chore: migrate rust/face-recognition to icp-cli

[marc0olo]

Summary

• Replace dfx.json with icp.yaml; custom build pipeline: cargo build --target wasm32-wasip1 → wasi2ic (rewrites WASI syscalls to IC system API) → wasm-opt (size + SIMD optimisation) → ic-wasm (embed Candid metadata); auto-installs wasi2ic and wasm-opt if missing
• Add sync step: upload-models-to-canister.sh runs after each deploy — downloads face detection model, auto-generates face recognition model (if Python 3.9–3.12 available), uploads both to the canister, skips if already loaded
• Restructure source: src/backend/ → backend/; update workspace Cargo.toml
• Upgrade ic-cdk 0.14 → 0.20, ic-stable-structures 0.6 → 0.7, ic-wasi-polyfill 0.4.1 → 0.13; remove unused colorgrad; update edition to "2024"; remove channel = "1.79" and channel = "1.79" pin from rust-toolchain.toml; rename wasm32-wasi → wasm32-wasip1 throughout
• Add models_ready() -> bool query so the frontend can detect unloaded models on page load and show an actionable setup message
• Fix detect() and embedding() in onnx.rs: replace .unwrap() with .ok_or_else() so callers receive a descriptive Err instead of a canister trap when models are not loaded
• Add post_upgrade model reload: reads model files from stable memory (WASI filesystem) with a metadata guard so upgrades don't require re-uploading 100MB+ of model data
• Replace Makefile + run.sh with test.sh; Test 4 is context-aware (models_ready() check) so it passes in both CI (no models) and locally (models loaded)
• Add CI workflow face-recognition.yml (icp-dev-env-rust:1.0.1, installs build-essential for wasi2ic/wasm-opt C++ deps); remove legacy rust-face-recognition-example.yaml
• Frontend: migrate webpack + dfx declarations to Vite + @icp-sdk/bindgen; align vite.config.js and actor.js with canonical icp-cli pattern (no dfx fallbacks); bump @icp-sdk/core 5.0→5.4, @icp-sdk/bindgen 0.2.2→0.4.0, vite 5.4.11→8.1.3; remove "Onchain" jargon from title/heading
• download-face-detection-model.sh: fix target path; force IPv4 (--inet4-only/-4)
• download-face-detection-model.sh: carry over backend/assets/image.png (embedded test image for benchmarking endpoints); update .gitignore to ignore *.onnx at project root
• Rewrite README.md: remove all "smart contract" jargon; add optional Python 3.9–3.12 prerequisite; document 3 deploy phases; explain model persistence vs persons database behaviour; add benchmarking endpoint usage with icp canister logs
• Update ic-file-uploader link to current repo (github.com/decide-ai/ic-file-uploader)
• Delete dfx.json, build.sh, webpack.config.js, run.sh, old src/ tree, old declarations, legacy CI

Test plan

• icp network start -d && icp deploy && bash test.sh passes — models downloaded/generated/uploaded automatically, all 4 tests green
• icp canister logs backend is empty after fresh deploy (no panics)
• Redeployment (icp deploy again) skips model upload (models_ready() returns true)
• Frontend shows setup warning when models not loaded; shows detection UI once loaded
• CI workflow face-recognition passes on push

🤖 Generated with Claude Code

[Copilot]

Pull request overview

Migrates rust/face-recognition from dfx to icp-cli, restructuring the example to the canonical backend/ + frontend/ layout, updating the Rust canister/toolchain, and switching the frontend build from webpack to Vite with @icp-sdk/bindgen.

Changes:

• Replaced dfx.json/Makefile-era flows with icp.yaml build+sync steps, plus new test.sh and model upload/download automation.
• Updated Rust backend dependencies/toolchain (incl. ic-cdk to 0.20, WASI target rename) and added models_ready() for the frontend.
• Migrated frontend to Vite + bindgen and removed legacy webpack + generated declarations.

Reviewed changes

Copilot reviewed 30 out of 42 changed files in this pull request and generated 1 comment.

