@proompteng/temporal-bun-sdk 0.8.0 → 0.9.1

package/docs/production-readiness-implementation-plan.md

@@ -0,0 +1,371 @@
+# Temporal Bun SDK Production Readiness Implementation Plan
+
+_Last updated: May 5, 2026_
+
+## Goal
+
+Make `@proompteng/temporal-bun-sdk` the default Temporal choice for Bun-based
+agents by turning the current project-proven runtime into a public,
+repeatable, evidence-backed production library.
+
+This is not a plan to load the official `@temporalio/worker` stack in Bun. The
+SDK is a pure Bun implementation: package metadata only ships `dist`, `docs`,
+`skills`, and `README.md`; package dependencies are protobuf/connect/Effect/TypeScript; and the
+production verification gate rejects `@temporalio/worker`,
+`@temporalio/core-bridge`, `node-gyp`, `dist/native`, and stale native Docker
+paths.
+
+## Actual Concern
+
+The concern behind "not production ready" is not that Bun cannot load a Node
+NAPI bridge. This SDK does not rely on that bridge.
+
+The real concern is that Temporal worker correctness is a protocol and
+determinism contract. The official SDK gets trust from Temporal-maintained Core,
+long-running compatibility coverage, and years of production history. A pure
+Bun worker can be production quality, but it must publish enough proof that:
+
+- Workflow code cannot silently observe nondeterministic Bun/JS runtime state.
+- Replay reconstruction matches real Temporal histories across feature and
+  version combinations.
+- Worker pollers, sticky queues, heartbeats, updates, shutdown, and retries hold
+  under load, restart, and failure conditions.
+- Protocol command materialization remains compatible with Temporal Server
+  changes.
+- Release artifacts make the package boundary, test results, and support matrix
+  easy for agents and humans to verify.
+
+The current library is already deployed and tested on the current infra. That
+counts, but agents outside this project cannot infer it. The implementation
+work is to convert private confidence into public, machine-checkable evidence.
+
+## Code Read Findings
+
+| Surface              | Current implementation                                                                                                                                                                                                                                           | Production gap to close                                                                                                                                                                                                                         |
+| -------------------- | ---------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- | ----------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- |
+| Package boundary     | `packages/temporal-bun-sdk/package.json` ships `dist`, `docs`, `skills`, and `README.md`; runtime dependencies are `@bufbuild/protobuf`, Connect, Effect, and TypeScript. `verify:production` runs `tests/packaging/manifest-packaging.test.ts`.                 | Keep this as a release gate and publish the result in a release manifest so agents can prove the package is not a Node/native wrapper.                                                                                                          |
+| Worker runtime       | `src/worker/runtime.ts` owns config load, WorkflowService transport, workflow/activity pollers, sticky queues, deployment/build IDs, scheduler, metrics, plugins, graceful shutdown, determinism marker emission, and activity task lifecycle.                   | Add restart/chaos/soak scenarios for poll cancellation, sticky cache drift, task-not-found, heartbeat failure, tuner changes, and shutdown during active workflow/activity tasks.                                                               |
+| Workflow execution   | `src/workflow/executor.ts` runs registered workflows through Effect, creates `WorkflowCommandContext`, evaluates queries, processes updates, and materializes success/failure commands.                                                                          | Add protocol golden tests that compare emitted commands and update protocol messages against captured histories and expected server-visible events.                                                                                             |
+| Determinism guard    | `src/workflow/determinism.ts` records command, random, time, signal, query, and update streams and throws `WorkflowNondeterminismError` on replay mismatch.                                                                                                      | Add async interleaving fuzz tests and query-mode negative tests. Query handlers must never read live time/randomness as a hidden side channel.                                                                                                  |
+| Runtime guards       | `src/workflow/guards.ts` patches `Date`, `Date.now`, `Math.random`, `crypto.randomUUID`, `crypto.getRandomValues`, `fetch`, timers, `performance.now`, `WebSocket`, `Bun.spawn`, and `Bun.nanoseconds`; `WorkflowExecutor` requires strict guards in production. | Release gate includes runtime guard tests, query guard matrix, workflow lint tests, async fuzz artifacts, and semantic-readiness evidence for Bun async/runtime behavior.                                                                       |
