@artblocks/abx-cli 0.1.0-alpha.42 → 0.1.0-alpha.44

package/CHANGELOG.md

@@ -1,238 +1,24 @@
 # @artblocks/abx-cli
 
-## 0.1.0-alpha.42
-
-### Minor Changes
-
-- 0fed512: `deploy-code`'s post-deploy setup (script chunks, PostParam schemas, dependency declarations,
-  on-chain-URI legs, reserve mints) no longer rides one atomic multicall no matter how large the
-  on-chain script is. A 60KB script (3 chunks) produced a setup transaction wanting 17,307,586 gas
-  against the ~16,777,216 (2^24) `eth_estimateGas` ceiling — not the block gas limit — and failed with
-  a useless "gas limit too high", leaving a verifiably half-configured contract on chain.
-
-  The setup now splits into gas-bounded transactions when it doesn't fit one: every script-chunk batch,
-  then every config (schema/dependency/on-chain-URI) batch, then every mint batch, in that order and
-  never mixed. Script-chunk writes and config setters are idempotent, so an interruption at any split
-  point is finished by the existing `deploy-code --resume` (which now also gas-batches whatever it
-  finds missing); mints ride strictly last because a mint is the one non-idempotent leg. A setup that
-  already fit one transaction is unaffected — same single combined multicall as before. There is no new
-  size-based refusal; the dry run's `transactions`/`approvals` reports the real transaction count and
-  cost for the whole plan.
-
-  `@artblocks/abx-sdk` exports the reusable pieces: `packCallsByGas` (a generic gas-bounded bin packer,
-  the call-level sibling of `planContentTxs`'s own batching), `estimateChunkGasForBytes`, and
-  `SETUP_LEG_GAS` (flat gas estimates for a schema/dependency/on-chain-URI/mint call).
-
-- 0fed512: Fixed three omissions in the structured deploy `--json` `plan` object (a cold-agent regression sweep
-  of issue #123's plan surfaced all three): `plan.custody` now carries the on-chain renderer address
-  (the same one the "On-chain renderer" step prints) and, on the single-file image lanes
-  (`deploy`/`deploy` `--copies`), file metadata for a staged `--onchain-image` (raw bytes, staged/
-  compressed bytes, sniffed MIME type, and a keccak256 of the local file) — both were visible in the
-  human dry-run/confirm prose but entirely absent from the JSON payload. `plan.transactions.legs` now
-  always ends with `'deploy'` when non-null, naming the deploy transaction itself instead of silently
-  omitting it from its own leg list (the `--resume` lane, which never deploys, is unaffected). The
-  `estimate` field's `null` on lanes that compute no cost figure was investigated and confirmed
-  genuinely absent — not fabricated, no new RPC calls added.
-
-  `DEPLOY_PLAN_SCHEMA_VERSION` is bumped to `2` for this shape change. The human dry-run/confirm output
-  is unchanged, byte for byte; this is additive to the JSON only.
+## 0.1.0-alpha.44
 
