[Community Fork] MiroFish-IT — European institutionally-calibrated fork (Eurostat, ISTAT, ESS, Hofstede at NUTS-2)

[Blubobo1010]

Hi everyone,

I've been working on a fork that adds institutional calibration for European (specifically Italian) synthetic populations and wanted to share it with the community.

What it does:
MiroFish-IT adds an Institutional Calibration Framework (ICF) on top of the original engine. Each simulated agent receives a regional profile calibrated on real EU institutional data at NUTS-2 granularity:

• Economic layer: Eurostat SDMX (GDP, PPS, employment by NUTS-2 region) + Banca d'Italia IBF (household wealth and savings)
• Cultural layer: Hofstede 6D dimensions + Schwartz values from European Social Survey (Round 10)
• Demographic/behavioral layer: ISTAT regional data (education, lifestyle, internet access)

The result: an agent from Lombardia and one from Campania receive different economic and cultural parameters based on real institutional data — not generic prompting.

What's included:

• calibration_service.py — ICF calibration engine with all 21 Italian NUTS-2 regions
• oasis_profile_generator.py — modified to inject ICF context into agent persona generation
• Pre-processed calibration profiles in JSON (calibration_profiles.json, calibration_texts.json)
• Data pipeline scripts to download/refresh from Eurostat, ISTAT, Banca d'Italia, ESS
• Italian UI translation
• Pre-configured demo scenario: Superbonus 110% regional debate (7 agents across 5 Italian regions)

All data sources are open and free (Eurostat SDMX API, ISTAT REST, Banca d'Italia open data, ESS ERIC open data).

Repo: https://github.com/Blubobo1010/MiroFish-IT

The framework is designed to be replicable for any EU country covered by Eurostat + ESS — the same calibration architecture applies to Germany, France, Spain, etc. with country-specific data.

I'm also working on an academic paper validating the calibration accuracy against historical Italian datasets — will share when it's ready.

Happy to answer questions, and open to contributions from others working on non-English/non-US synthetic populations.

— Andrea

[lordpba]

Hi! Is very similar of what I am doing :)

https://github.com/lordpba/CivesAI

[666ghj]

Note

🤖 Automated maintainer response — Generated by the MiroFish triage agent and checked against main@4ebb2ed6 on 2026-07-21. Posted with maintainer authorization; a maintainer may revise it.

Thanks for sharing MiroFish-IT. Regional and institutional calibration is an interesting community direction.

For clarity, MiroFish-IT is an independently maintained community fork, not an official MiroFish distribution. The maintainers have not audited its datasets, data licenses, calibration methodology, security, validation claims, or compatibility with current official main.

Please pin the upstream MiroFish commit supported by the fork and document data provenance, licensing, validation methodology, and upgrade instructions in its own repository. Users should review and test it independently. We will keep this Discussion open under Show and tell.