Add production-ready Rust coordinator migration path

[Gajesh2007]

Summary

Adds the production Rust coordinator and a safe migration path while retaining the Go coordinator as a rollback-compatible fallback. This includes frozen cross-language contracts, protocol v2 and Swift provider support, durable owner-fenced billing/recovery, full HTTP surface parity, external migrations, dual-stack deployment, observability, fault/load validation, and signed cutover gates.

Before

flowchart LR
  subgraph Behavior_Before[Behavior]
    C1[Consumer request] --> G1[Go coordinator]
    G1 --> V1[Swift provider protocol v1]
    V1 --> G1 --> C1
  end
  subgraph Code_Before[Code]
    H1[Go HTTP handlers] --> R1[Go registry and routing]
    R1 --> S1[Go store and billing]
    S1 --> D1[PostgreSQL]
  end

Loading

After

flowchart LR
  subgraph Behavior_After[Behavior]
    C2[Consumer request] --> A2[Rust API adapters]
    A2 --> F2[FleetActor and RequestTask]
    F2 --> V2[Swift provider protocol v2]
    V2 --> J2[Signed durable terminal]
    J2 --> L2[Atomic ledger settlement]
    L2 --> C2
  end
  subgraph Code_After[Code]
    P2[Versioned protocol and pure reducers] --> W2[Supervised Rust warm plane]
    W2 --> Q2[Durable jobs ledger and recovery]
    Q2 --> D2[Externally migrated PostgreSQL]
    O2[Ownership handoff deploy and cutover gates] --> W2
    O2 --> G2[Go guarded fallback]
  end

Loading

Linked issue

Not linked.

Test plan

• go test ./coordinator/... -count=1
• Real-PostgreSQL store, migration, ownership, rollback, ledger, and recovery suites
• cargo fmt --all -- --check
• cargo clippy --workspace --all-targets --all-features --locked -- -D warnings
• cargo test --workspace --all-features --locked
• cargo build --workspace --all-features --locked
• SQLx offline metadata check against the latest migrated schema
• make contracts-check
• make coordinator-rs-fault-test
• Python pilot/load/cutover regression suites and shell cutover security tests
• Swift protocol files parsed/typechecked during implementation; full MLX-backed swift test remains a macOS CI/hardware gate
• Docker dual-stack image build and rollback rehearsal (added as CI validation; Docker unavailable in the local agent VM)

Components touched

• coordinator (Go)
• coordinator-rs (Rust replacement)
• provider (Rust, legacy)
• provider-swift (Swift CLI)
• console-ui (Next.js)
• enclave (Swift)
• infra / CI / release
• docs

Protocol / interface changes

• No protocol/interface changes
• Yes — protocol v2, negotiation, binary framing, durable terminal/ACK, replay fencing, HTTP/config/CLI surfaces, and matching Rust/Swift contract tests are included

Notes for reviewers

• Go remains built into the dual-stack image and can serve only after the additive-schema rollback guard passes.
• Serving startup performs no schema mutation; coordinator-migrate owns all DDL/checksums.
• Paid starts are single-attempt and durable: reserve → prepare → resize/authorize → start → signed terminal → atomic settlement → ACK.
• Essential tasks, mailboxes, provider writers, request pipes, DB workers, telemetry, and external calls are bounded and supervised.
• No production deploy or cutover was performed. Remaining operational gates are intentionally external: macOS MLX hardware validation, Docker rehearsal, reviewed scheduled soak baseline, staged human-approved cutover/bakes, and 90-day Go-retirement evidence.

  

---

^{Need help on this PR? Tag /codesmith with what you need. Autofix is disabled.}

[vercel]

The latest updates on your projects. Learn more about Vercel for GitHub.

Project	Deployment	Actions	Updated (UTC)
d-inference	Ready	Preview	Jul 13, 2026 6:52am
d-inference-console-ui-dev	Ready	Preview	Jul 13, 2026 6:52am
d-inference-landing	Ready	Preview	Jul 13, 2026 6:52am