walrus-verity 0.1.4

Proof-chain registry for AI agents — backed by Walrus for persistent, portable, verifiable memory

verity

A proof-chain registry for AI agents — backed by Walrus for persistent, portable, verifiable memory.

---

What it does

verity gives AI agents structured, portable memory:

feature → claim → test → evidence → release
                                       │
                              verity push ──► Walrus blob ID (immutable)
                              verity pull ◄── restore anywhere, any agent

1. Proof-chain registry — track what an agent claimed, what it tested, and what it proved, all in a single verity.json file.
2. Agent memory layer — push the registry to Walrus (or MemWal) and pull it back in any future session, on any machine, by any agent.

Built for the Sui Overflow hackathon, Walrus track.

---

Install

pip install walrus-verity

# With MemWal support
pip install "walrus-verity[memwal]"

---

Quick start

verity init --repo-id repo:my-project

verity add feature feat:auth "User authentication"
verity add claim   clm:auth.t1 "Login succeeds" --feature feat:auth
verity add test    tst:auth.unit "Unit test" --claim clm:auth.t1 --kind unit --path tests/test_auth.py
verity add evidence evd:auth.run1 "CI run" --test tst:auth.unit --artifact artifacts/run1.json --status passed

verity validate          # → OK
verity release 1.0.0     # fail-closed — all verified claims need passed evidence
verity push              # → blob: AbCdEfGh…

# Any agent, any machine, any future session:
verity pull AbCdEfGh…

$ verity validate
OK

$ verity release 1.0.0
Released rel:1.0.0 at 2026-05-23T19:08:25Z
  claims: clm:auth.t1

$ verity push
blob: AbCdEfGhIjKlMnOpQrStUvWxYz0123456789

$ verity log
  1.  [walrus]  2026-05-23T19:08:25Z  AbCdEfGhIjKlMnOpQrStUvWxYz0123456789

---

Documentation

Topic
CLI Reference	All commands: init, add, validate, release, push, pull, log, site
Python API	VeritySession, low-level functions, custom backends
Schema Reference	verity.json fields, ID prefixes, status values, validation rules
Walrus Setup	Testnet, mainnet, custom endpoints
MemWal Setup	Env vars, delegate keys, namespace isolation
Multi-Agent Patterns	Handoff pattern, audit trail, dry-run
Scripts	store_project_context.py — seed MemWal with verity project knowledge

---

Acknowledgements

The proof-chain model — feature → claim → test → evidence → release — is directly inspired by the ssot-registry project, licensed under Apache 2.0. verity adapts that model for AI agents and Walrus-backed persistence.

---

Contributing

See CONTRIBUTING.md. Run tests with uv run pytest.

License

MIT — see LICENSE.