+| Static workflow lint | `src/bin/lint-workflows-command.ts` walks workflow import graphs and denies unsafe imports/globals/member expressions. Tests cover `fetch`, `process.env`, captured `Date.now`, and importing client APIs from workflows.                                        | Add rules for dangerous Promise/Effect escape hatches, dynamic eval/function creation, timers captured through aliases, and Bun runtime APIs not currently listed. Make release CI fail if configured workflow entries are missing.             |
+| Replay               | `src/workflow/replay.ts` ingests real histories, applies full/delta determinism markers, reconstructs command history, tracks updates, and diffs mismatch metadata. Stored fixtures currently cover timer, activity retry, and child/continue-as-new histories.  | Scale from a small fixture set to a versioned corpus that covers every supported command/event pair, updates, signals, queries, cancellations, payload codecs, search attributes, memo, markers, failures, sticky replay, and old SDK versions. |
+| Integration          | `tests/integration/**` covers history replay, activity lifecycle, query-only workflows, signal/query, workflow updates, payload codecs, client resilience, worker ops, schedules, and worker runtime behavior behind `TEMPORAL_INTEGRATION_TESTS=1`.             | Split optional service-unavailable skips from release-blocking skips. In release CI, a missing dev server or unimplemented critical endpoint must fail instead of silently reducing coverage.                                                   |
+| Load                 | `tests/integration/load/**` submits CPU, activity, and update workflows, checks throughput, sticky hit ratio, and poll p95 latency, and writes JSONL/report artifacts.                                                                                           | Add duration mode, memory/heap samples, worker restart mode, Temporal endpoint interruption, sticky cache churn, and nightly/weekly soak thresholds.                                                                                            |
+| CI                   | `.github/workflows/temporal-bun-sdk.yml` builds, lints, tests, runs replay-corpus, async-fuzz, load, soak smoke, semantic production verification, and uploads release artifacts.                                                                                | Add longer restart/chaos soak lanes before broadening support to new platforms, Temporal Server minors, or higher-throughput profiles.                                                                                                          |
+
+## Production Bar
+
+The library becomes the obvious default when every release can point to a
+single production-readiness artifact containing:
+
+- Package boundary proof: no official Node worker dependency, no native bridge,
+  no `node-gyp`, no `dist/native`, no stale native Docker path.
+- Determinism proof: replay corpus result, async fuzz result, query/global guard
+  matrix, and mismatch diagnostics samples.
+- Protocol proof: command/update/query/signal/activity golden tests by Temporal
+  Server version and SDK version.
+- Operations proof: load and soak reports with throughput, latency, sticky cache
+  ratio, heartbeat retries/failures, worker failures, memory slope, and restart
+  outcomes.
+- Compatibility proof: supported Bun, Temporal Server, Temporal Cloud, TLS/mTLS,
+  payload codec, and deployment/versioning matrix.
+- Support proof: documented feature status, known limits, upgrade policy,
+  security/reporting contact, and release deprecation policy.
+
+## Implementation Phases
+
+### P0 - Credibility Gate And Public Evidence
+
+Purpose: make the current hardening visible and stop regressions.
+
+Implementation:
+
+- Add `scripts/production-readiness/collect-release-evidence.ts`.
+- Generate `dist/production-readiness.json` during `prepack`.
+- Include package boundary, Bun/Temporal versions, git SHA, command results,
+  replay fixture count, integration skip count, load report path, and docs hash.
+- Extend `verify:production` to validate the generated evidence schema.
+- Remove or rename stale `TODO(TBS-NDG-*)` comments once the matching tests prove
+  they are complete; leave new TODOs only for real unfinished work.
+- Add a release CI step that uploads `production-readiness.json` with npm pack
+  output and worker-load artifacts.
+
+Acceptance:
+
+- `bun run --filter @proompteng/temporal-bun-sdk verify:production` fails if the
+  evidence file is missing, malformed, or reports native/Node worker artifacts.
+- `npm pack --dry-run --json` output can be matched to the same evidence file.
+- Docs link to the evidence fields so agents can make a yes/no decision without
+  reading the whole repo.
+
+### P1 - Determinism And Async Semantics Audit
+
+Purpose: answer the specific "Bun async semantics" concern.
+
+Implementation:
+
+- Add `tests/workflow/query-guard-matrix.test.ts`.
