EcoIQ LegacySafe AI
EcoIQ already existed as a climate intelligence and industrial modernisation platform. On 2026-07-01 we started building LegacySafe AI as a new AI agents module inside EcoIQ: a permission-aware, agentic change-management layer for enterprise legacy systems.
Demo project: Samruk Energy Legacy Modernisation Demo · Samruk Energy
Model-agnostic by design: switch providers without weakening permission-aware retrieval.
Judge Demo Checklist
Everything below is verifiable live, in this module, in under 3 minutes.
- Same question, different roles — different correct answers
- Restricted evidence named and excluded, never silently dropped
- Source revoked → its chunks revoked → its derived memory revoked
- Audit log entry created for every ask / permission-demo / revoke
- Dependency & lineage graph, built with NetworkX
- Prompt-injection document treated as content, not instruction
Conduct Legacy Modernisation
Reads legacy documents and produces a controlled modernisation plan from only the allowed evidence.
BASEDAIBasedAI Permission-Aware Memory
Every chunk and derived memory carries its own access level — public, engineering, finance, or executive.
BASEDAIDeterministic Retrieval Guard
Access is enforced before retrieval by a pure function — never by asking an LLM for permission.
BASEDAILineage + Revocation
Revoke a source and every chunk and derived memory built from it goes dark in the same request.
BASEDAIAudit Logs
Every question, retrieval, blocked source, allowed source, and revocation is logged as it happens.
CONDUCTDependency Graph
NetworkX-built map of the project, its documents, and derived memories, exported as JSON.
CONDUCTFull EcoIQ Modernisation Plan
Plans clean energy generation, heat replacement, equipment upgrades, grid optimisation, predictive maintenance, finance, procurement, worker transition, and Justice/Maqasid impact.
Why Conduct
- Reads legacy documents, code snippets, process manuals, ESG reports, and policy documents
- Maps dependencies between systems, documents, teams, risks, and required changes
- Produces controlled change proposals and modernisation plans
- Supports human-in-the-loop approval
Why BasedAI
- Permission-aware memory — access enforced before retrieval, not after generation
- Deterministic permission checks — never an LLM decision
- Source lineage tracked from raw document to derived memory
- Revocation propagates: source → chunks → derived memories
- Audit log for every question, retrieval, blocked source, and allowed source
EcoIQ's core platform, brand, and proprietary logic (including the Justice & Maqasid Intelligence framework) are intended to remain founder/company-owned IP, subject to legal registration and professional advice. Selected public-benefit components — such as educational templates and demo data — may be shared separately for NGOs, schools, and community use under an appropriate open or community-use licence.