Construction Management AI Assistant

Problem it solves

Construction management details are often split across scheduling, quality, safety, labor, and documentation. Engineers frequently face conclusions without sources, numbers without traceability, and inconsistent assumptions across roles. This skill focuses on tying professional judgments to lookupable standards, auditable calculation steps, and verifiable outputs.

How it works

• Aggregated roles: one agent switches among scheduling, quality, safety, labor, and documentation roles according to task intent.
• Standard lookup: it searches references/standards/规范数据库.json or runs authority_search.py to bind conclusions to standard numbers and clauses.
• Engineering calculation: it calls calc_engine.py for CPM, man_days, and labor_cost calculations instead of relying on mental math.
• Role handshakes: team_state.py uses publish/get/handoffs to pass structured data, such as schedule quantities triggering labor headcount calculation.
• Gate validation and reuse: gate_validator.py blocks unsourced numbers and vague wording before output; validated conclusions can be saved through quality_feedback.py.

Boundaries

It does not make major commercial or legal liability decisions for the user. Estimates must be labeled as non-standard, reference-only experience. The current version is a single-agent persona switcher, not a true parallel multi-agent team.