Tip | AI-Assisted Install: Just tell any AI assistant: |
Typell is the type-theoretic verification engine that powers PanLL's neurosymbolic intelligence.
It provides the "best of the best" type system coverage for formally verified database queries:
Dependent types — types that depend on values (Pi, Sigma)
Linear types — resources used exactly once (no data leaks, no double-reads)
Session types — protocol safety (connections always closed, transactions atomic)
Quantitative Type Theory — resource quantity tracking (rate limits, cost analysis)
Effect systems — explicit side effect tracking (read/write/memory)
Modal types — contextual access control (transaction-scoped data)
Proof-carrying code — cryptographic proof certificates attached to queries
NOT a standalone IDE — it is a verification engine consumed by PanLL and other tools
NOT a pane in PanLL — it is the intelligence behind all three panes
NOT a database — it validates queries, it does not store data
NOT a theorem prover — it delegates complex proofs to Echidna
┌─────────────────────────────────────────────┐ │ PanLL (Development Environment) │ │ ├─ Pane-L ← constraints from Typell │ │ ├─ Pane-N ← reasoning from Typell │ │ └─ Pane-W ← validated results from Typell │ └──────────────┬──────────────────────────────┘ │ Verification Protocol (JSON-RPC) ┌──────────────┴──────────────────────────────┐ │ Typell (Verification Kernel) │ │ ├─ Type Checker (bidirectional) │ │ ├─ Proof Engine (gen/verify/cache/cert) │ │ ├─ Effect Tracker │ │ ├─ Session Protocol Manager │ │ └─ Language Backends │ │ ├─ VCL-dt++ (VeriSimDB) │ │ ├─ GQL-dt++ (LithoGlyph) │ │ └─ KQL-dt++ (QuandleDB) │ └─────────────────────────────────────────────┘ ↑ ↑ ↑ VS Code CLI/CI Databases extension pipelines (direct use)
| Backend | Database | Status | Strategy |
|---|---|---|---|
VCL-dt++ | VeriSimDB | Port from ReScript | 8-modality queries, cross-modal proofs, hexad types |
GPNL-dt++ | LithoGlyph | Bridge (Lean 4) | Knowledge graphs, RATIONALE clause, refinement types |
KRL-dt++ | QuandleDB | Design from scratch | Category-theoretic schema, equality saturation, HoTT |
| Layer | Technology | Rationale |
|---|---|---|
Formal Specs | Idris2 ( | Dependent types prove type system soundness. Zero believe_me. |
Kernel | Rust ( | Performance, safety, Tauri compatibility (PanLL backend). |
FFI | Zig ( | C ABI compatibility per hyperpolymath standard. |
Protocol | JSON-RPC | Language-agnostic. Any consumer can talk to Typell. |
Phase 0: Vision capture and repo scaffolding — complete.
See docs/design/DESIGN-2026-03-01-typell-vision.md for the full design vision.
See .machine_readable/6a2/STATE.a2ml for current project state.
ABI → Idris2 (
src/abi/*.idr) — type definitions with dependent type proofsFFI → Zig (
ffi/zig/src/*.zig) — C-compatible function implementationsGenerated C Headers → Auto-generated from Idris2 ABI (
generated/abi/*.h)
See docs/design/ABI-FFI-README.md for complete documentation.
File:
0-AI-MANIFEST.a2ml(read this FIRST)Purpose: Universal entry point for ALL AI agents
Key invariant: Typell serves PanLL — never competes
typell/ ├── 0-AI-MANIFEST.a2ml # AI manifest (start here) ├── spec/ # Formal specifications │ ├── protocol/ # Verification Protocol spec │ ├── type-system/ # Type system specs │ └── proof/ # Proof system specs ├── src/ │ ├── abi/ # Idris2 formal specs │ ├── kernel/ # Rust verification kernel │ │ ├── checker/ # Bidirectional type checker │ │ ├── proof/ # Proof engine │ │ ├── effects/ # Effect tracker │ │ ├── session/ # Session protocol manager │ │ └── protocol/ # JSON-RPC server │ └── backends/ # Language backends │ ├── vcl/ # VCL-dt++ (VeriSimDB) │ ├── gql/ # GQL-dt++ (LithoGlyph) │ └── kql/ # KQL-dt++ (QuandleDB) ├── ffi/zig/ # Zig FFI implementation ├── generated/abi/ # Auto-generated C headers ├── integrations/ # Consumer integrations │ ├── panll/ # PanLL (primary) │ ├── vscode/ # VS Code extension │ ├── cli/ # Command-line interface │ └── ci/ # CI/CD plugins ├── docs/design/ # Design documents ├── tests/ # Test suite ├── container/ # Stapeln container ecosystem └── .machine_readable/ # Machine-readable metadata
You don’t need to read this README to get started. Just say this to any AI:
Set up Typell from https://github.com/hyperpolymath/typellThe normal installation instructions remain below for anyone who prefers manual setup.
MPL-2.0
Copyright (c) 2026 Jonathan D.A. Jewell (hyperpolymath) <j.d.a.jewell@open.ac.uk>