions under increased FATF monitoring, those that have active OFAC sanctions, or those that fail our bank compliance tests. We also cannot provide grants if your payment bank account is located in those same locations. Please review the following list.
Are you a resident of any jurisdiction on that list? No
Will your payment bank account be located in any jurisdiction on that list? No
Grant Specifics
Amount of money requested and justification with a reasonable breakdown of expenses
Total Amount Requested: $35,000 USD
| Month | Milestone | Budget (USD) |
|---|---|---|
| Month 1 | Design Documents & Project Foundation | $7,000 |
| Month 2 | Markdown Editor, Note Storage, Encryption & Offline Support | $7,000 |
| Month 3 | Bidirectional Linking, Documents & Graph Layer | $7,000 |
| Month 4 | Search, Graph Visualisation, PWA & Navigation UI | $7,000 |
| Month 5 | AI Integration, Demo Video & Final Release | $7,000 |
| Total | 5 months | $35,000 |
Justification:
- Proven delivery record: Enhanced S5.js completed 25% ahead of schedule under previous Sia Foundation grant, with deliverables exceeding technical requirements (10.6× under bundle size target)
- Prior experience building both the graph data patterns (FabstirDB) and the AI infrastructure (Platformless AI with RAG/vector search already production-validated)
- London-based development costs
- Senior-level expertise in decentralised storage, graph databases, and AI inference systems
- The grant funds application development and integration — the underlying infrastructure (Enhanced S5.js, FabstirDB v2, Platformless AI SDK, vector database) already exists, significantly reducing delivery risk
High-level architecture overview
┌─────────────────────────────────────────────────────────┐
│ SiaMind (Browser PWA) │
│ │
│ ┌──────────────┐ ┌──────────────┐ ┌──────────────┐ │
│ │ Markdown │ │ Document │ │ AI Chat │ │
│ │ Editor │ │ Viewer │ │ (User │ │
│ │ (Block-based)│ │ (PDF/Text) │ │ Opt-in) │ │
│ └──────┬───────┘ └──────┬───────┘ └──────┬───────┘ │
│ │ │ │ │
│ ┌──────┴─────────────────┴──────────────────┴───────┐ │
│ │ SiaMind SDK (TypeScript) │ │
│ │ • Note CRUD, linking, search │ │
│ │ • Document upload/retrieval │ │
│ │ • FabstirDB v2 graph layer (GUN-compatible API) │ │
│ │ • Local cache (IndexedDB) + sync │ │
│ └──────────────────┬────────────────────────────────┘ │
│ │ │
│ ┌──────────────────┴────────────────────────────────┐ │
│ │ FabstirDB v2 + Enhanced S5.js │ │
│ │ • Graph storage (nodes, edges, bidirectional) │ │
│ │ • XChaCha20-Poly1305 encryption (client-side) │ │
│ │ • CBOR serialisation for structured data │ │
│ │ • HAMT sharding for scalable storage │ │
│ │ • Deterministic paths for notes/docs │ │
│ └──────────────────┬────────────────────────────────┘ │
│ │ │
│ ┌──────────────────┴────────────────────────────────┐ │
│ │ Platformless AI SDK (User Opt-in) │ │
│ │ • Encrypted WebSocket to GPU hosts │ │
│ │ • RAG: embed docs → vector search → LLM answer │ │
│ │ • Vision: OCR uploaded images/PDFs │ │
│ │ • Web search integration │ │
│ └───────────────────────────────────────────────────┘ │
└─────────────────────────────────────────────────────────┘
│ │
▼ ▼
┌──────────────┐ ┌──────────────────┐
│ Sia Network │ │ Platformless AI │
│ via S5 │ │ GPU Hosts │
│ │ │ (Decentralised) │
│ • Notes │ │ │
│ • Documents │ │ • Embeddings │
│ • Indexes │ │ • LLM Inference │
│ • Graph │ │ • Vision/OCR │
└──────────────┘ └──────────────────┘
Security best practices:
- All data encrypted client-side with XChaCha20-Poly1305 (built into Enhanced S5.js) before upload to Sia
- User identity derived deterministically via Enhanced S5.js (no passwords stored)
- AI queries transmitted over end-to-end encrypted WebSocket connections via Platformless AI protocol
- No server-side component — SiaMind is a static site that communicates directly with Sia via Enhanced S5.js and with Platformless AI hosts
- Content Security Policy headers to prevent XSS
- Subresource Integrity for all external dependencies
- All dependencies audited and pinned to specific versions
Following the Sia Foundation Development Guide: all code open-source (MIT licence), clean commit history, comprehensive test coverage (>85%), CI/CD pipeline with automated testing, documentation in repository README and dedicated docs.
