Available now
- Private PDF uploads and saved review results.
- Reducto r-1 parsing with word/cell source coordinates and image-backed Deep Extract for structural fields.
- AI comparison of a selected design reference and shop drawing submission.
- Nine trade/review modules, multiple supporting sources and additional submittal PDFs.
- Separate project libraries with linked, independently stored revisions.
- Assigner/reviewer fields, evidence checklists and manual review stamps.
- Page citations, source quotations, reviewer notes, and recorded decisions.
- PDF review exports.
Accounts and workspaces
Create an account with your email address and password, then confirm your email. Each account opens a private structural workspace. Projects, document revisions, review decisions and PDF delivery are checked against workspace ownership. Passwords and email verification are managed by Auth0. Trace stores an opaque HttpOnly session cookie. Signing out revokes that application session. The account guide covers verification and password recovery.Review boundaries
Trace can miss discrepancies, misread a drawing, or produce an incorrect finding. It does not perform structural capacity calculations, certify code compliance, or autonomously issue fabrication approval. Manual review stamps record the assigned engineer’s decision. A qualified engineer must verify all findings and complete the required engineering review. Steel member, weld and plate discrepancies were tested with fictional PDFs, including delegated source selection and an existing-work exclusion. Real steel, concrete, wood-on-podium and wood-on-slab validation remains pending sample documents from the engineering team. Other modules are implemented but have not been field-validated.Current limits
Keep the review tab open. Results are saved after successful completion. If a review is interrupted, first check for a saved result before retrying.
Storage and AI processing
Documents and results are stored in Supabase. With document reading enabled, starting AI document review sends selected PDFs to Reducto for parsing and Deep Extract. Saved extractions are stored privately and reused for the same document revision. OpenAI then receives the original PDFs and extracted evidence for comparison. If Reducto is unconfigured in a development installation, AI reviews send the original PDFs directly to OpenAI. The application requestsstore: false; this setting alone is not a promise of zero data retention across all providers.
Both provider API keys stay on the server. Reducto r-1 is currently a preview model. Low-confidence and unlocated extracted values remain subject to engineer verification. Moving a document to Trash retains its PDF and review history. Permanent removal requires an authorized support request. Contact the workspace owner for retention and permanent-removal requirements.