 ### Patch Changes
 
-- 0fed512: `--dry-run` used to report `"status": "would-succeed"` from an unconstrained `eth_call` that never
-  checked whether the signer could actually PAY for the send — a funded on-chain sweep hit a
-  `replace-script --dry-run` that said "would succeed" for a send whose real gas need was 49% higher
-  and then failed. The simulation now reuses `pinGas` (the exact gas selection the real send makes),
-  prices it against the signer's current balance, and reports one of three honest states — `would-
-succeed` (state check passed AND affordable), `would-revert` (a proven failure, including
-  "insufficient funds" with the actual numbers), or `unknown` (neither could be established — never
-  collapsed into either of the others). `simulation.estimatedGas`/`estimatedCostWei`/`balanceWei` carry
-  the real numbers; `transaction.gasFloor` remains a proven MINIMUM used to detect an implausible
-  estimate, never a prediction of total cost.
-
-  `abx replace-script`'s own `gasFloor` (restated for its batched multicall) was 49% below what a real
-  send needed — not wrong about what it claimed, but incomplete: it counted only SSTORE2's 200
-  gas/byte code-deposit cost, omitting the CREATE opcode's fixed 32,000 gas per chunk write and the
-  transaction's own intrinsic calldata cost (both unconditional, provable physics, unlike a simulated
-  estimate). It now includes both, computed from the real encoded calldata rather than approximated —
-  still a lower bound, never an invented number.
-
-- 0fed512: `abx lock-field` now checks which scope (token or collection) actually serves a field's value before
-  locking — a token-scope value wins over collection when both are set, so locking the scope that
-  ISN'T serving the value used to report "would succeed" (and would succeed) while freezing an empty
-  or overridden slot, not what a viewer sees. It's refused by default, naming the correct command
-  (`--collection` or `--token <id>`); `--force-field` proceeds anyway with a loud warning instead of a
-  silent no-op, for the legitimate case of deliberately locking an empty slot forever. `--dry-run` and
-  a real send see this identically.
-
-  `abx verify` (no `--remote`) already failed fast when a project wasn't registered on this node, but
-  its message didn't say a resolver-served project could skip local registration entirely — an agent
-  following it literally (`abx add <addr>`) could walk into a full historical log scan and repeated
-  rate limits on a project actually meant to be checked with `abx verify <addr> --remote <name>`. The
-  message now names both options.
-
-- Updated dependencies [0fed512]
-  - @artblocks/abx-sdk@0.1.0-alpha.32
-  - @artblocks/abx-indexer@0.1.0-alpha.33
-  - @artblocks/abx-storage@0.1.0-alpha.32
-  - @artblocks/abx-token-api@0.1.0-alpha.35
-
-## 0.1.0-alpha.41
-
-### Minor Changes
-
-- c068a0f: `abx artifacts <address> [--token <id>]` reads a token's `artifacts` manifest directly, instead of
-  fetching and parsing the whole served tokenURI document. Reports every entry plus every
-  producer-registered effect row — current AND stale — each labeled against the token's active
-  `inputsHash`, so a param change that re-addressed a render is diagnosable rather than silently
-  absent. `--remote <name|url>` reads what a hosted resolver actually reports (the surface that owns a
-  hosted project's real artifact set), mirroring the existing `abx verify` / `abx verify --remote`
-  split. Neither "not registered here" nor "not registered on that remote" fails the command — both
-  return a stable `{surface, registered, available, reason, entries, effects}` JSON shape.
-  `@artblocks/abx-token-api` exports the underlying `tokenArtifacts()` read for direct SDK/server use.
-- c068a0f: `abx attach` now distinguishes the on-chain field write from off-chain SERVING — the write can
-  succeed while nothing is able to serve it. It prints the canonical, directly-fetchable URL for every
-  attached key (`{base}/{chainId}/{address}[/<id>]/data/<key>`) so nobody hand-assembles the route, and
-  probes whether a resolver actually answers for the project right now, warning when none does (with
-  no resolver base baked in at all, or with one baked in that isn't currently reachable). `--collection`
-  now correctly reads `contractURIBase` (it previously always read `tokenURIBase`, even for a
-  collection-scope attach).
-- c068a0f: Local data-directory resolution (`.abx-self-host` / `ABX_DATA_DIR`) is now centralized in one SDK
-  resolver (`resolveDataDir` from `@artblocks/abx-sdk/node`) instead of five independent copies of the
-  same fallback across the CLI, indexer, and storage packages — the setup that could silently split a
-  project's SQLite projection from its managed Arweave key into two different directories.
+- Updated dependencies [3ddba71]
+- Updated dependencies [9065128]
+  - @artblocks/abx-sdk@0.1.0-alpha.34
+  - @artblocks/abx-token-api@0.1.0-alpha.37
+  - @artblocks/abx-storage@0.1.0-alpha.34
+  - @artblocks/abx-indexer@0.1.0-alpha.35
 