Timeline with measurable objectives and goals
Total Duration: 5 months
Milestone 1: Design Documents & Project Foundation (Month 1)
Objective: Produce design documents for the full application and establish a working project foundation with Sia storage connectivity and authentication.
| Task | Description | Measurable Outcome |
|---|---|---|
| 1.1 | Design documents | Data model design, storage path conventions, UI wireframes, component architecture, authentication strategy, and API surface for each module. Published in repository. |
| 1.2 | Project scaffolding | Next.js/React app with TypeScript, CI/CD pipeline, linting, test framework |
| 1.3 | Enhanced S5.js storage integration | Basic connectivity — upload, download, list, and delete operations working against S5 portal |
| 1.4 | Authentication flow | Sign-in flow deriving S5 identity (approach determined during design — wallet-based, social login, or hybrid) |
| 1.5 | Unit & integration tests | >85% coverage for storage integration and authentication |
Success Criteria:
- Design documents published covering data model, architecture, UI wireframes, and auth strategy
- User can sign in and store/retrieve data on Sia
- All tests passing in CI
Milestone 2: Markdown Editor, Note Storage, Encryption & Offline Support (Month 2)
Objective: Deliver a functional encrypted note editor with persistence to Sia and offline capability.
| Task | Description | Measurable Outcome |
|---|---|---|
| 2.1 | Note data model | CBOR schema for notes (content, metadata, timestamps, tags) stored via deterministic S5 paths |
| 2.2 | Client-side encryption | XChaCha20-Poly1305 encryption/decryption of all note data before upload |
| 2.3 | Markdown editor UI | Block-based Tiptap editor with formatting toolbar, keyboard shortcuts, live preview |
| 2.4 | Note CRUD operations | Create, read, update, delete notes — encrypted and persisted to Sia via Enhanced S5.js |
| 2.5 | Local cache layer | IndexedDB cache for offline access; sync queue for pending uploads |
| 2.6 | Conflict resolution | Last-write-wins with timestamp-based merge for edits made offline |
| 2.7 | Basic sidebar navigation | Note list, create new note, sort by date/title |
| 2.8 | Tests | >85% coverage for editor state, encryption, CRUD, and offline sync |
Success Criteria:
- User can create/edit/delete Markdown notes with a polished editor
- Notes are encrypted before upload and decrypted on retrieval
- App works offline via local cache and syncs when reconnected
- All tests passing in CI
Milestone 3: Bidirectional Linking, Document Upload & Graph Layer (Month 3)
Objective: Integrate FabstirDB v2 for the knowledge graph and add document management.
| Task | Description | Measurable Outcome |
|---|---|---|
| 3.1 | FabstirDB v2 integration | Bidirectional note relationships via FabstirDB v2’s GUN-compatible API on Enhanced S5.js |
| 3.2 | Wiki-style linking | [[note-name]] syntax parsed in editor; creates graph edges in FabstirDB v2 |
| 3.3 | Backlinks panel | Each note displays a list of all other notes that link to it |
| 3.4 | Document upload | Upload PDF, text, and image files to Sia via Enhanced S5.js |
| 3.5 | Document-to-note linking | Link uploaded documents to notes; documents appear as referenced resources |
| 3.6 | Tag system | Tag notes and documents; filter by tags in sidebar |
| 3.7 | Tests | >85% coverage for linking, graph operations, and document upload |
Success Criteria:
- User can create wiki-style links between notes and see backlinks
- PDF/text/image documents can be uploaded and linked to notes
- FabstirDB v2 maintains bidirectional graph relationships
- All tests passing in CI
Milestone 4: Search, Graph Visualisation, PWA & Navigation UI (Month 4)
Objective: Full-text search, interactive graph view, polished navigation, and PWA packaging — delivering a complete, production-ready core application.
