vaid-pop 0.1.0

Canonical Python proof-of-possession (PoP) request signer for the VAID standard.

vaid-pop

Canonical Python proof-of-possession (PoP) request signer for the VAID standard.

This is the single Python definition of the PoP signing contract. A VAID-bound request carries a fresh, replay-protected Ed25519 signature; this package builds the four x-synthera-* headers that carry it. Import it directly — no extra runtime, no framework.

from vaid_pop import RequestSigner

signer = RequestSigner(vaid=vaid_doc, private_key=agent_key)
headers = signer.sign_headers("POST", "/vaid/mint", body_bytes)
# -> {"x-synthera-vaid": ..., "x-synthera-timestamp": ...,
#     "x-synthera-nonce": ..., "x-synthera-signature": ...}

The x-synthera-* header names are the VAID wire contract — the prefix is the fixed header namespace a conforming verifier reads, not a package dependency.

The firewall

Cross-language byte-identity is the whole point. The signer is locked against the frozen cross-language vector (vendored here at vaid_pop/vectors/operator_pop_v1.json), which the Rust client (vaid-client) and the verifier (vaid_pop) assert against too. The pop-conformance CI job proves Rust output == Python output == vector byte-for-byte. A mismatch is a hard blocker.

Contract: RFC 8785 (JCS) over the camelCase RequestAuthPayload → SHA-256 → pure Ed25519 over the 32-byte digest → raw 64-byte signature.

Standardized on rfc8785 for JCS — the byte-for-byte JCS implementation the vector is proven against. Do not substitute another JCS library without re-proving byte-equality.

Relationship to other packages

• vaid-client (Rust) is the cross-language peer, proven against the same vector.
• Language-specific agent integrations depend on this package and re-export RequestSigner.
• Clients consume this directly to authenticate a VAID-bound request.