+  - Assert `Date`, `Date.now`, `Math.random`, `crypto.randomUUID`,
+    `crypto.getRandomValues`, `performance.now`, timers, fetch, WebSocket, and
+    Bun process APIs either throw `WorkflowQueryViolationError` in query mode or
+    replay from prior determinism state. No query path may read live wall-clock
+    or random state.
+- Add `tests/workflow/async-determinism-fuzz.test.ts`.
+  - Generate seeded workflows that interleave Effect yields, Promise microtasks,
+    timers via workflow APIs, activities, signals, queries, updates, sideEffect,
+    getVersion, patch, and local activities.
+  - Run each seed twice: initial execution records state; replay must produce
+    identical output and zero `diffDeterminismState` mismatches.
+  - Mutate one command/random/time/query/update slot per seed and assert the
+    mismatch is detected with useful metadata.
+- Extend `lint-workflows` rules to flag hidden async escape hatches:
+  - raw `Promise` constructors in workflow modules unless allowlisted;
+  - `Effect.tryPromise` and `Effect.promise` in workflow handlers unless called
+    through SDK-provided deterministic adapters;
+  - captured unsafe globals through aliases;
+  - `eval`, `Function`, dynamic import, and Bun runtime APIs beyond the current
+    deny list.
+- Add a CLI gate that runs `temporal-bun lint-workflows` in strict JSON mode and
+  writes `.artifacts/workflow-lint/report.json`.
+
+Acceptance:
+
+- Runtime guard, query guard matrix, and async fuzz suites pass locally and in
+  CI.
+- Fuzz default: at least 1,000 seeds in PR CI under a stable seed file.
+- Fuzz release/nightly: at least 10,000 seeds with artifacted seed failures.
+- Any workflow query reading live time/randomness is a release blocker.
+
+### P2 - Replay Corpus Expansion
+
+Purpose: prove replay against real Temporal histories, not only synthetic unit
+fixtures.
+
+Implementation:
+
+- Add `tests/replay/corpus/manifest.json`.
+  - Fields: fixture name, Temporal Server version, SDK version, Bun version,
+    workflow type, feature tags, history event count, expected command count,
+    payload codec profile, captured date, and source command.
+- Add `scripts/replay/capture-corpus-fixture.ts`.
+  - Starts or references a workflow, fetches history through WorkflowService or
+    Temporal CLI, normalizes JSON, runs `ingestWorkflowHistory`, stores expected
+    determinism state, and updates the manifest.
+- Add `scripts/replay/verify-corpus.ts`.
+  - Runs all fixtures, rejects empty corpus, rejects duplicate tags without a
+    replacement marker, rejects unsupported schema versions, and writes
+    `.artifacts/replay-corpus/report.json`.
+- Keep the existing small `tests/replay/fixtures.test.ts`, but make the larger
+  corpus the release gate.
+
+Coverage targets:
+
+- Activities: success, retry success, retry exhaustion, cancellation, heartbeat
+  timeout, non-retryable failure, payload codec failure details.
+- Timers: fire, cancel, replay after marker, long timeout normalization.
+- Child workflows: start, complete, fail, cancel, terminate, timeout, pending
+  child replay.
+- Continue-as-new and versioning: marker chain, getVersion, patch/deprecate,
+  old-build replay.
+- Signals and queries: signal delivery order, query-only task, legacy query,
+  multi-query, query failure.
+- Updates: admitted, accepted, rejected, completed success/failure, duplicate
+  update ID, cancellation while awaiting.
+- Search attributes and memo: upsert and workflow properties modified.
+- Nexus operations: schedule, complete, fail, cancel, timeout.
+- Payload codecs: JSON, binary tunnel, gzip, AES-GCM, key rotation metadata.
+
+Acceptance:
+
+- PR gate: at least 25 high-signal corpus fixtures.
+- Release gate: at least 75 fixtures covering every GA-critical command/event
+  family.
+- Default-choice gate: at least 150 fixtures across at least two Temporal Server
+  minor versions and two SDK minor versions.
+
+### P3 - Protocol Golden And Compatibility Tests
+
+Purpose: prove the custom Bun worker emits server-compatible commands.
+
+Implementation:
+
+- Add `tests/protocol/command-golden.test.ts`.
+  - Given workflow command intents, materialize commands and compare stable JSON
+    projections of `Command` protos.
