saffron-ai 1.3.2 → 1.3.4

This diff represents the content of publicly available package versions that have been released to one of the supported registries. The information contained in this diff is provided for informational purposes only and reflects changes between package versions as they appear in their respective public registries.
package/CHANGELOG.md CHANGED
@@ -5,6 +5,388 @@ the VS Code extension (`ChathurangaJayasinghe.saffron-vscode`) and the
5
5
  JetBrains plugin (`ai.saffron.jetbrains`). Dates are publish dates. Each
6
6
  release heading names its main feature: `## <version>: <title>`.
7
7
 
8
+ ## 1.3.4: Checks that wait for a settled page, and credentials kept masked
9
+
10
+ Released 2026-11-07.
11
+
12
+ Runner
13
+
14
+ - An Outline row no longer passes on another row's data in three more
15
+ ways: a check naming a column that no step of the recording writes ("the
16
+ `<plan>` plan shows ...") binds the recording to its row, a row's number
17
+ is found where the page shows it grouped or with a decimal comma ("Up
18
+ to $1,000" for 1000), and a request body's number is taken for a
19
+ constant only where the body itself writes the column. A style check's
20
+ property is a value a row can hold. A check run in the session a column
21
+ names (`As "<who>"`) follows that column; one run in another session
22
+ (`As "admin"`) does not, and a step writing the column counts only when
23
+ it comes before the check in the same session. A row's number is also
24
+ found negative ("−$1,000") and grouped by an apostrophe or a no-break
25
+ space ("CHF 1'000").
26
+ - An amount check fails when the target shows another currency than the
27
+ captured amounts, and captures in two currencies are refused. A symbol
28
+ is compared with the codes it can stand for: "$5" may be USD 5 or CAD 5,
29
+ never NOK 5 or EUR 5, "€" is EUR, and a dollar its letters name is that
30
+ dollar ("US$", "CA$", "A$", "R$" and others); every capture is compared
31
+ with every other. A lowercase word that is also a code ("Top 10",
32
+ "150 eur") is compared with codes only. A URL check reads a
33
+ path whole past `-`, `.`, `_`, `~` and `+`, and past `:` before a letter
34
+ ("/checkout" is not "/checkout-failed" or "/checkout_failed"; a host
35
+ before its port, "localhost:3000", still matches), and a word starts as
36
+ it ends: after a letter and a separator it is part of another ("paid"
37
+ is not in "/orders/not-paid"; in a query "+" and a space still separate
38
+ words, so "shoes" is in "?q=red+shoes"), a number also after a
39
+ digit ("42" is not in "/1-42"), while an id-prefixed slug keeps its words
40
+ ("blue-shirt" is in "/products/42-blue-shirt") and "example.com" is in
41
+ "shop.example.com". A doubled separator ("/checkout--failed") is one, in
42
+ replay and in specs `saffron eject` writes. A committed check of a path
43
+ the app shows with a slug after it ("/products/42" on
44
+ "/products/42-blue-shirt"), or of one word inside a slug ("shirt" or
45
+ "US" in "/en-US/cart"), fails after upgrading: record it again with the
46
+ whole slug, or check the id another way. A currency code is read beside the
47
+ number, capitals first, so "Top up NOK 150" is in NOK, and "NET 150
48
+ EUR", "150 eur" and "EUR (150.00)" are in euros. Text inside `<noscript>`, or inside an SVG
49
+ definition never drawn where it stands, is not shown text, whether the
50
+ check's target is the container or that element itself.
51
+ - API tests: a group's `same` check compares values exactly (two int64
52
+ payment references a digit apart differ, as do 0.1 and
53
+ 0.10000000000000001, and a number and a string of its digits), a number
54
+ in a body that JSON cannot hold as written is refused, in any spelling
55
+ (`9007199254740993`, `9007199254740993.0`, `9.007199254740993e15`,
56
+ `0.10000000000000001`; it would be sent rounded to another value: send
57
+ it as a string). A row or capture that held such a number and passed
58
+ before is refused after upgrading, a `differs` check whose row value is
59
+ empty is refused, an expected decimal with more digits than a number
60
+ holds is refused, and a group's contract check ties the recording to
61
+ the contract, so `api pin` re-records it.
