Skip to main content

Digital employees for investment and construction control

CodeGraph combines three specialized roles in a controlled analytical workflow. They assess input quality, identify candidates for CAPEX risk, and.

User Guides

CodeGraph combines three specialized roles in a controlled analytical workflow. They assess input quality, identify candidates for CAPEX risk, and distinguish reported progress from verified execution. The current interface demonstrates this workflow with bundled sample data; adapt the roles and inputs to the target industry process.

Roles and outcomes

Roles and outcomes
Role Job Practical outcome
DASHA — Data Acceptance, Screening & Harmonization Agent Assesses the availability, structure, completeness, and consistency of project data An input-quality profile, data gaps, and boundaries for subsequent analysis
CARA — CAPEX At Risk Agent Relates cost, schedule, physical quantities, and assets CAPEX-risk candidates, mapping conflicts, and assumptions for human review
VERA — Verified Execution Reporting Agent Checks whether claims about actual execution have supporting evidence A confidence level and a list of missing certificates, quantities, or financial evidence

DASHA determines whether input is fit for analysis; the accountable owner certifies the engineering data. CARA raises risk candidates while the budget owner makes budget decisions. VERA combines the schedule percentage with completion evidence.

How the roles work together

A typical sequence starts with DASHA, continues with CARA, and finishes with VERA. The workflow transfers a typed evidence package between steps instead of an unstructured retelling of the previous answer. An operator may also start with an individual role for a focused question.

Each result remains independent. DASHA’s conclusion covers its own analysis, CARA’s observation covers its own scenario, and VERA records the security evidence. Every step exposes its scope, missing inputs, and available follow-up actions.

What the current version demonstrates

The /dashboard/onboarding-research page provides the investment_construction_dynamic preset. For this preset, the interface automatically attaches two bundled demo files: a workbook with project and cost data and a schedule excerpt. Custom file upload is not supported on this page at present.

The bundled data demonstrates role responsibilities and evidence transfer. Its file names and table structure are not required formats for a real project.

Usage boundaries

The pack helps estimators, planners, site-control specialists, and project managers prepare questions for review. Every result requires professional judgment and must not be the sole basis for budget, schedule, payment, or acceptance decisions. Review the execution status and any fallback provenance before using a finding.

Verifiable contract sources

This article is checked against these implementations:

  • src/investment_construction_agents/pack.py — role names and responsibilities;
  • src/investment_construction_agents/dynamic_workflow.py — workflow actions and typed evidence transfer;
  • src/api/services/core_services/chat_service.py — preset execution, statuses, and metadata;
  • frontend/src/pages/OnboardingResearchChat.tsx — the user-facing controls and bundled inputs.