Document Index
/documents
Curated, searchable index of publicly available MoH and partner-ministry documents per pilot country.
Ship phase
Governance & module status
Funder-grade governance from v0.0. Every change is a public pull request. Every take-down request is logged with its outcome. From v0.5 you can watch month-to-date AI token spend per module from the same page as the site itself.
Repo
github · public
Licence
text CC-BY · code MIT
DOI service
Zenodo
Take-down SLO
5 business days
The argument
Every feature ships as a discrete module with its own owner, budget, surface, and on/off switch. A funder can sponsor a single module. If any module is turned off, the rest still build. This is the funder protection — and it is the operational protection too, because it means a delay on one piece never blocks the others.
The twenty modules fall into four families: Data spine (the public information layer), People (researchers and funders made routable), AI-powered (frontier models applied in narrow, edited loops), and Operations (the governance, distribution and audit layer that makes the rest credible).
Live · build status
Each module has an owner, a surface, a budget and an on/off switch. A funder can sponsor a single module. If any module is turned off, the rest still build. This board is generated from the content collection — it is the same source of truth the team plans against.
Live in v0.0
8/20
Families
4
Phases
v0.0 → v1.5
The information layer the rest of the platform stands on.
/documents
Curated, searchable index of publicly available MoH and partner-ministry documents per pilot country.
Ship phase
/portfolio
The annual portfolio of country-specific modelling questions co-formulated with each ministry.
Ship phase
/dossiers
Build-ready packages per question — schema, starter pack, supervisor, adviser, budget ask.
Ship phase
/sessions
Dialogue + discovery session notes, with ground rules and outcomes.
Ship phase
/outputs
Open-access working papers, methods notes, policy briefs — DOI-deposited via Zenodo.
Ship phase
/countries
One dashboard page per pilot country — anchor, dossiers, documents, sessions, researchers.
Ship phase
Researchers and funders made visible and routable.
/researchers
Per-researcher profiles — photo, bio, institution, methods, dossiers, open-to-support tag.
Ship phase
/researchers/spotlight
Monthly featured researcher rotating across pilot countries, pushed to partner networks.
Ship phase
/support
Single discoverable surface for funders — every project with a question, researcher, supervisor, budget ask.
Ship phase
Frontier models, applied in narrow loops with editor review.
/tools/triage
Upload a public-domain document → structured summary + 3–5 candidate research questions + draft dossier scaffold. Hand-holding refinement loop. Flagship AI module.
Ship phase
/tools/coauthor
Interactive Claude-powered assistant for a researcher with an assigned dossier — drafts concept-note sections, lit review, methods. Dossier-aware and discipline-aware (Nick lens in the system prompt).
Ship phase
/tools/coach
Guided walkthrough from "I have this question" to "here's the appropriate Bayesian / spatial / time-series approach", with starter-code references and an adviser routing recommendation.
Ship phase
/tools/search
Natural-language search over the indexed corpus with citations. Default is citation-list (deterministic); AI synthesis is opt-in.
Ship phase
/tools/funding-match
Periodically scrapes public funder calls (Wellcome, Mastercard, IDRC, Gates Africa, GCC, Carnegie, AESA, DPI funders) and matches each researcher / dossier to a personalised list, pushed monthly via newsletter.
Ship phase
editor-only
EN↔FR translation pipeline for content authoring; supports bilingual DRC content with reviewer-in-the-loop.
Ship phase
The governance, distribution and audit layer.
Git + Decap CMS
All edits via PR; Decap CMS for non-dev editors from v0.1; auto-deploy on merge. Funder-grade governance from day one.
Ship phase
/about/governance
Honour MoH take-down requests within 5 business days even though documents indexed are already public. Public log of every request.
Ship phase
/about/governance
Live, public, quarterly-refreshed metrics — readership, dossiers, outputs, AI-token spend per module, uptime.
Ship phase
Buttondown
Monthly digest — new dossiers, spotlight researcher, what the ministries are reading. M14 funder matches inserted per-recipient from v1.5.
Ship phase
/api/v1
Read-only public JSON feeds — researchers, dossiers, outputs — so partner directories (MAP, IDEM-LAB, Vector Atlas, AIMS alumni) can ingest.
Ship phase
Source · src/content/modules/ · Updated on every public commit.
Editorial workflow
The audit trail is the governance. Every word on the site is reachable as a public commit with an editor, a timestamp, and a diff.
Draft on a branch
Editor opens a branch — either via git locally, or via Decap CMS in the browser from v0.1. Drafts are markdown in src/content/.
Pull request, public
Every change is a PR against main. AI-generated drafts are tagged and require a human editor before merge.
Review · methods · facts
Lead consultant signs off on every PR. Methods content is additionally reviewed by the named working-modeller adviser for that dossier.
Auto-deploy on merge
Merge triggers a Netlify build. The audit trail is the git history — every change is signed, dated, and addressable.
Take-down policy
Every document we index is already in the public domain. We still honour ministry take-down requests — within five business days, with a public log of every request and its outcome. Trust takes years to build and one uncorrected mistake to erode. The policy is the floor. M17 implements it.
Window
5 days
Working days from request receipt.
Record
Public log
Every request, every outcome — from v0.5.
Standing
Ministry
Originating ministry of the document, or its successor.
Public token-spend
From v0.5, the public analytics dashboard shows month-to-date AI spend per module — M10 triage, M11 co-author, M13 search, M15 translation. The line item, the model, the tokens, the cost. No black box. Funder-grade by default, not by request.
Preview · v0.5
MTD · USD
Values placeholder · live numbers from v0.5
Budget transparency
The $50,000 scenario, split three ways. Engagement is the largest line because the work is fundamentally about people in rooms. Platform and AI are the smallest defensible number. The contingency is real — not aspirational.
v0.0 scenario · 12 months
$50,000
$25k floor also planned
$29,000 of $50k scenario
$12,000 of $50k scenario
$9,000 of $50k scenario
Roadmap
Six releases. Each one ships a real artefact a funder or ministry can read. v0.0 is what you are looking at now.
v0.0
Q3 2026
Funder prototype
Currentv0.1
Q4 2026
First real content
v0.5
Q2 2027
The collaborative is on
v1.0
Q4 2027
Country-led
v1.5
Q2 2028
Compound the network
v2.0
TBD
Continental
Plan updates · committed quarterly via PR · last updated when this page was deployed.