refactor(runtime): delete legacy SQLite state adoption

[ScriptedAlchemy]

Summary

Delete the two pre-1.0 compatibility paths createSqliteStateDriver gained in #201. The owner has approved breaking changes for old on-disk stores, so neither path gets a fallback reader or a migration shim.

• The root-mode opener no longer looks for <sanitized id>-<first 12 hex of utf8(id)>.sqlite or renames it and its -wal/-shm sidecars to <sanitized id>-<sha256(id)[0:16]>.sqlite. An old file stays on disk untouched and a fresh store opens beside it.
• agent_state_journal.result_state is now NOT NULL in the CREATE TABLE. The PRAGMA table_info + ALTER TABLE ... ADD COLUMN upgrade, the row.result_state ?? row.state read fallback, the replay-to-revision fallback in #committedState, and the "legacy journal row ... has no recoverable committed result" migration-failure branch are gone.
• One generic schema check replaces the ALTER: after CREATE TABLE IF NOT EXISTS, initialize compares each table's PRAGMA table_info column list with the current schema and fails closed with AgentStateError('corrupt', ...). Without it an old journal opened cleanly and the first dispatch surfaced a raw SQLite "no column named result_state" as unavailable. The check names no old format.

Kept as-is: kernel formats 1/2, the 'compact' row stored as 'reset', the pending-migration loop that migrates result_state between schema versions, and the notice ledger migrations under src/notices/** (another PR).

Removal matrix

Removed	Callers migrated	Deleted surface
legacySanitizedFileName	none remained after the adopter was deleted	private helper in src/state/sqlite.ts
Rename-on-open adopter (legacy file + -wal/-shm sidecars)	open() now returns resolve(join(root, sanitizedFileName(id)))	existsSync/renameSync imports, two ENOENT/race branches
PRAGMA table_info + ALTER TABLE agent_state_journal ADD COLUMN result_state	CREATE TABLE declares result_state TEXT NOT NULL; a generic column-list check fails closed	in-place journal upgrade
row.result_state ?? row.state fallback	JournalRow.result_state: string; #committedByKey reads it directly	replay-to-revision fallback in #committedState, its db parameter
"legacy journal row ... has no recoverable committed result" recovery	migration loop reads row.result_state unconditionally	migration-failure branch and head-row recovery branch
Test adopts the legacy root database and live WAL sidecars	none	legacyFileName, currentFileName, holdUncheckpointedLegacyEvent
Tests fails closed when a non-head legacy event..., migrates a legacy journal-head result..., preserves legacy reset input...	replaced by keeps the original commit input while migrating each committed result (current writer, v1 → v2 open)	createLegacyMigrationDatabase, clearLegacyEventResults, createLegacyHeadOnlyDatabase, valueCounterDefinition

Added test: fails closed with a typed corrupt error when a table does not match the current schema (drops result_state from a current store and asserts the corrupt error names table agent_state_journal).

Consequence for old stores

Measured on the built dist with a hand-seeded pre-#201 database:

• Old filename under root: open() creates probe-counter-271e964079f6c266.sqlite and reads { revision: 0, state: initial }; probe-counter-70726f62652f.sqlite keeps its mtime and is never opened.
• Journal without result_state (any path): open() rejects with AgentStateError code corrupt, message State 'probe/counter' table agent_state_journal at '<file>' does not match the current schema (has revision, kind, name, payload, state, to_version, idempotency_key, committed_at; expected revision, kind, name, payload, state, result_state, to_version, idempotency_key, committed_at).

Delete or recreate such stores before upgrading. No docs page under docs/, packages/rsc-runtime/README.md, or website/docs/{en,zh} described either legacy behavior (grepped for the filename scheme, result_state, and adoption prose), so the changeset is the only prose change.

Verification

Run in this worktree on top of origin/main (64492dd):

• pnpm build: pass.
• pnpm typecheck: pass.
• pnpm lint: pass (1518 files, 88 rules).
• pnpm test:unit: 4486 passed, 0 failed, 6 skipped. packages/rsc-runtime/tests/state-sqlite.test.ts alone: 55 passed.
• pnpm test:integration:run packages/rsc-runtime/tests/notices-sqlite-cross-process.test.ts packages/rsc-runtime/tests/state-packaging.test.ts packages/rsc-runtime/tests/state-sqlite-cross-process.test.ts: 10 passed, 0 failed.
• pnpm test:packed packages/agent-bundle/tests/packed-readonly-state-root.test.ts: 1 passed.
• pnpm docs:site:build not run: no public export or website page changed.

Follow-up

Codex threads addressed in #850. The P2 nullable-column case now fails at open: the schema check compares each column's NOT NULL flag. The P1 docs gap is closed in reference/runtime-environment.mdx (en, zh), which says pre-current stores are not read or migrated and how to recover.

[changeset-bot]

🦋 Changeset detected

Latest commit: 05c5f1e

The changes in this PR will be included in the next version bump.

This PR includes changesets to release 2 packages

Name	Type
@agent-bundle/runtime	Minor
create-agent-bundle	Patch

Not sure what this means? Click here to learn what changesets are.

Click here if you're a maintainer who wants to add another changeset to this PR

[chatgpt-codex-connector]

Codex Review Summary

This comment shows the latest Codex review activity on this pull request.

Review	Status	Commit	Review trigger
📝 Code Review	✅ Completed 2026-09-25T00:38:14.833045Z	2e34d54	PR opened

ℹ️ About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

• Open a pull request for review
• Mark a draft as ready
• Comment "@codex review" or "@codex security review".

Codex reacts with 👀 while any review is running, comments if it has suggestions, and reacts with 👍 once all reviews finish with no findings.

[chatgpt-codex-connector]

💡 Codex Review

Here are some automated review suggestions for this pull request.

Reviewed commit: 2e34d54156

ℹ️ About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

• Open a pull request for review
• Mark a draft as ready
• Comment "@codex review".

If Codex has suggestions, it will comment; otherwise it will react with 👍.

Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".

[pkg-pr-new]

npm i https://pkg.pr.new/ScriptedAlchemy/agent-bundle@837

npm i https://pkg.pr.new/ScriptedAlchemy/agent-bundle/create-agent-bundle@837

npm i https://pkg.pr.new/ScriptedAlchemy/agent-bundle/rsc-markdown-stream@837

npm i https://pkg.pr.new/ScriptedAlchemy/agent-bundle/@agent-bundle/runtime@837

commit: 05c5f1e