Neutral DeFi Risk Intelligence Aggregator
A neutral, open aggregator of what DeFi risk feeds say about Ethereum protocols — side by side, verbatim, with no composite scoring.
DeFi risk intelligence is fragmented. Before deploying capital, users — from retail participants to institutional allocators — must navigate dozens of dashboards, rating services, and research feeds. The information exists; there is no neutral layer that puts it in one place. This project is that layer: for each major Ethereum protocol, it shows the coverage status of every risk feed in the registry, the governance facts (with provenance), and what each feed says, verbatim.
The mental model is oracle diversity: no single feed is canonical for something this
important. The aggregation is the value. We add no synthesis, no editorial judgement, and
no score of our own. See CHARTER.md.
Status: prototype. This repository is an early, in-development reference built for the Ethereum Foundation App Relations Neutral DeFi Risk Intelligence Aggregator RFP. Feed ratings shown in the UI are currently illustrative samples pending provider verification and are labelled as such. The protocol set, design, data model, and governance docs are real.
- A protocol × feed coverage matrix for the top Ethereum DeFi protocols by funds at risk (TVL, or volume where TVL is not applicable), sourced from DefiLlama.
- A per-protocol detail page (tabbed): governance facts with provenance tags, one feed card per provider (methodology one-liner + coverage status + verbatim rating + source link), an on-chain control timeline, incident and audit history, and — unique to this project — a transaction-transparency link into the protocol's live on-chain activity.
- An AI intelligence layer (preview) — the first tab on each detail page. A neutral,
citation-grounded per-protocol assessment of the registry's own sourced facts, plus a
living "what changed" investigative note (who / when / why / what) when the underlying data
moves. It never produces a score, grade, or ranking — an explainer of facts, not a judge,
cordoned off as a clearly-labelled, non-authoritative layer (charter-safe by construction).
Currently an illustrative mockup; the evaluation harness that keeps it neutral, and the rest of
its roadmap, is in
docs/ROADMAP.md§2. - A methodology page stating exactly what the project does and does not do, the full feed registry, and the data-provenance tags.
- A plain-language FAQ answering what a DeFi risk feed is, what coverage means, and why the registry aggregates feeds rather than scoring them — each answer restating the no-composite-scoring stance.
- Not a rating agency. We never produce a composite or proprietary risk score. We show
what others publish. The constraint is binding and documented in
CHARTER.md. - Not editorial. Where feeds disagree, we show the disagreement. We do not adjudicate.
| Layer | File | Source of truth |
|---|---|---|
| Seed protocols + metadata | lib/data/protocols.ts |
Hand-curated; ranked against DefiLlama |
| Live TVL | data/tvl-snapshot.json |
DefiLlama API (scripts/refresh-tvl.mjs) |
| Feed registry | lib/data/feeds.ts |
Hand-curated from the RFP registry + additions |
| Coverage + verbatim ratings | lib/data/coverage.ts |
Curated now; automated per feed at M2 |
| Governance facts | lib/data/protocols.ts (governance) |
On-chain / verifiable where possible |
| AI assessment layer (preview) | lib/data/ai-preview.ts |
Illustrative mockup; live pipeline post-grant |
Every datum carries a provenance tag: onchain, verified (confirmed against a
primary source), self-reported, or sample (illustrative placeholder, prototype only).
Infrastructure. The onchain data and the transaction-transparency links are produced
by Sieve (MIT OR Apache-2.0), our open-source Ethereum
indexer. Sieve syncs directly over the Ethereum P2P network — no RPC provider, no API keys,
no rate limits — into Postgres with a GraphQL API. Because the indexer is open and re-runs
over public P2P, every on-chain figure here is independently reproducible.
This is a community-correctable public good. There are three contribution paths, all via GitHub so every change is reviewable and attributable:
- Correct a fact (a wrong coverage status, a stale rating, a governance detail) —
open an issue with the
correctiontemplate, or edit the relevant file inlib/data/and open a PR. Cite the source. Maintainers verify against the cited source before merge. - Add a protocol — open an issue with the
new-protocoltemplate including the DefiLlama slug and rationale (funds at risk). Accepted protocols are added tolib/data/protocols.ts. - Add or update a feed provider — open an issue with the
new-feedtemplate describing the provider's focus, type, and whether it exposes machine-readable output (which determines automated vs. manual coverage). Seelib/data/feeds.ts.
Review SLA: corrections triaged within 5 business days. A correction is merged only once its cited source has been checked. Merged corrections are credited in the PR and in the contributors list.
npm install
npm run dev # http://localhost:3000
npm run refresh:tvl # pull latest TVL from DefiLlama into data/tvl-snapshot.json
npm run buildAGPL-3.0-or-later. The full codebase and data layer are open and community-correctable.
Maintained by the Rails team (rails.finance) as a public good. See the proposal and
CHARTER.md for the stewardship commitment.