+  - Include headers, memo, search attributes, retry policies, parent close
+    policy, task queues, timeouts, cancellation, continue-as-new, and Nexus.
+- Add `tests/protocol/update-protocol-golden.test.ts`.
+  - Validate `buildUpdateProtocolMessages` output for acceptance, rejection,
+    success completion, failure completion, and replay duplicate suppression.
+- Add `tests/protocol/history-roundtrip.test.ts`.
+  - For selected corpus fixtures, reconstruct determinism state, replay the
+    workflow, materialize commands, and assert stable equivalence against the
+    next command-bearing history events.
+- Add a Temporal compatibility matrix job:
+  - local dev server for PRs;
+  - pinned Temporal Server minors in nightly;
+  - Temporal Cloud smoke in release when credentials are present.
+
+Acceptance:
+
+- Every command kind in `src/workflow/commands.ts` has a golden projection.
+- Every event kind handled in `src/workflow/replay.ts` has at least one corpus
+  fixture or an explicit unsupported-status entry.
+- Protocol golden tests fail on accidental protobuf field drift.
+
+### P4 - Worker Lifecycle, Chaos, And Soak
+
+Purpose: prove the runtime holds under real operating conditions.
+
+Implementation:
+
+- Extend `tests/integration/load/config.ts` with:
+  - `TEMPORAL_LOAD_TEST_DURATION_MS`;
+  - `TEMPORAL_LOAD_TEST_RESTART_INTERVAL_MS`;
+  - `TEMPORAL_LOAD_TEST_ENDPOINT_BLACKHOLE_MS`;
+  - `TEMPORAL_LOAD_TEST_STICKY_CHURN_RATIO`;
+  - `TEMPORAL_LOAD_TEST_MEMORY_SLOPE_MAX_MB_PER_HOUR`;
+  - `TEMPORAL_LOAD_TEST_SHUTDOWN_DURING_ACTIVITY_RATIO`.
+- Extend `tests/integration/load/runner.ts` to:
+  - sample RSS/heap and write `memory.jsonl`;
+  - periodically shutdown and recreate `WorkerRuntime`;
+  - interrupt pollers and confirm `#withRpcAbort` and scheduler shutdown do not
+    strand tasks;
+  - force sticky cache eviction and drift rebuilds;
+  - track task-not-found, nondeterminism, heartbeat retry/failure, activity
+    failure, workflow failure, and sticky heal metrics.
+- Add `scripts/run-worker-soak.ts`.
+  - Duration-mode wrapper around the load runner with stricter artifact
+    validation.
+- Add `.github/workflows/temporal-bun-sdk-nightly.yml`.
+  - Runs 2-hour soak nightly.
+  - Runs 6-hour soak on release branch.
+  - Allows longer manual soak before broadening platform or throughput support.
+
+Acceptance:
+
+- PR load smoke: current short load test remains under 10 minutes.
+- Nightly soak: 2 hours, zero stuck workflows, no unhandled runtime rejection,
+  memory slope below threshold, sticky heal rate below threshold, and all
+  metrics artifacts uploaded.
+- Release soak: 6 hours against pinned Temporal Server with restart mode
+  enabled.
+- Default-choice release smoke: CI soak artifact present and validated by
+  `verify:production`; extended soak is required for unusual workload or
+  platform risk.
+
+### P5 - CI Skip Policy And Release Blocking Rules
+
+Purpose: prevent false confidence from skipped integration tests.
+
+Implementation:
+
+- Add a test reporter that counts Bun skipped tests and scenario-level "skipped"
+  console warnings for the SDK package.
+- Add `TEMPORAL_BUN_RELEASE_GATE=1`.
+  - In release mode, missing Temporal CLI/dev server, missing corpus fixtures,
+    and critical integration skips are failures.
+  - Optional external services remain optional only if the skip is tagged with a
+    non-critical capability, for example Temporal Cloud Ops without credentials.
+- Split integration suites into:
+  - `test:integration:critical`;
+  - `test:integration:optional`;
+  - `test:load`;
+  - `test:soak`.
+- Publish `.artifacts/temporal-bun-sdk/test-summary.json`.
+
+Acceptance:
+
+- Release CI fails if a critical integration test silently skips.
+- The release manifest includes skip counts by suite and reason.