| Task | Description | Measurable Outcome |
|---|---|---|
| 4.1 | Full-text search | Client-side search index built from note/document content; instant results |
| 4.2 | Graph visualisation | Interactive force-directed graph showing note/document connections |
| 4.3 | File tree sidebar | Hierarchical note/folder navigation with drag-and-drop organisation |
| 4.4 | Recent notes & quick switcher | Quick access to recent notes; Cmd/Ctrl+K switcher |
| 4.5 | PWA packaging | Service worker for app shell caching, manifest, install prompt |
| 4.6 | Responsive layout | Desktop, tablet, and mobile layouts with collapsible sidebar |
| 4.7 | Developer documentation | API reference, architecture docs, deployment guide |
| 4.8 | Tests | >85% coverage for search, graph rendering, navigation, and PWA |
Success Criteria:
- Full-text search returns results across all notes and documents
- Graph view renders note/document connections interactively
- App installable as PWA on desktop and mobile
- Navigation is polished and responsive across screen sizes
- All tests passing in CI
Milestone 5: AI Integration, Demo Video & Final Release (Month 5)
Objective: Integrate Platformless AI for RAG-powered chat over the knowledge base, record a comprehensive demo video showcasing the full application, and deliver final documentation and release. Note: AI features are delivered as part of this grant but are designed as a user-facing opt-in — SiaMind works fully without AI enabled.
| Task | Description | Measurable Outcome |
|---|---|---|
| 5.1 | Platformless AI SDK integration | Connect to decentralised GPU hosts via encrypted WebSocket |
| 5.2 | Document embedding pipeline | Chunk notes/documents → generate embeddings via Platformless AI hosts |
| 5.3 | Vector index on Sia | Store embedding vectors on Sia via Enhanced S5.js; retrieve relevant chunks for queries |
| 5.4 | AI chat panel & RAG pipeline | Chat interface with query → vector search → LLM response with source citations |
| 5.5 | Demo knowledge base | Seed SiaMind with sample notes, documents, and links to showcase all features in the demo |
| 5.6 | Demo video | Comprehensive walkthrough video demonstrating: sign-in, note creation, Markdown editing, document upload, wiki-linking, backlinks, graph visualisation, search, PWA install, and Platformless AI chat with knowledge base |
| 5.7 | User guide & final documentation | End-user guide, getting started tutorial, FAQ |
| 5.8 | Final testing & polish | >85% overall coverage; performance benchmarks documented; zero critical bugs |
Success Criteria:
- User can chat with their knowledge base and receive answers with source citations via Platformless AI
- Demo video published showing end-to-end usage of SiaMind including AI features
- App deployed and publicly accessible for committee testing
- Comprehensive user guide and developer documentation published
- All tests passing in CI; zero critical bugs
What are your plans for this project following the grant?
SiaMind is designed to be a standalone, open-source tool that anyone can use immediately after the grant concludes. Post-grant plans include:
- Platformless AI integration — SiaMind’s SDK will serve as the knowledge management layer within Platformless AI, giving its users Obsidian/NotebookLM-style capabilities through the existing chat interface and SDK without needing to interact with SiaMind directly. As part of Platformless AI, SiaMind will reach researchers, journalists, privacy advocates, and knowledge workers who care about data sovereignty.
