Thesis slice
Prove the v19 serialization thesis by generating one git-warp boundary transport contract and deterministic codec through the Bunny/Wesley GraphQL profile path.
Why
The v19.0.0 milestone requires serialization boundaries to be deterministic and generated where drift can occur. Bunny/Wesley already proves the GraphQL SDL -> generated Rust/TypeScript DTO -> manifest/witness pattern. git-warp needs one concrete boundary pilot so this becomes release evidence, not architecture intent.
Proposed pilot boundary
Checkpoint fact/shard transport is the preferred first target because it is storage-facing, deterministic, and less entangled with network auth than sync envelopes.
Acceptance evidence
- A git-warp GraphQL boundary schema is checked in for the selected pilot boundary.
- Bunny/Wesley profile generation emits transport DTOs and deterministic codec artifacts or a deliberately scoped first emitter slice with witnesses.
- Generated artifacts include schema hash, generator version, scalar/wire-profile witnesses, and manifest integrity evidence.
- git-warp consumes the generated boundary through an adapter/port;
src/domain/** does not import concrete generated codecs directly.
- The decode/admission path constructs validated runtime domain nouns before behavior branches on data.
- Tests prove deterministic round-trip bytes and reject malformed transport.
Ordering
This should start after #169 removes the remaining domain defaultCodec / defaultCrypto / defaultTrustCrypto default-selection leaks, so the generated codec lands at the boundary instead of becoming a new core dependency leak.
Related: #695, #169.
Thesis slice
Prove the v19 serialization thesis by generating one git-warp boundary transport contract and deterministic codec through the Bunny/Wesley GraphQL profile path.
Why
The v19.0.0 milestone requires serialization boundaries to be deterministic and generated where drift can occur. Bunny/Wesley already proves the GraphQL SDL -> generated Rust/TypeScript DTO -> manifest/witness pattern. git-warp needs one concrete boundary pilot so this becomes release evidence, not architecture intent.
Proposed pilot boundary
Checkpoint fact/shard transport is the preferred first target because it is storage-facing, deterministic, and less entangled with network auth than sync envelopes.
Acceptance evidence
src/domain/**does not import concrete generated codecs directly.Ordering
This should start after #169 removes the remaining domain
defaultCodec/defaultCrypto/defaultTrustCryptodefault-selection leaks, so the generated codec lands at the boundary instead of becoming a new core dependency leak.Related: #695, #169.