Futures-grade market data that never touches exchange-licensed feeds.
Signed, reproducible and free to redistribute.
Website · Open archive · Architecture · Roadmap · Admission policy · Contributing
Status: Phases 0–2 active; live and historical services in development. See the roadmap and the changelog.
Exchange groups control their prices through data contracts. Anyone who receives those prices, or data derived from them, inherits the contract. This project takes a different route. It collects prices for the same economic exposure from places where that exposure trades in the open, and combines them with published, reproducible methods:
- on-chain order-book venues, read from our own nodes
- swap trades that US law requires to be published free of charge
- public-domain cash-market and rates data from US agencies
Arbitrage keeps these prices close to exchange prices. The link is economic, not contractual. No exchange-licensed data, and nothing derived from it, ever enters the repository, the archive or the network.
- Tapes: trade-by-trade tapes with native aggressor side, for every market on each venue, signed block by block
- Books: order-by-order books, for the venues that publish them
- Order flow: open footprint, delta, volume profile and heatmaps
- Bars and archive: bars from 1 second to 1 day, and a versioned archive that anyone can download over HTTPS, BitTorrent or IPFS and verify against signed manifests anchored on-chain
- Composites: composite prices per exposure (for example "US large-cap index front", "WTI front", "gold")
- Dated-contract equivalents for perpetual markets, from an open cost-of-carry model
- Open reference close: an open daily reference close, recomputable by anyone, and an on-chain oracle with no exchange data in its lineage
- Rates and positioning: rate curves, cash-market prices and positioning data
- Open instrument registry: contract specs, rolls, session calendars
It does not aim to provide exchange prices, official settlements, or execution-grade marks for positions held at an exchange. See docs/requirements.md.
- Clean lineage over coverage. A source joins only if nothing upstream of it carries an exchange data contract. See docs/admission-policy.md.
- Provenance on every record. Source, venue, license and lineage travel with the data. Node operators sign what they publish.
- Reproducible outputs. Composites, conversions and settlements use published methods, and anyone can recompute them from the archive.
- Honest labels. Proxies are called proxies. The basis to the reference market is measured and published, not hidden.
- No single point of control. Active-active ingest, signed blocks that any relay can carry, and a content-addressed archive from day one; independent operators as the mesh federates. No project token.
You need rustup (it installs the pinned toolchain from
rust-toolchain.toml automatically) and a C++ compiler
with make, for the bundled DuckDB and zstd builds. A full build takes a
while the first time.
git clone https://github.com/codeurge/bleh.xyz.git
cd bleh.xyz
cargo build --workspace
cargo test --workspaceRun a local gateway on a synthetic feed and stream from it with the reference client:
# Terminal 1: demo gateway on 127.0.0.1:8080, signed with an in-memory key
cargo run -p gateway -- demo --ephemeral-key
# Terminal 2: stream trades for 60 seconds
cargo run -p mesh-client -- --url ws://127.0.0.1:8080/v0/ws --sub '*|trade' --duration 60Verify a day of the public archive, as a mirror would (archive runbook, manifest spec):
cargo run -p origin --bin archive -- verify \
--url https://data.bleh.xyz/reference-normalized --archive reference-normalized \
--date 2026-10-28 --trusted 0x<signer address>Build the service container images from a commit (reproducible, tagged by git SHA; see deploy/):
deploy/docker/build.sh # every target, tagged with HEAD's SHA
deploy/docker/build.sh origin archive # selected targetsThe reference-normalized archive is public at
data.bleh.xyz/reference-normalized:
signed daily manifests (ADR 0017)
with OpenTimestamps proofs, BitTorrent v2 magnets in
_manifests/magnets.txt,
IPFS CIDs and static Iceberg metadata. Data is CC0; some admitted sources
carry a notice that must stay with their values. Read the
terms of use before you redistribute.
| Specification | Contents |
|---|---|
| observation-v0 | Record model (JSON Schema, Arrow schema) |
| gateway-v0 | Live WebSocket protocol: subscriptions, signed blocks, inclusion proofs |
| manifest-v0 | Signed daily archive manifests and verification |
| archive-layout-v0 | Archive file layout |
| registry-v0 | Instrument registry (JSON Schema) |
| docs/specs/ | Every spec and method |
| Path | Contents |
|---|---|
| crates/ | Rust workspace: origin (node reader, book builder, block signer, archive CLI), gateway, bus, book, archive, history-api, swap-pollers, reference-pollers, mesh-types, contracts |
| tools/ | mesh-client (reference client), hl-synth (synthetic feed), hl-backfill |
| fixtures/synthetic/ | Synthetic test data; the only place data files may live without an admission |
| deploy/ | Container images, compose profiles, Terraform, systemd units, observability |
| docs/ | Design documents, specs, protocol, runbooks, decisions, admissions |
| brand/ | Logo, palette, site and design system |
The layout is decided in ADR 0007.
| Document | Contents |
|---|---|
| docs/requirements.md | What consumers need, and the targets this project sets |
| docs/sources.md | The source register: tiers, terms, and what still needs checking |
| docs/admission-policy.md | Rules a source must meet to join |
| docs/terms.md | Terms of use and disclaimer for bleh.xyz and its data |
| docs/architecture.md | Data model, nodes, composites, settlement, transport, archive |
| docs/coverage.md | Expected coverage by asset class |
| docs/roadmap.md | Phases, exit criteria, open questions |
| docs/operations/ | Runbooks for every service, deploys, keys and incidents |
| Record | Contents |
|---|---|
| docs/decisions/ | Decision records (ADRs) |
| docs/admissions/ | Source admission register; Hyperliquid node and archive fields are admitted |
| .agents/plans/ | Tracked plans; Phases 0–2 are active |
| CHANGELOG.md | Notable changes |
Contributions are welcome. Read CONTRIBUTING.md and AGENTS.md first.
- Changes go through pull requests (the repository owner may push
directly). Run
scripts/check.shbefore you open one; CI runs the same checks. - Commits follow Conventional Commits
and need a DCO sign-off (
git commit -s). - A new source needs an admission record before any code that ingests it can merge.
- Never contribute exchange-licensed data in any form, not even a small sample for a test.
Everyone taking part follows the code of conduct.
Report vulnerabilities and suspected data contamination privately, as described in SECURITY.md. Do not open a public issue.
Code is licensed under Apache-2.0. Data, registry entries and method documents are dedicated to the public domain under CC0 1.0. See ADR 0002. Records from some admitted sources carry a notice that must stay with the values; see the terms of use.