Potential risks that will affect the outcome of the project
| Risk | Mitigation | Impact | Probability |
|---|---|---|---|
| S5 portal availability or latency | Aggressive local caching with write-through sync. User works against local IndexedDB cache; background sync to Sia. Proven pattern from Enhanced S5.js grant. Multiple S5 portals can be configured as fallbacks. | Medium | Low |
| Platformless AI host availability for RAG | AI features are user opt-in — the core note-taking and knowledge graph work without AI. Platformless AI infrastructure is already production-validated. | Low | Low |
| Browser storage limitations for large knowledge bases | Use IndexedDB for hot cache only; full data on Sia. Implement lazy loading and pagination for large note collections. HAMT sharding in Enhanced S5.js handles large directory structures efficiently. | Low | Low |
| FabstirDB v2 integration complexity | FabstirDB v2 was written from scratch specifically for Enhanced S5.js integration and is already working with Fabstir v2. The graph patterns are proven and the GUN-compatible API is stable. | Low | Very Low |
| Editor complexity | Use Tiptap (ProseMirror-based), a battle-tested open-source editor framework, rather than building from scratch. Reduces risk significantly. | Low | Very Low |
Development Information
Will all of your project’s code be open-source?
Yes. All code developed under this grant for SiaMind will be open-source under the MIT licence, with no closed-source components.
Leave a link where code will be accessible for review.
- SiaMind Repository: https://github.com/Fabstir/siamind (to be created upon approval)
- Previous work:
- Enhanced S5.js — Sia Foundation grant, completed November 2025
- FabstirDB v2 — Graph database rewritten from scratch for Enhanced S5.js, GUN-compatible API, integrated with Fabstir v2. Repository will be published to GitHub ahead of the Karma Sanctum Soho live demo on April 29th, 2026. (Original FabstirDB available here for reference.)
- fabstir-vectordb — Production vector database on S5
- Platformless AI Whitepaper — Decentralised AI inference protocol
Track record with the Sia Foundation:
-
Enhanced S5.js (Standard Grant, 2025) — Completed 25% ahead of schedule. Delivered path-based filesystem, HAMT sharding, XChaCha20-Poly1305 encryption, CBOR serialisation, media processing, deterministic identity from wallet, CID-based downloads. Bundle size came in at 10.6× under the requirement. All milestones delivered on or ahead of schedule with comprehensive documentation. Grant marked as Complete.
-
Fabstir Media Player (Prior grants) — Multiple successful grants building decentralised streaming infrastructure on Sia.
-
Platformless AI (Independent) — Decentralised AI inference protocol with Anthropic and OpenAI API compatibility layers, supporting agentic AI workflows with full tool use, multi-turn orchestration, and streaming. Smart contracts independently audited by Hacken, a leading blockchain security firm.
-
FabGraph (Two proposals, both rejected) — I took the practical approach: I rewrote FabstirDB from scratch as FabstirDB v2, integrating it directly with Enhanced S5.js. Where the original FabstirDB was a wrapper around OrbitDB on IPFS, FabstirDB v2 is built natively for S5.js while preserving the GUN-compatible API. It delivers the core graph capabilities — bidirectional relationships, traversal, and indexed lookups — that real applications need. SiaMind is the consumer application that puts FabstirDB v2 to work.
Have you developed a proof of concept?
Yes, the critical infrastructure components already exist and are production-tested:
- Enhanced S5.js provides the storage primitives (path-based read/write, encryption, CBOR, HAMT sharding) — completed under Sia Foundation grant
- FabstirDB v2 provides the graph database layer — rewritten from scratch to integrate natively with Enhanced S5.js, with a GUN-compatible API for bidirectional relationships, currently working with Fabstir v2. FabstirDB v2 and Fabstir v2 will be published to GitHub ahead of the public announcement and live demo at the Karma Sanctum Soho event in London on April 29th, 2026.
- Platformless AI provides the AI inference layer (RAG/vector search, host-side embeddings, encrypted WebSocket sessions) (demo)
- fabstir-vectordb proved vector similarity search works on S5/Sia storage (demo)
SiaMind combines these proven components into a consumer-facing application. The core technical risk — “can you build a responsive knowledge tool on decentralised storage?” — has already been answered by the Enhanced S5.js and FabstirDB v2 work.
Do you agree to submit monthly progress reports?
Yes. Progress reports will be submitted monthly in the Sia Forum grants thread, following the Foundation’s required format.
Contact info
Email: [email protected]
Any other preferred contact methods: Sia Discord (@jules_lai)