62
+ - A provider's daily quota, "exhausted your capacity" and "hit your
63
+ limit" stop the run's AI sessions as other usage limits do. A key a
64
+ vendor quotes with its middle starred or elided (`sk-...XwVu`) is
65
+ masked, every Cursor error masks CURSOR_API_KEY, including one the SDK
66
+ throws, and an Antigravity error that ends in JSON names its
67
+ message. A `verify` scenario path that leaves the project through a
68
+ link is refused.
69
+ - Specs `saffron eject` writes open web pages only (never `data:`,
70
+ `file:` or `javascript:`) and upload only the project's own files (no
71
+ dotfiles, nothing outside the project, no link to a dotfile or out of
72
+ the project, no hard link, no folders), as replay does; they refuse a
73
+ `differs` whose value a blank row or capture empties, and a body number
74
+ they cannot send exactly. An Examples cell holding such a
75
+ number marks its test fixme instead of stopping the eject.
76
+ - A `"""` or a triple backtick at the start of a description line no
77
+ longer opens a doc string, which left every `StepSet` line after it
78
+ unread: as in Gherkin, a doc string opens only after a step in a
79
+ scenario, Background, StepSet or Fixture, never in a Feature, Rule or
80
+ Examples description, in the runner and the language server alike.
81
+ - An API that answers with a refused connection in its body or a field
82
+ (its own database's) is a failure, never "not run, environment
83
+ missing": only a request that got no response, or the browser's own
84
+ navigation error, says nothing answered.
85
+ The dashboard's recent failures pass over reds the app did not cause
86
+ when telling whether a scenario still fails and for how long.
87
+ - Recording: a credential an API handed out during the session is masked
88
+ in everything the browser tools show the agent (a token in a page URL
89
+ was shown whole, and an earlier value of a credential that changed); a
90
+ page the agent opens with an `{env:...}` value in its URL must be on the
91
+ app's origin or one the scenario writes, a URL with no scheme read as
92
+ the browser opens it; code a check runs through `browser_evaluate` that
93
+ clicks, submits, sets a value, navigates or sends a writing request is
94
+ refused, as a click in a check is;
95
+ with a browser already running (`transport.browser` or a
96
+ `transport.module`), recordings run one at a time, as two sessions on
97
+ one browser closed each other's pages.
98
+ - A password or user name in a base URL is masked wherever a URL quotes
99
+ it (`https://qa:***@staging.test`): the report's errors, the run
100
+ history, proposals, `status`, `diff`, bundles, traces and the API
101
+ transcript, and the agent is never shown it. Only there: the same word
102
+ elsewhere in a message is left as it is.
103
+ - A local notation beside the number (`100 kr`, `Rs. 500`, `5,00 zł`) is a
104
+ currency of its own in an amount check, so captures in "kr" fail a total
105
+ shown in euros. A URL check reads a percent-encoded letter beyond a
106
+ separator as a letter ("/checkout" is not in "/checkout-%C3%A9chou%C3%A9").
107
+ - A proposal or recording the run cannot read (no permission, a folder in
108
+ its place) is listed as unreadable instead of stopping `status`, `diff`,
109
+ `accept --all` and the run's closing summary. The editor flags a
110
+ `{data:...}` reference only where the runner reads it (a step, its table
111
+ or doc string), never in a title, a description or a comment.
112
+ - A Then recorded again under the same words that now expects another
113
+ value ("Pending" where it checked "Paid", a status 500 where it expected
114
+ 201, another JSON value or URL), or reads another element named the same
115
+ way (test id "status" become "badge"), is held for review as a weakened
116
+ check (a row or data token holding the same value is the same): a stale re-record no longer replaces a check that just
117
+ failed. A heal that changed nothing is red, never a proposal, also with
118
+ `--no-verify`. The usage an agent session spent stays counted when its
119
+ proof replay cannot start.
