posted an update

What Research OS Is Research OS is described as “a governed AI research workspace that turns source material into an inspectable chain of evidence, claims, analysis, verification, and a final publication.”

Its core purpose is to produce trustworthy research artifacts, not just answers. Every run produces:

Typed Evidence

Claims

Analysis

Verification results

A Markdown Publication

A SHA‑256‑backed Research Receipt

The system rejects any model output that cites sources outside the approved set: “If GPT‑5.6 cites a source outside the approved project source set, Research OS rejects the result and rolls back the entire workflow.”

Architectural Foundations The document outlines a deliberately layered architecture:

Frontend & Interface Streamlit dashboard (current UI)

Planned migration to full FastAPI‑only UI integration

Backend FastAPI (typed, versioned API)

Pydantic (structured output validation)

SQLAlchemy 2 + repository/unit‑of‑work

SQLite (hackathon build), migration‑ready for PostgreSQL

Alembic migrations

AI Governance Layer Provider‑neutral AI interface + registry

GPT‑5.6 via OpenAI Responses API

Constitutional runtime enforcing:

model constraints

token limits

workflow authorization

GitHub Integration GitHub App authentication

Repository metadata, README, Issues, PRs imported as research sources

Persistence sink exists; full workflow integration is WIP

Governance & Provenance Atomic transaction boundary at API request

Adversarial tests verifying rollback on unauthorized citations

Typed lineage records: Source → Evidence → Claim → Analysis → Verification → Publication → Receipt

Verified Features The document lists a large set of completed features, including:

Full FastAPI backend with OpenAPI docs

SQLAlchemy domain models + reversible migrations

Provenance protection + rollback

Modular GitHub integration

JWT‑ready auth boundary

25 passing tests, 93% coverage

Dockerized deployment (non‑root container)

Mythar integration (ISF v0.4 validation)

CHEA Ω∞ Level 0 substrate:

Execution Envelopes

Runtime Broker

Broker Selection Records

Constitutional Execution Records

Public judge demo deployment

Narrated YouTube demo

Current Work in Progress PR #3 blocked by upstream ruleset requiring a nonexistent “Main” check

Judge dashboard, Quick Start, Docker deployment, Mythar integration, CHEA substrate already published in fork

GitHub import → persistence → workflow integration not yet fully claimed as complete

Public judge deployment live at research‑os‑judges.streamlit.app

Planned Milestones Fix upstream required‑check configuration and merge PR #3

Complete GitHub import persistence integration

Add request/response examples to all OpenAPI endpoints

Implement role‑based auth

Add background jobs for GitHub sync

Expand Docker deployment into full Compose stack

Promote judge deployment to upstream v0.1.0 after merge

Add URL/document/academic‑paper adapters

Add human approval checkpoints

Add evaluation suites, observability, PostgreSQL

Add managed signing keys for cryptographically signed receipts

Add isolated sandbox adapters + hardware/multi‑party attestation for higher CHEA assurance levels

Why It Matters Research OS argues that AI research must be both fast and accountable, and that evidence‑first workflows with strict provenance controls are the path to trustworthy AI‑assisted research.

Log in or sign up for Devpost to join the conversation.