-  A handful of READ commands (`status`, `state`, `verify`, `doctor`, `capabilities`, `tokens`,
-  `tokenuri`, `contracturi`, `inspect`, `minter show`) now search upward from the current directory,
-  git-style, for an already-existing `.abx-self-host` before falling back to creating one at cwd —
-  bounded at the home directory, a `.git` repository root, or the filesystem root, and never crossing
-  into an unrelated directory tree. Running one of these from a project subdirectory (e.g. a
-  `contracts/` folder) now finds the same node a run from the project root would, instead of quietly
-  reporting an empty one. `abx status`'s existing `data: <path>` line — and a new one-line notice —
-  name the directory whenever it was found this way. Every WRITE command is unaffected: it always
-  resolves strictly to the current directory, exactly as before, so nothing is ever created somewhere
-  you didn't `cd` into. `ABX_DATA_DIR`, when set, still always wins outright and disables the search
-  entirely.
-
-- c068a0f: Every `deploy` / `deploy-series` / `deploy-code` `--json` emit — dry-run and real send, across every
-  product lane and `deploy-code --resume` — now carries a versioned, structured `plan` object alongside
-  the existing fields: wallet transaction/approval count, expected signer/owner/royalty/minter/payee
-  roles, custody and URI-resolution choices, the mint plan, a cost estimate where one is computed, every
-  warning that run raised, and (code lanes) the per-surface thumbnail/traits/postParams disposition plus
-  dependency-registry info. A field a lane has no answer for is reported as `null`, never silently
-  absent. The human dry-run/confirm output is unchanged, byte for byte.
-- c068a0f: `deploy-code --copies --code-dir <dir>` now works: an EditionCode project can point `code` at a
-  hosted directory build (index.html + assets uploaded to ipfs/arweave/cloud), the same as the
-  existing SeriesCode (721) lane.
-- c068a0f: `deploy-code --copies --image-base <url>` now works: an EditionCode project can bake a
-  deterministic per-id off-chain still (`{base}/{id}.png`), the same field write as the existing
-  SeriesCode (721) lane, mutually exclusive with an on-chain `--image-renderer`. The effect runner's
-  render sweep excludes an id with zero live copies (no mint-time seed drawn yet), mirroring how a
-  burned 721 token is already excluded.
-- c068a0f: `deploy-code --resume <address>` now works against an EditionCode (`--copies`) target, not just
-  SeriesCode. The mint leg diffs per id against that id's own `totalSupply(id)` — pass the same
-  content flags the original deploy used, plus `--mint-count`/`--mint-amount` if it premint any ids.
-- c068a0f: `reconstructProject()` and `reconstructIncremental()` accept `toBlock: 'safe' | 'finalized'` in
-  addition to a literal block number and the existing `'latest'` default. The tag is resolved to a
-  concrete inclusive block number (via the new `resolveBlockTag()`) before any `eth_getLogs` scan
-  starts, and that number — never the tag — is what lands in the persisted `ProjectState.toBlock`. An
-  RPC that doesn't support the requested tag throws the new `BlockTagUnavailableError` instead of
-  silently falling back to `latest`. `SelfHostIndexer#reindex` gained a matching `blockTag` option.
-  Existing callers (the `'latest'`/unset default) are unaffected.
-
-  `abx index --to-block safe|finalized` stops a re-index at a reorg-safe boundary.
-
-- c068a0f: The Solidity renderer scaffold (`abx scaffold renderer` / `abx scaffold solidity`) now ships
-  `script/Preview.s.sol`: a Forge script that runs the actual `IAbxFieldRenderer.render` interface
-  against representative token id / seed / PostParam inputs (all overridable via env vars) and writes
-  the SVG/JSON output to `preview-out/` for human inspection — `forge test` proves the renderer
-  behaves, but never let you look at what it drew. Reverts are surfaced clearly; the script is explicit
-  that its gas output is a local-EVM sanity check, not a production gas measurement.
-- c068a0f: Add `abx replace-script <address> --script <file>` — a safe, first-class way to replace an UNLOCKED
-  code project's on-chain program (`SeriesCode`/`EditionCode`, before `abx lock-script`). It refuses
-  outright on a locked script or a non-code target, diffs the replacement by content against what's
-  on-chain (an index that already matches is never re-sent), handles a smaller replacement by queuing
-  the trailing `removeLastScriptChunk` calls it needs, folds every write and remove into ONE atomic
-  transaction (a revert can never leave a half-applied script), and reads the completed script back to
-  verify it reassembles exactly to the file before reporting success. Supports `--dry-run` and every
-  signing lane.
-
-  The SDK gains the primitives this is built on: `prepareSetScriptChunk`/`prepareRemoveLastScriptChunk`
-  (no `prepare*` wrapper existed for `OnChainScript.sol`'s writers before this — every caller hand-
-  encoded `encodeFunctionData`), plus `planScriptReplace`/`verifyScriptReplace`, which take a small
-  reader interface (mirroring `resume.ts`'s own reader seam) so the diff and post-write verification
-  are testable without a chain.
-
-- c068a0f: SQLite maintenance for existing and long-running self-hosted stores (issue #113). `SqliteStore`
-  gains `autoVacuumMode`/`vacuumStats` (detection), `vacuumConvert` (a one-time, explicit full
-  `VACUUM` for a store that predates incremental auto-vacuum), and `runIncrementalVacuum` (a bounded
-  reclaim pass). `SelfHostIndexer.startVacuumMaintenance` runs that bounded reclaim automatically
-  between chain-watch ticks — never inline with a request. The new `abx vacuum [status|convert|
-incremental]` CLI command exposes both halves to an operator. Also fixes a pragma-ordering bug that
-  left EVERY store — not just pre-existing ones — stuck at `auto_vacuum='none'` despite the earlier
-  incremental auto-vacuum fix (`PRAGMA auto_vacuum` must precede `PRAGMA journal_mode = WAL`, not just
-  the schema's first `CREATE TABLE`).
-- c068a0f: `abx state` now reports every irreversible lock a collection can carry — token/contract URI, script,
-  dependencies, param hooks, and the standard `METADATA_FIELD` set — for both ERC-721 and ERC-1155
-  families, not just `paramHooks.locked`. Each lock is a distinct `true` (frozen) / `false` (open) /
-  `null` (unread — never collapsed into `false`) reading, in both human and `--json` output, alongside
-  the exact owner command that freezes it.
-- c068a0f: `abx verify --json` gains a sibling `availability` field (render/serve-readiness: `available` /
-  `partial` / `unavailable` / `unknown`), kept separate from the integrity-only `ok`/`contentIntegrity`
-  that the command's exit status still reflects exactly as before. A locator-committed `image` field
-  (ipfs/arweave/url/url-template) this command doesn't re-fetch now reports as a `verified: null`
-  content check instead of silently reading as no commitment at all. Fixes a side-bug: `abx verify
---remote <r> --json` used to route around `--json` entirely and emit no payload — it now emits the
-  same shape as the local lane, plus a `"not-checked"` `contentIntegrity` state for when byte
-  integrity genuinely wasn't asked (no credential, rejected token, or an older resolver).
-- c068a0f: Additive for SDK consumers — a new exported function, nothing existing changed shape or behavior.
-
-  `abx verify` now compares a collection's actual `tokenURIRenderer` and `animation_url` generator
-  pointers against the current canonical singletons for its chain, reporting each as current or
-  not-current (never treating an older-but-working deployment as broken). Adds the SDK helper
-  `isCurrentGenerator` alongside the existing `isCurrentRenderer` probe.
+## 0.1.0-alpha.43
 