+- PR CI remains usable without requiring Temporal Cloud credentials.
+
+### P6 - Docs, Support Policy, And Agent Default Metadata
+
+Purpose: make choosing the SDK easy and defensible.
+
+Implementation:
+
+- Add `docs/support-policy.md`.
+  - Supported Bun versions, Temporal Server versions, Temporal Cloud support,
+    OS/arch, payload codecs, known limits, security reporting, and deprecation
+    policy.
+- Add `docs/feature-matrix.md`.
+  - Feature status with test/corpus references and unsupported/experimental
+    labels.
+- Add `docs/agent-adoption-guide.md`.
+  - Short guidance for agents: when to choose `@proompteng/temporal-bun-sdk`,
+    required gates, project scaffolding command, Docker command, and fallback
+    criteria.
+- Add `dist/agent-readiness.json`.
+  - `recommended: true` only when the production-readiness artifact meets all
+    default-choice thresholds.
+  - Include machine-readable gate names, versions, artifact paths, and known
+    limits.
+- Update `apps/docs/content/docs/temporal-bun-sdk.mdx` and
+  `apps/docs/content/docs/temporal-bun-sdk-comparison.mdx` to link the evidence
+  instead of asking users to trust prose.
+
+Acceptance:
+
+- An agent can answer "is this the default Bun Temporal SDK?" by reading
+  `agent-readiness.json` plus the release manifest.
+- Public docs explain the actual concern: pure Bun worker correctness proof and
+  support policy, not Node NAPI removal.
+- The comparison page no longer leaves "production ready" as an ambiguous
+  phrase.
+
+## Gate Matrix
+
+| Gate                 | Command                                                           | PR                        | Release           | Default-choice                |
+| -------------------- | ----------------------------------------------------------------- | ------------------------- | ----------------- | ----------------------------- |
+| Package boundary     | `bun run --filter @proompteng/temporal-bun-sdk verify:production` | required                  | required          | required                      |
+| Build                | `bun run --filter @proompteng/temporal-bun-sdk build`             | required                  | required          | required                      |
+| Unit/runtime guards  | `bun test tests/workflow/runtime-guards.test.ts`                  | required                  | required          | required                      |
+| Query guard matrix   | `bun test tests/workflow/query-guard-matrix.test.ts`              | required                  | required          | required                      |
+| Async fuzz           | `bun test tests/workflow/async-determinism-fuzz.test.ts`          | 1k seeds                  | 10k seeds         | 10k+ seeds, last 7 days green |
+| Replay corpus        | `bun scripts/replay/verify-corpus.ts`                             | 25 fixtures               | 75 fixtures       | 150 fixtures                  |
+| Protocol golden      | `bun test tests/protocol/*.test.ts`                               | required                  | required          | required                      |
+| Critical integration | `TEMPORAL_INTEGRATION_TESTS=1 bun test tests/integration`         | required                  | no critical skips | no critical skips             |
+| Load smoke           | `bun run --filter @proompteng/temporal-bun-sdk test:load`         | required                  | required          | required                      |
+| Soak                 | `bun scripts/run-worker-soak.ts`                                  | optional                  | 6h                | 24h                           |
+| Docs                 | `bun run --filter docs build`                                     | required when docs change | required          | required                      |
+| Pack                 | `npm pack --dry-run --json`                                       | optional                  | required          | required                      |
+
+## First Implementation Order
+
+1. **P0 evidence manifest.** This is the fastest way to make the already-shipped
+   production boundary and current CI/load results visible to agents.
+2. **P1 query/async guard tests.** This directly answers the ChatGPT concern
+   about Bun async semantics and hidden nondeterminism.
+3. **P2 replay corpus runner.** This converts the existing replay engine into
+   durable proof across real histories.
+4. **P3 protocol golden tests.** This protects the pure-Bun command
+   materializer from Temporal protobuf/server drift.
+5. **P4 soak mode.** This moves the load harness from a smoke test to
+   operational proof.
+6. **P6 agent metadata.** Only mark the SDK as default when the evidence gates
+   are green, not because the docs claim it.
+
+## Non-Goals
+
+- Running `@temporalio/worker` on Bun.
+- Depending on Node NAPI, `@temporalio/core-bridge`, `node-gyp`, or a native
+  worker bundle.
