[satran004]

Stacked on #209. Part of #207: PR 4 of the Script Lab series, the what-if lab.

What users get

/transactions/{hash}/fork ("Fork this transaction" on the Scripts tab) rebuilds a transaction from Yaci Store and lets you change it. Every script runs again about a quarter of a second after each edit, next to the on-chain result.

What you can edit:

• Redeemers of every script, and the datums of inputs being spent by scripts:
  • as JSON in the detailed schema ({constructor, fields}, {int}, {bytes}, {list}, {map}), with readable integers, or as raw CBOR
  • validated as you type, with per-field reset
• Required signers: remove or add key hashes.
• Validity range: from and until slots, with the UTC time shown; clear a field for no bound.
• Outputs: address, lovelace and each token amount, or remove an output.
• Mint (including new assets), withdrawals, fee.
• Rules: another epoch's cost model, or protocol 10 or 11.

What you see:

• Per script: on chain against fork, with status and CPU/memory, the % change, and Now passes / Now fails when an outcome flips.
• Error and last trace of failing forks.
• A summary: "2 changes · 1 outcome changed", with Reset.

What you can do with a run:

• Debug the fork: opens the PR 3 debugger inline, on the forked run, and reopens it after each edit.
• Export fixture: a JSON file with everything needed to run it again elsewhere:
  • script, target, cost model, limits
  • the VM arguments, ScriptContext CBOR
  • the expected result and the declared budget
  • the fork's edits

Nothing is submitted; the fork exists only in the browser.

How

• replay/fork.ts: edits are applied to the store-shaped transaction before it is rebuilt. Ledger ordering, purposes, the julc#244 context corrections and the complete redeemer map therefore stay consistent for every script without special cases. Unit-tested on a recorded Mainnet transaction.
• The debugger accepts a prepared run (prepared, syncUrl) for forks, and hides "Link to this step", since a fork isn't addressable by URL.
• Older epochs have shorter cost models (Plutus V2: 175 parameters at epoch 500, against 332 now). julc runs protocol 10/11 rules, so it fills the newer builtins' costs with the maximum, and the lab explains this when you pick such an epoch.

Example (Mainnet withdraw-zero transaction 1da5cd…)

• Setting the reward validator's redeemer to Constr 0 [] makes it Now fail (HeadList: empty list, −99% CPU); the two spend scripts are unaffected.
• Epoch 500's cost model makes the reward validator 0.02% cheaper in CPU.
• Protocol 10 gives the same budget as 11 for this transaction.

Not included

A julc-testkit Java snippet in the export. Loading a script from CBOR spans julc modules (one is in the decompiler), so a generated snippet would be fragile. The JSON fixture is language-neutral.

Tests

• npm run check: 0 errors, 0 warnings.
• vitest: 56 passing. New: fork edits for redeemers, an inline datum, signers, validity, fee, outputs, mint, withdrawals, protocol, another epoch's parameters.
• Playwright script-lab.spec.ts: all 4 tests pass against a production build in Chrome. The new one covers an edit, the summary, invalid JSON, Reset and a fixture download.
• 19 devnet smoke tests pass.

🤖 Generated with Claude Code

[beanbot-gh]

Initial automated review · Aster

Hi @satran004, thanks for the PR! 🙏 This is an automatic first check by an AI reviewer. It does not replace maintainer review and may be wrong.

Reviewed commit: 087d5ab87e06966e423116707c3e983d8f2e9ba3

Issues found:

• applications/viewer/src/lib/julc/fork/DataEditor.svelte:35 — Large JSON integers are silently changed before encoding: entering {"int":9007199254740993} runs the script with 9007199254740992, because JSON.parse has already rounded the number. Please reject unsafe numeric values and require quoted decimal strings, or use a lossless parser.
• applications/viewer/src/lib/julc/fork/ForkLab.svelte:207 — Export remains enabled while a new edit is waiting for its rerun, so it combines the current edits/epoch with the previous run's arguments, cost model and expected result. Please export the snapshot associated with the completed run, or disable export throughout the debounce and evaluation.

Not tested: Static review only; type checks, Vitest, Playwright and live julc/Yaci Store execution were not run. GitHub reported no checks for this commit.