# Mizan An **Application Framework Interface (AFI)** — one decorator on a server function, a typed client generated, invalidation automatic, caching protocol-driven. Any backend, any frontend, one wire protocol. For the wire protocol, package layout, and codegen state, see [`CLAUDE.md`](CLAUDE.md). Architecture deep-dives live in [`docs/`](docs/). Open work is tracked in [`ROADMAP.md`](ROADMAP.md) and [`ISSUES.md`](ISSUES.md). --- ## Backend Adapter Parity **Django is the maximal rubric** — it implements the full AFI surface. Every other adapter is measured against it. A cell is marked supported only when *that adapter* wires the capability into its own dispatch surface (not merely that a shared core primitive exists). Legend: ✅ full · ◑ partial · ❌ absent · — not applicable to this transport | Capability | Django | FastAPI | Rust / Axum | Tauri | TypeScript | |---|:---:|:---:|:---:|:---:|:---:| | RPC call dispatch (`{result, invalidate}`) | ✅ | ✅ | ✅ | ✅ ¹ | ✅ | | Named-context bundle fetch | ✅ | ✅ | ✅ | ✅ | ✅ | | Invalidation — JSON body | ✅ | ✅ | ✅ | ✅ | ✅ | | Invalidation — `X-Mizan-Invalidate` header | ✅ | ❌ | ❌ | — ¹ | ✅ | | Invalidation auto-scoping (three-tier) | ✅ | ✅ | ✅ | ✅ | ✅ | | Origin-side HMAC cache | ✅ | ❌ | ❌ | ❌ | ✅ | | WebSocket channels | ✅ | ❌ | ◑ ² | ❌ | ❌ | | Forms (schema / validate / submit) | ✅ | ❌ | ◑ ³ | ❌ | ❌ | | Formsets | ✅ | ❌ | ❌ | ❌ | ❌ | | API shapes (ORM query projection) ⁴ | ✅ | — | — | — | — | | Auth guards (`auth=True/'staff'/'superuser'/callable`) | ✅ | ✅ | ❌ | ◑ ⁵ | ❌ | | JWT auth (access / refresh, session validation) | ✅ | ❌ | ❌ | ❌ | ❌ | | MWT (edge identity token) | ✅ | ❌ | ❌ | — | ❌ | | SSR bridge | ✅ | ❌ | ❌ | — | ❌ | | PSR (`render_strategy` in manifest) | ✅ | ❌ | ❌ | — | ✅ | | Edge manifest export | ✅ | ❌ | ❌ | — | ✅ | | Codegen IR export (KDL) | ✅ | ✅ | ✅ ⁶ | ✅ ⁶ | ❌ | | Session / CSRF init endpoint | ✅ | ◑ ⁷ | ◑ ⁷ | — | ❌ | | Function discovery / registration | ✅ | ✅ | ✅ | ✅ | ✅ | | Auth-provider integration (allauth) | ✅ | ❌ | ❌ | ❌ | ❌ | **Notes** 1. Tauri's transport is Tauri IPC (a single `#[tauri::command]` envelope), not HTTP. Invalidation rides in the JSON response body; there is no header channel, so the header row is N/A. 2. Rust/Axum declares `Transport::Websocket` in the IR/macro but routes no Axum WebSocket handler yet. 3. Rust/Axum carries `is_form`/`form_role` trait stubs but no validate/submit endpoint. 4. "API shapes" is Django's django-readers queryset projection — ORM-coupled. Every adapter carries typed input/output through the KDL IR; the *projection primitive* itself is Django-only. 5. Tauri's `FunctionSpec` carries `auth`/`private` fields, but the dispatch path does not enforce them. 6. Rust/Axum and Tauri *are* the IR authority via the `#[mizan::client]` macro + linkme registry; the codegen links the crate directly (`build_ir()` / the `export-ir` bin) rather than fetching over HTTP. 7. FastAPI and Rust/Axum expose `GET /session/` returning a null CSRF token for wire parity; real CSRF is Django-only. ### Reading the columns - **FastAPI** — the AFI-common HTTP subset: dispatch, contexts, body invalidation, auth guards, IR export. Channels / forms / SSR / cache are deliberately delegated to native FastAPI equivalents. - **Rust / Axum** — core dispatch + context bundling + compile-time registry, and the *server-side* IR authority. No HTTP-layer cache, auth enforcement, or edge surface yet. - **Tauri** — the same Rust core over IPC for desktop/mobile. Edge, SSR, MWT, and CSRF are structurally inapplicable to a local app. - **TypeScript** — the edge/protocol reference: HMAC cache, edge manifest, and PSR strategy, proving the cache + invalidation protocol is language-agnostic. It is not a codegen IR source. ### Keeping this honest This table is a **snapshot**, and a hand-maintained snapshot drifts the moment an adapter gains or loses a target. The enforcing layer is the AFI conformance suite at [`tests/afi/`](tests/afi/), which today gates **IR-shape parity** — the same fixture through Django, FastAPI, and the Rust adapter must emit byte-identical KDL (`test_codegen_parity.py`). It does **not yet** assert *runtime capability* parity (e.g. "every adapter that claims the header transport actually emits `X-Mizan-Invalidate`," "every adapter that claims `auth=` actually rejects an unauthorized call"). Extending `tests/afi/` with per-capability assertions — one row of this table per asserted behavior — is what turns the table from documentation into a red-on-regression contract.