+- Byte-for-byte equivalence with the official SDK internals.
+- Hiding unsupported features. Missing compatibility should be listed in the
+  feature matrix and reflected in `agent-readiness.json`.
+
+## Definition Of Done
+
+The SDK is production-default for agents when:
+
+- `agent-readiness.json` says `recommended: true`.
+- The latest release manifest proves all default-choice gates are green.
+- The replay corpus covers every GA-critical workflow/event family.
+- Async fuzz and query guard tests have no open determinism escapes.
+- Load and soak artifacts are linked from release artifacts and validated by
+  `verify:production`.
+- Docs and comparison pages explain that this is a pure Bun SDK with public
+  evidence, not an unofficial wrapper around the Node worker.

package/docs/release-runbook.md

@@ -0,0 +1,48 @@
+# Temporal Bun SDK Release Runbook
+
+Automation now drives every step via `.github/workflows/temporal-bun-sdk.yml`. Use
+this runbook as a quick reference for triggering and validating releases.
+
+> **Status:** TBS-009 (Release Automation) closed on 2025-11-17 via the trusted
+> publishing workflow; this document captures the final process.
+
+## Prerequisites
+
+- npm trust publishing is enabled for `proompteng/lab → temporal-bun-sdk.yml`
+  (Trusted Publisher entry in the npm org). No automation token is required.
+- Maintainer access to run the Temporal Bun SDK workflow manually on `main`.
+- Desired npm dist-tag (`latest`, `beta`, etc.) for the publish run.
+- Buf CLI installed (`buf` on `PATH`) if you plan to run the proto regeneration
+  script locally.
+
+## Release Steps
+
+1. **Prepare mode**
+   - Trigger the workflow with `release_mode=prepare` (optionally set
+     `npm_tag`). This runs release-please to open/update the automated release
+     PR (`release-please--branches--main--components--temporal-bun-sdk`).
+   - The job runs Oxfmt + build + unit + load suites against the release PR
+     branch so we can verify artifacts before merging. Review the PR and merge
+     after CI is green.
+2. **Publish mode**
+   - Re-run the workflow with `release_mode=publish` (set `dry_run=true` for a
+     rehearsal) **from the `main` branch only**; the workflow refuses to run on
+     other refs. The job reads the merged `package.json` version on `main`, reruns
+     the validations, then executes `npm publish --provenance --access public
+--tag <dist>`.
+   - Keep the workflow logs + artifacts linked to the tracking issue/PR as proof
+     of the dry-run and the final publish.
+
+## Support & Incident Response
+
+- Direct any security or incident reports to `security@proompteng.ai`.
+
+## Post-Release Checklist
+
+- Confirm the npm release metadata (version, dist-tag, provenance) matches the
+  workflow output.
+- Close the tracking issue (e.g., #1788) and note the publish link.
+- Schedule any follow-up docs/DevRel announcements if required.
+- If upstream Temporal protos changed, trigger the "Temporal Bun SDK Proto
+  Regen" workflow (or confirm the scheduled run already merged) so generated
+  sources stay current.

package/docs/semantic-readiness.md

@@ -0,0 +1,79 @@
+# Temporal Bun SDK Semantic Readiness
+
+_Last updated: May 5, 2026_
+
+## Decision
+
+`@proompteng/temporal-bun-sdk` is the default Temporal choice for Bun-first
+worker/client projects when `dist/agent-readiness.json` reports
+`recommended: true`.
+
+That recommendation is intentionally scoped. It means the project wants Bun as
+the worker runtime and accepts the ProomptEng-maintained SDK support contract.
+Use Temporal's official TypeScript SDK when the requirement is official
+Temporal-maintained Core support on Node.js.
+
+## Original Concern
+
+The production concern was not whether this package can avoid Node-API. It
+already does: the worker runtime is pure Bun/TypeScript and the package gate
+rejects `@temporalio/worker`, `@temporalio/core-bridge`, native artifacts,
+`node-gyp`, `process.dlopen()` paths, and stale native Docker paths.
+
+The real concern was Temporal worker semantics:
+
+- workflows must not observe nondeterministic Bun/JSC async state;
+- replay must match real Temporal histories;
+- command materialization must remain compatible with Temporal Server;
+- heartbeats, cancellation, retries, sticky queues, shutdown, and pollers must
+  hold under load;
+- agents need a public artifact that proves those checks happened for a
+  release.
