[Grohle]

Summary

Complete implementation of VeriModel: a local-first IFC quality gate + quantity takeoff + BCF export, as a Rust engine with a static HTML interface (Tauri v2) and a CLI sharing the same core.

• crates/core (pure Rust, no IFC dependency): data model, project.yaml schema with validation + JSON schema, YAML rule packs with 20 check kinds and $config.* parameters, config-generated rules, documented score/verdict, IDS evaluation (requirement quoted), QTO aggregation with unit-rate costing and hand-check samples, BCF 2.1/3.0 writer validated against the official buildingSMART XSDs, report writers (report.html/json, failures.csv, qto.csv/xlsx/json), run ledger with SHA-256 hashes, append-only audit log. Deterministic outputs (enforced by tests).
• crates/ifc_reader: the only crate that touches IFC. An embedded IfcOpenShell extractor streams NDJSON facts (header, units converted to SI, spatial tree, elements, psets, quantities, classification, materials, geometry metrics, IDS results) to Rust. IFC2x3 / IFC4 / IFC4.3 verified.
• crates/assistant: optional read-only copilot (small GGUF model via Candle, downloaded on demand), BM25 retrieval over built-in docs, imported documentation (docs → knowledge.yaml converter) and the current report; answers are validated against the report before the UI can apply them. Never part of a verdict.
• src-tauri + frontend/: 3-step Load → Review → Export UI, EN/ES, dark/light, metric/imperial, filters, pagination, element detail, run history with diffs, assistant panel, demo model that loads in one click.
• crates/cli: check, qto, clash (explicit "not implemented"), validate-config, inspect, doctor, ask, docs-to-yaml.
• examples/: residential / office / infrastructure project.yaml, custom rule pack, IDS 1.0 sample, rates.csv, generated demo pavilion.
• tests/: end-to-end on real IFC fixtures (golden expectations on a deliberately broken model, QTO reconciliation, genericity with two configs, determinism, IDS, three schemas) plus unit tests in every crate.
• CI (fmt, clippy -D warnings, tests, CLI smoke run, bundles) and a tag-triggered release workflow producing Windows .msi/.exe, macOS .dmg, Linux .AppImage/.deb and CLI binaries.
• Licence: PolyForm Noncommercial 1.0.0.

Test plan

• cargo test --workspace green (core 60, reader 8, e2e 10, CLI 15, desktop 19, assistant)
• cargo clippy --workspace --all-targets -- -D warnings clean
• CLI runs on all fixtures; broken fixture exits 1 (FAIL) with the expected findings
• BCF opened with bcf-client and validated against BCF-XML 2.1 and 3.0 XSDs
• Frontend walkthrough in headless Chromium with zero console errors
• cargo tauri build -b deb produces VeriModel_1.0.0_amd64.deb
• Release workflow run on tag v1.0.0

---

Generated by Claude Code

[Grohle]

CI status: the Rust tests (core, reader, end-to-end) job is red only at the cargo fmt --all -- --check step. The unformatted files are src-tauri/src/{ops,lib,error,logging}.rs and crates/cli/src/{main,render,error}.rs; all tests and clippy pass locally. The rustfmt fix rides the next push, which also brings the assistant panel and the header/logo work currently in progress on those same files.

---

Generated by Claude Code

[Grohle]

CI update: rustfmt is now clean. The remaining red is cargo clippy -D warnings on two spots in crates/assistant (manual_range_contains in knowledge.rs, large_enum_variant on Task in tasks.rs); CI's clippy (1.99) is stricter than the local one (1.97). Both fixes are in the assistant backend change that lands with the next push.

---

Generated by Claude Code