 ### Patch Changes
 
-- c068a0f: `abx storage upload <path>` now stats the path before reading it, so a directory, a missing file, or
-  an unreadable file gets an actionable error instead of a raw Node errno (`EISDIR: illegal operation
-on a directory, read`, a bare `ENOENT` stack). The directory message points at the commands that
-  actually upload folder content — `abx deploy-series --dir` and `abx deploy-code --code-dir` — since
-  `storage upload` only ever handles a single file.
-- c068a0f: `abx submit-app` no longer fails silently on RPC catch-up: the mint's token id is now recoverable
-  even under `--json` if metadata configuration can't proceed, the first metadata write after a fresh
-  mint retries a few times with backoff before giving up, and a caller who exhausts that bound gets an
-  explicit message naming the token id and confirming a re-run picks up exactly where it left off
-  (never a duplicate mint — the collection is already `claimed`). `--json`'s `withJson` helper also now
-  flushes whatever payload a command already `emit()`-ed before a later step throws, instead of leaving
-  `--json` callers with nothing on stdout when a command fails partway through.
-- Updated dependencies [c068a0f]
-- Updated dependencies [c068a0f]
-- Updated dependencies [c068a0f]
-- Updated dependencies [c068a0f]
-- Updated dependencies [c068a0f]
-- Updated dependencies [c068a0f]
-- Updated dependencies [c068a0f]
-- Updated dependencies [c068a0f]
-- Updated dependencies [c068a0f]
-- Updated dependencies [c068a0f]
-- Updated dependencies [c068a0f]
-  - @artblocks/abx-token-api@0.1.0-alpha.34
-  - @artblocks/abx-sdk@0.1.0-alpha.31
-  - @artblocks/abx-indexer@0.1.0-alpha.32
-  - @artblocks/abx-storage@0.1.0-alpha.31
-
-## 0.1.0-alpha.40
-
-### Patch Changes
-
-- 64248d9: Establish the public source baseline and refresh dependency constraints.
-- Updated dependencies [64248d9]
-  - @artblocks/abx-indexer@0.1.0-alpha.31
-  - @artblocks/abx-sdk@0.1.0-alpha.30
-  - @artblocks/abx-storage@0.1.0-alpha.30
-  - @artblocks/abx-token-api@0.1.0-alpha.33
-
-## Unreleased
-
-Public release history begins here.
+- 9242a05: Prepare package metadata, release notes, and public-facing source comments for the initial public
+  source release.
+- Updated dependencies [9242a05]
+  - @artblocks/abx-indexer@0.1.0-alpha.34
+  - @artblocks/abx-sdk@0.1.0-alpha.33
+  - @artblocks/abx-storage@0.1.0-alpha.33
+  - @artblocks/abx-token-api@0.1.0-alpha.36