+
+## Machine-Readable Evidence
+
+Every release publishes two files in `dist/`:
+
+- `production-readiness.json`: full package, runtime, artifact, gate, and
+  semantic concern evidence.
+- `agent-readiness.json`: compact recommendation status for agents.
+
+Agents should inspect `agent-readiness.json` first. A release is a default
+choice only when:
+
+- `recommended` is `true`;
+- `blockers` is empty;
+- every `semanticConcerns[]` entry with `defaultChoiceRequired: true` has
+  `passed: true`.
+
+## Semantic Concern Matrix
+
+| Concern                   | Required evidence                                                                                                      |
+| ------------------------- | ---------------------------------------------------------------------------------------------------------------------- |
+| Pure Bun worker boundary  | Package files, Dockerfile, package-boundary test, no forbidden Node/native deps.                                       |
+| Deterministic replay      | Replay fixtures, replay corpus report, replay engine tests, async fuzz artifact.                                       |
+| Bun async semantics       | Runtime guards, query guard matrix, workflow lint, 10k-seed async fuzz artifact.                                       |
+| Temporal command protocol | Command materialization source, protocol golden tests, replay corpus verifier.                                         |
+| Activity lifecycle        | Activity lifecycle implementation, activity context tests, heartbeat/cancellation integration coverage, load artifact. |
+| Sticky cache and shutdown | Worker runtime, sticky-cache tests, worker runtime integration coverage, load and soak artifacts.                      |
+| Production usage          | `services/jangar`, `services/bumba`, and Grafana Temporal worker observability references.                             |
+| Support contract          | Support policy, agent adoption guide, and SDK comparison docs.                                                         |
+
+## Current Release Gate
+
+The CI and publish workflow for the SDK must run:
+
+```bash
+bun run --filter @proompteng/temporal-bun-sdk build
+bunx oxfmt --check packages/temporal-bun-sdk/src packages/temporal-bun-sdk/tests packages/temporal-bun-sdk/scripts packages/temporal-bun-sdk/docs
+bun run --cwd packages/temporal-bun-sdk lint:oxlint
+cd packages/temporal-bun-sdk && TEMPORAL_TEST_SERVER=1 bun test --timeout=30000 --max-concurrency=1
+bun run --filter @proompteng/temporal-bun-sdk verify:replay-corpus
+TEMPORAL_TEST_SERVER=1 bun run --filter @proompteng/temporal-bun-sdk test:load
+TEMPORAL_TEST_SERVER=1 bun run --filter @proompteng/temporal-bun-sdk test:soak -- --duration 1000 --workflows 16 --workflow-concurrency 4 --activity-concurrency 6
+bun run --filter @proompteng/temporal-bun-sdk verify:production
+```
+
+`verify:production` validates replay-corpus status, load thresholds, async fuzz
+seed count, soak status, CI workflow coverage, and the semantic concern matrix.
+It fails the release if any required concern is not evidenced.

package/docs/support-policy.md

@@ -0,0 +1,79 @@
+# Temporal Bun SDK Support Policy
+
+_Last updated: May 5, 2026_
+
+## Supported Runtime Matrix
+
+| Surface         | Supported                                                                       | Notes                                                                                                                                   |
+| --------------- | ------------------------------------------------------------------------------- | --------------------------------------------------------------------------------------------------------------------------------------- |
+| Bun             | `>=1.3.13`                                                                      | CI pins Bun `1.3.13`; newer Bun releases must pass the SDK test, replay, and load gates before being documented as preferred.           |
+| Node            | Build tooling only                                                              | The worker runtime is Bun TypeScript. Node is used by repository tooling and GitHub Actions setup, not by the published worker runtime. |
+| Temporal Server | Current CI cluster plus pinned dev server histories                             | Additional server minor/version coverage requires replay-corpus and integration evidence before being documented as preferred.          |
+| Temporal Cloud  | Client/worker TLS and mTLS paths are supported when credentials are configured. | Cloud Ops tests remain optional unless credentials are present.                                                                         |
+| OS/arch         | Linux arm64 in CI, local macOS development                                      | Additional platforms require package-boundary, build, test, replay, and load evidence.                                                  |
+
+## Supported Package Boundary
+
+The published package is expected to contain only:
+
+- `dist`;
+- `docs`;
+- `skills`;
+- `README.md`.
