Skip to content

[ADR] Select runtime, store, queue, migration tooling, and repository layout #12

Closed

Description

@gnanirahulnutakki

Outcome

Choose an implementation stack that can meet the established correctness and operational bounds.

Scope

  • Compare runtime/framework, relational store, queue, migration tooling, module boundaries, and supported versions.

Acceptance criteria

  • Alternatives are assessed for correctness, recovery, security, performance, operability, and monthly baseline cost.
  • Replaceable store and queue interfaces are identified.
  • The accepted ADR pins supported toolchain versions.

Test and evidence

  • Accepted ADR with rejected alternatives and benchmark assumptions.

Dependencies

  • Blocked by #11.

References

Activity

  1. added
    priority:p0Critical prerequisite or correctness boundary
    decision-requiredRequires an explicit architecture or governance decision
    area:foundationGovernance, architecture, and repository foundation
    on Sep 1, 2026
  2. gnanirahulnutakki commented on Sep 1, 2026

    @gnanirahulnutakki
    MemberAuthor

    Completion evidence — 2026-09-01

    ADR 0002 is accepted and merged through PR #69:
    #69

    Acceptance mapping

    • Alternatives: runtime/framework, relational store, queue, migration tooling, and module boundaries are assessed across correctness, recovery, security, performance, operability, and monthly baseline cost. Rejected choices and their assumptions are recorded in ADR 0002.
    • Replaceable interfaces: ADR 0002 defines the StateStore and ReconciliationQueue capability and failure contracts, including atomic accepted-work recording and deliberately weak at-least-once/unordered delivery semantics.
    • Supported pins: Go 1.27.0, PostgreSQL 18.6, pgx 5.10.0, sqlc 1.31.1, and goose 3.27.3 are recorded in the ADR and machine-readable version manifest with primary evidence.
    • Benchmark evidence: the ADR binds qualification to ADR 0001's exact topology, data occupancy, request mix, payload distribution, latency, failure, recovery, queue, and cost assumptions rather than external microbenchmarks.

    Exact merge proof

    Validation on the merge commit

    • shellcheck docs/architecture/stack-evaluation/verify.sh
    • docs/architecture/stack-evaluation/verify.sh — passed
    • docs/architecture/cost-model/verify.sh — passed
    • Markdown lint passed with only repository-existing MD013 long-table/URL handling excluded
    • git diff --check — passed
    • every external ADR and version-manifest link returned HTTP success on 2026-09-01
    • zero unresolved review threads

    Reviews

    • Codex code review completed cleanly on 5ea22b2.
    • Codex security review completed cleanly on 5ea22b2 with no security issues.
    • The configured CodeRabbit context passed. CodeRabbit declined a substantive retry because its incremental review allowance was rate-limited; it produced no findings. This limit was not represented as a substantive approval.

    Cost boundary

    The selected PostgreSQL plus SQS envelope is USD 355.83/month small and USD 682.13/month target, already included in ADR 0001. The whole profiles remain USD 937.50 and USD 2,627.40, leaving USD 62.50 and USD 22.60 headroom. No recurring service was added.

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Metadata

Metadata

Assignees

No one assigned

    Labels

    area:foundationGovernance, architecture, and repository foundationpriority:p0Critical prerequisite or correctness boundary

    Type

    Projects

    No projects

      Relationships

      None yet

      Development

      No branches or pull requests

      Issue actions