Collaborate · Architecture · Contribute
How MESSAI Works
Architecture reference, tech stack inventory, persona playbooks, and first-day setup guide for contributors and integration partners.
Platform at a glance
MESSAI is an Nx 21 monorepo with a Next.js 15 web app, 9 open-source MESS-* packages, a Python ML engine (PyMC, GP-SCM, Nougat), and a Supabase Postgres database. Everything runs on Vercel + Supabase in production.
23,470
Papers
ResearchPaper
195,846
Extractions
ExtractedParameterData
825
Canonical params
ParameterDefinition
2,812
DAG edges
ParameterEdge
7,776
PDFs in R2
20 GB
45
AI chat tools
4 subsets
32
LLM models wired
8 providers
21
3D reactor models
plus 7 archetypes
30
Electrode materials
16 anode · 14 cathode
179
API routes
app/api/**
9
Open packages
mess-* mirrors
7
Per-class predictors
MFC/MEC/MES/MDC/MMRC/MNRC/MBES
Open source
9 MESS-* packages under open-source/mess-*/ in this monorepo. Public read-only mirrors at github.com/Messai-io. Apache 2.0 license.