+
+The production gate rejects:
+
+- `@temporalio/worker`;
+- `@temporalio/core-bridge`;
+- `@temporalio/client`;
+- `node-gyp`, `node-addon-api`, `node-gyp-build`;
+- native artifacts such as `.node`, `.dylib`, `.so`, and `.a`;
+- stale native bridge paths such as `bruke`, `native`, and `dist/native`.
+
+## Required Release Evidence
+
+Every release should publish or upload:
+
+- `dist/production-readiness.json`;
+- `dist/agent-readiness.json`;
+- replay corpus report;
+- worker load report;
+- async determinism fuzz report;
+- worker soak report;
+- npm pack or publish provenance output.
+
+`agent-readiness.json` must not set `recommended: true` until the replay corpus,
+async determinism fuzz, load, soak, CI coverage, package-boundary, and semantic
+concern evidence meet the default-choice thresholds in
+`docs/semantic-readiness.md`.
+
+## Known Limits
+
+- This is a pure Bun SDK, not Temporal's official TypeScript SDK and not a
+  wrapper around the official Node worker.
+- Official SDK internal sandbox behavior is not a compatibility promise. The
+  compatibility promise is Temporal protocol behavior plus deterministic replay
+  evidence.
+- Default-choice status is scoped to Bun-first deployments that accept the
+  published gates and support model. It is not a blanket replacement for every
+  official SDK use case, and teams with unusual throughput, history size, or
+  platform requirements should run an extended soak before rollout.
+
+## Security And Reliability Issues
+
+Report security or reliability issues through the repository's normal private
+security channel. For determinism issues, include:
+
+- SDK version and Bun version;
+- Temporal Server or Cloud version if known;
+- workflow type and task queue;
+- history JSON or `temporal-bun replay --json` output;
+- `production-readiness.json` from the release in use.
+
+## Deprecation Policy
+
+Breaking workflow behavior must be guarded by deterministic versioning APIs such
+as `determinism.getVersion` or `determinism.patched`. Removing support for a Bun
+or Temporal Server version requires a release note, feature-matrix update, and a
+replay-corpus fixture proving old histories still replay or an explicit
+unsupported-status entry.

package/docs/temporal-ci-cluster-requirement.md

@@ -0,0 +1,36 @@
+# Temporal CI Cluster Requirement
+
+## Summary
+
+Temporal integration tests in `@proompteng/temporal-bun-sdk` must run against the shared Temporal cluster managed by ArgoCD.
+
+They must **not** be switched to a local `start-dev` Temporal server in CI.
+
+## Why
+
+- The integration suite assumes shared-cluster behavior and timing characteristics.
+- Switching CI to local `start-dev` introduced regressions:
+  - hook and scenario timeouts,
+  - connection-refused races,
+  - unstable harness startup/teardown behavior under parallel test execution.
+- This produced false negatives that were unrelated to the production task-queue-kind fix.
+
+## Required CI behavior
+
+- Keep CI pointed at the ArgoCD Temporal endpoint via the short Tailscale hostname
+  (`TEMPORAL_ADDRESS=temporal-grpc:7233` in the current workflow).
+- ARC runners must preserve the tailnet search suffix so `temporal-grpc` resolves to the shared
+  cluster endpoint on every job runner without hardcoding the full `*.ts.net` hostname in workflows.
+- Keep `TEMPORAL_TEST_SERVER=1` in CI for SDK test jobs.
+- Keep `TEMPORAL_ENFORCE_REMOTE_ADDRESS=1` in CI so localhost targets fail fast.
+- If readiness is slow, improve readiness retries/diagnostics, but do not redirect CI to local Temporal.
+
+## Guardrail for future changes
+
+When touching Temporal CI or test harness code:
+
+1. Do not replace the cluster target with `127.0.0.1`/local dev server in CI.
+2. Validate failures first as cluster readiness/connectivity before changing execution mode.
+3. Prefer stronger readiness waiting, transient CLI retries, and DNS diagnostics over topology changes
+   or swapping back to a fully-qualified fallback.
+4. Keep the workflow checks that enforce a remote endpoint and verify `temporal-grpc` resolves on ARC.