Contractual Risk Intelligence Engine.

Forge reads your project documents, extracts signals and governing rules, resolves delay attribution, and computes net contractual exposure — automatically.

Beta Notice: Forge is in active development. Features described here are continuously refined based on real-world testing.

The Reasoning Pipeline

Forge operates through a structured reasoning chain. Each layer feeds the next — from raw text to a defensible financial conclusion.

📄 Documents → Signals + Constraints

The engine separates two types of intelligence from every document: Signals (events that change project state, requiring triage) and Constraints (governing rules that define how signals translate into impact — LDs, penalties, weather entitlements, notice windows).

🔗 Activity Linkage

Each signal is semantically matched to its corresponding CPM activity using pgvector similarity search, composite scoring (semantic + temporal + WBS + trade), and configurable confidence thresholds.

⚖️ Responsibility Attribution

The LLM infers the responsible party for each delay signal: CONTRACTOR, OWNER, THIRD_PARTY, or FORCE_MAJEURE. Attribution confidence is tracked and displayed.

🔄 Concurrency Resolution

Delay intervals are projected onto the activity timeline and swept using a sweep-line algorithm. Overlapping windows are resolved into net attribution: contractorDays, ownerDays, concurrentDays.

⚓ Milestone Anchoring

Delay pressure propagates upstream to project milestones. Forge computes Risk Drift — predicting future slippage from emerging signals that have not yet hit the schedule baseline.

💰 Confirmed vs. Potential Exposure

The engine separates Confirmed Exposure (materialized delay) from Potential Exposure (emerging risk signals). Leading indicators are now clearly separated from trailing stats.

🚨 Automated Urgency Scoring

Every risk signal is automatically scored into one of three tiers — Act Now, Monitor, or Watch — based on critical path proximity, float buffer, and contractual severity. This ensures the most time-sensitive issues surface first, regardless of confirmed cost.

Reference

System Architecture →

NestJS modules, reasoning services, and the data model.

Financial Exposure Engine →

Dual exposure modeling, constraint resolver, and concurrency adjustment.

Intelligence Feed →

Signal lifecycle from extraction to triage to causal locking.

Claims Defense →

Evidence trails, attribution records, and export for dispute resolution.

Project Q&A Assistant →

Chat-based interface for querying the project knowledge base.

Safety Pattern Detection →

AI clustering to surface systemic safety risks from observations.

Subcontractor Scorecards →

Automated performance ratings per supplier, derived from observations.

Guides

Getting Started →

Upload your schedule, ingest documents, invite your team.

Triaging Signals →

The triage inbox, CPM impact preview, and accept/dismiss lifecycle.

The Schedule Screen →

Gantt float encoding, baseline overlay, Deviation View, and Impact Drivers.

Supply Chain Intelligence →

Pre-crime monitoring for vendor events that affect your schedule.

From Information System to Reasoning Engine

Traditional construction tools are read-only. They display data — Gantt charts, cost reports, dashboards. You manually synthesise what connects a late concrete pour to a contractual penalty.

Forge crossed a different threshold. It doesn't just display information; it reasons about it. It evaluates contract rules, interprets delay causality, resolves concurrent responsibility, and computes the financial consequence — deterministically. That's what separates analytics from intelligence.

The strategic goal: answer the questions construction executives actually ask — "Which delays are truly ours? What is our real penalty exposure? Are we protected by concurrency? Which signals need attention right now?" — in real time, not retrospectively.