Show a summary per file

File	Description
rust/face-recognition/webpack.config.js	Removed legacy webpack config (Vite migration).
rust/face-recognition/upload-models-to-canister.sh	Reworked model download/generate/upload logic for icp-cli sync step.
rust/face-recognition/test.sh	Added icp-cli based smoke tests.
rust/face-recognition/src/frontend/src/.ic-assets.json5	Removed legacy asset headers config (old layout).
rust/face-recognition/src/frontend/assets/.ic-assets.json5	Removed legacy asset headers config (old layout).
rust/face-recognition/src/declarations/backend/index.js	Removed dfx-generated JS declarations.
rust/face-recognition/src/declarations/backend/index.d.ts	Removed dfx-generated TS declarations.
rust/face-recognition/src/declarations/backend/backend.did.js	Removed dfx-generated did JS.
rust/face-recognition/src/declarations/backend/backend.did.d.ts	Removed dfx-generated did TS.
rust/face-recognition/src/declarations/backend/backend.did	Removed dfx-generated did file (old location).
rust/face-recognition/rust-toolchain.toml	Updated target list to wasm32-wasip1 and removed channel pin.
rust/face-recognition/README.md	Rewritten docs for icp-cli deploy phases and model persistence behavior.
rust/face-recognition/package.json	Simplified to workspace root delegating to frontend/.
rust/face-recognition/icp.yaml	Added icp-cli canister definitions with custom build pipeline + sync script.
rust/face-recognition/frontend/vite.config.js	Added Vite config + bindgen + dev-server cookie/proxy wiring.
rust/face-recognition/frontend/src/index.js	Updated frontend entry to use new actor + models-ready banner.
rust/face-recognition/frontend/src/actor.js	Added canonical cookie-based actor wiring (no dfx fallback).
rust/face-recognition/frontend/package.json	Added frontend package with Vite + SDK deps.
rust/face-recognition/frontend/main.css	Added/updated styling for new UI assets.
rust/face-recognition/frontend/logo_transparent.png	Added UI image asset.
rust/face-recognition/frontend/logo_small.png	Added UI image asset.
rust/face-recognition/frontend/loader.svg	Added loader asset.
rust/face-recognition/frontend/index.html	Updated page title/heading and loads src/index.js as module.
rust/face-recognition/frontend/favicon.ico	Added favicon asset.
rust/face-recognition/download-face-detection-model.sh	Adjusted download flags (force IPv4) and target path behavior.
rust/face-recognition/dfx.json	Removed dfx project file.
rust/face-recognition/Cargo.toml	Updated workspace member path to backend/.
rust/face-recognition/build.sh	Removed legacy build script (replaced by icp.yaml build steps).
rust/face-recognition/backend/src/storage.rs	Added WASI filesystem helper functions for model files.
rust/face-recognition/backend/src/onnx.rs	Added models_ready() + replaced model unwrap() with descriptive errors.
rust/face-recognition/backend/src/lib.rs	Added canister models_ready() endpoint + post-upgrade model reload + candid export.
rust/face-recognition/backend/src/benchmarking.rs	Improved benchmarking endpoint docs and formatting helper comment.
rust/face-recognition/backend/Cargo.toml	Upgraded deps, removed unused dep, bumped edition, updated versions.
rust/face-recognition/backend/backend.did	Added models_ready() to the Candid interface.
rust/face-recognition/.gitignore	Updated ignores for ONNX and bindgen output; removed dfx artifacts.
rust/face-recognition/.cargo/config.toml	Removed legacy per-target config (moved to icp.yaml build flags).
.github/workflows/rust-face-recognition-example.yaml	Removed legacy dfx-based CI workflow.
.github/workflows/face-recognition.yml	Added new icp-cli-based CI workflow using icp-dev-env-rust:1.0.1.

Comments suppressed due to low confidence (2)

rust/face-recognition/backend/backend.did:50

• setup_models returns Result<(), String> in Rust, but backend/backend.did declares setup_models": () -> (). Since you embed this .did into WASM metadata and also use it for frontend bindgen, the interface will be inconsistent and can break callers. Either change the Rust signature to return () or (recommended) update the .did to match the Result (e.g. variant { Ok: null; Err: Error } / variant { Ok: null; Err: text }).
  rust/face-recognition/backend/src/lib.rs:138
• setup_models currently calls storage::bytes(...) which does std::fs::read(...).unwrap(). If the model files haven't been uploaded yet, this will trap the canister instead of returning a descriptive error. Since setup_models already returns a Result, consider guarding for missing files and returning an Err (and also align the error type with the existing Error { message } candid record for consistency).

---

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.

[Copilot]

Pull request overview

Copilot reviewed 30 out of 42 changed files in this pull request and generated 1 comment.

Comments suppressed due to low confidence (1)

rust/face-recognition/backend/src/lib.rs:168

• post_upgrade currently ignores the result of setup(...). If model reload fails (corrupt/partial files), the canister silently continues with models unloaded, making debugging difficult and potentially causing confusing runtime errors. Logging the error (or trapping explicitly) would make failures visible in canister logs.

[Copilot]

Pull request overview

Copilot reviewed 30 out of 42 changed files in this pull request and generated 1 comment.

Comments suppressed due to low confidence (1)

rust/face-recognition/backend/Cargo.toml:5

• edition = "2024" deviates from the Rust canister examples in this repo (which consistently use edition 2021, e.g. rust/hello_world/backend/Cargo.toml:4). Using 2024 also increases the minimum required Rust toolchain and can break builds in the pinned dev/CI container if it doesn't ship a new-enough rustc. Consider sticking to edition 2021 unless there is a concrete 2024-only requirement.