Scenario Lab: source and research notices

This browser edition retrieves public evidence from ClinicalTrials.gov, Europe PMC, UniProt, PubChem and openFDA, freezes a hashed snapshot in IndexedDB, and asks the visitor’s BYOK AI to explore fixed synthetic stakeholder roles. A separate final AI request integrates the evidence, assumptions and discussion into qualified inference, evidence-linked findings and a conceptual drug/vaccine development flow when relevant. No model weights ship with the Scenario Lab.

AI requests go from the browser directly to the provider the visitor configured (Anthropic, OpenAI, Google Gemini, OpenRouter, Hugging Face or a custom OpenAI-compatible endpoint), or through the optional stateless relay at /api/ai-relay.php when the visitor enables it. Keys are kept in the visitor’s browser (tab, session or device, their choice). The provider’s own terms apply to the content sent.

Integration code: AGPL-3.0-or-later. CesiumJS (world view): Apache-2.0, see licence.

FACT means a preserved provider field, not independent clinical confirmation. DERIVED means a documented transformation. ASSUMPTION covers hypotheses and unverified imported source attribution. Generated personas, statements, inference, citations and proposed development stages are SIMULATED. Known citation IDs are checked against the frozen evidence; scientific support and correctness require independent review.

The final request receives up to eight records with bounded field excerpts, explicit assumptions and the saved stakeholder discussion. This is not a systematic review. Development flows are conceptual research plans with proposed deliverables and validation gates. Scenario Lab does not execute chemistry, docking, laboratory synthesis, vaccine formulation, manufacturing or clinical models.

Research only. General-purpose language models are not medically validated. No efficacy, safety, patient outcome, approval, eligibility, dosimetry or geographic prediction is supported.