package/dist/commands/deploy.js

@@ -34,7 +34,7 @@ import { describeSchema, editionSchemaAdvisory, parseSchemaSpecs } from '../sche
 import { parseSeriesTraits } from '../series-traits.js';
 import { decodeOnChainJson } from '../served.js';
 import { openWalletSession, signHotSequence, signTx } from '../signer.js';
-// ── plan-object warning capture (issue #123) ─────────────────────────────────────────────────────
+// ── plan-object warning capture ──────────────────────────────────────────────────────────────────
 // The structured `--json` plan object (deploy-plan.ts) reports "the warnings the lane raised"
 // alongside every other computed field — but ~400 `warn()` call sites in this file exist purely as
 // human prose, and rewriting each to ALSO push onto a collector would be exactly the kind of
@@ -926,7 +926,7 @@ export async function cmdDeployBody(flags, serveAfter, emit) {
     // never disagree. Counts TX signatures only: a storage upload signed by a connected wallet
     // (Arweave via --storage-signer eth) is a message signature, not a transaction, and is already
     // named separately (see the `remoteEthUpload` narration) — it does not add to this count.
-    // Also captures the plan object's `custody.image` file metadata (issue #123 follow-up): the SAME
+    // Also captures the plan object's `custody.image` file metadata: the SAME
     // bytes read here for the tx-count math, hashed/typed locally (no chain read) — never re-derived
     // at the emit site below.
     let planImageFile = null;
@@ -1073,7 +1073,7 @@ export async function cmdDeployBody(flags, serveAfter, emit) {
         // could act on that the actual deploy will not land at. `saltPinned` still says why.
         emit(jsonSafe({
             command: 'deploy', dryRun: true, sent: false, address: explicitSalt ? predicted : null, chain: CHAIN, chainId: resolveChain(CHAIN).id, factory, salt, saltPinned: !!explicitSalt, name, symbol,
-            // The structured plan (issue #123) — built from the SAME locals the readout above just
+            // The structured plan — built from the SAME locals the readout above just
             // printed; nothing here is re-derived or newly computed. `deployer` is a real wallet address
             // regardless of salt pinning (only the PREDICTED CONTRACT address carries that caveat, and
             // this plan doesn't repeat it — see `address`, above), so roles are safe to report unconditionally.
@@ -1538,7 +1538,7 @@ export async function cmdDeployOneOfOneEditionBody(flags, emit) {
     // The edition paths shipped without this and without the readout line below, so `--copies` previews
     // silently dropped the approval count the skill promises "every preview" prints — leaving an agent
     // with nothing to tell the creator about how many wallet prompts to expect.
-    // Also captures `custody.image` file metadata (issue #123 follow-up) — see the 1/1 twin's identical
+    // Also captures `custody.image` file metadata — see the 1/1 twin's identical
     // local for why it's computed here rather than re-derived at the emit site.
     let planImageFile = null;
     const approvals = onchainImage