120
+ - An Outline row's short value (a word under 5 characters with no digit,
121
+ such as "dev") is written as its `<column>` only where it stands alone:
122
+ inside "devices panel" it stays as recorded, so another row no longer
123
+ checks a text never on the page. A secret Playwright cut short with "…"
124
+ in a trace line, or a captured credential cut short in a message, is
125
+ masked.
126
+ - Replay: a "not shown" check looks inside the app's own iframes too (its
127
+ origin, or one it writes, shown on the page), as a "shown" check does, so
128
+ an error drawn in one no longer passes both; another origin's frame (a
129
+ chat widget's "Close") and a hidden one (a silent sign-in renewal) are
130
+ not searched. A "not shown", count, "differs" or "URL does
131
+ not contain" check, and a response check, passes only once the requests
132
+ the last action started have answered and the passing state has held for
133
+ half a second: a list still growing, a placeholder drawn during a
134
+ refresh or a row gone for a moment no longer passes. A long-poll the page
135
+ opened earlier is not waited for. A response check fails when the very
136
+ request it matched (its URL and method, and the body or JSON it checks)
137
+ was answered again with another status (a POST answered 200 and then
138
+ 500); a refetch, a child path or a GraphQL query on the same endpoint is
139
+ another request.
140
+ - `accept --step-set` refuses a StepSet that ran in a named browser session
141
+ (`As "seller"`) and a proposal that fails an Examples row: filed for
142
+ reuse, its steps would seed other sessions' or rows' steps. A Nordic
143
+ price ("150,- kr", "150:- kr") keeps its currency. A base URL's Basic
144
+ header an API echoes is masked; a URL in a query beside an email is not
145
+ taken for a password; step text accept writes into the feature file
146
+ keeps its URLs whole.
147
+ - Recording: a capture's target is read as replay reads it before it is
148
+ recorded, and a value it does not show is refused, so a later check never
149
+ compares with another element; a status the step names exactly ("answer
150
+ 404") is expected exactly, never as its class, and a status check names a
151
+ status the step states; a step performed again drops its requests'
152
+ responses too, so a later check never reads a dropped one; a response
153
+ check is verified with the step's table values, as replay reads it.
154
+ - Config: a key that names a known one misspelt (`maxCostUSD`, `stict`) is
155
+ refused, naming the key meant, and any other unknown key is said; a
156
+ `capabilities` value in quotes, an unknown `snapshotMode` and a
157
+ `hookTimeoutMs` of 0 are refused; an empty `SAFFRON_ENV` keeps the
158
+ config's `env`. `status --json` and the MCP `project_status` mask a
159
+ password in any URL and a header value written whole. `run --plan`
160
+ judges a pending proposal as the run does: a stale one is a recording,
161
+ one filed on another machine is not replayed, a row-bound one is
162
+ refused. The MCP `verify` tool says when the run did not finish cleanly
163
+ (interrupted, scenarios not run, a failed hook).
164
+ - A heal whose agent session could not start keeps the replay's failure,
165
+ never "not run"; a fixture that failed in the app fails the scenarios
166
+ behind it, never "environment missing"; the usage earlier agent
167
+ sessions of a recording spent stays counted when a later one fails.
168
+ - A password in the base URL, or a user name alone (`https://TOKEN@host`),
169
+ is masked in the report, the run history and bundles. A run with `--filter` or of part of the features folder is
170
+ partial in the history, so trends compare suite runs only. A red the
171
+ app did not cause never makes a scenario flaky on the dashboard.
172
+ - A CSV data file with a quote never closed is refused, naming the row,
173
+ instead of reading every line after it as one field. The language
174
+ server no longer stalls on a line holding a long run of spaces.
175
+ - Replay lets the first candidate that finds the recorded element decide a
176
+ check, shown or not: a hidden recorded element no longer passes a text,
177
+ visible, value or enabled check on a fallback elsewhere, a list check
178
+ (`every {data:...}`) no longer passes on a fallback or another frame
179
+ listing more, and a watched message counts only on the element watched,
180
+ never on a log line elsewhere. A currency-first value (`$5.00`) in a
181
+ pattern no longer matches its negative (`-$5.00`).
182
+ - A refused API request's reply to the agent masks a credential an API
183
+ handed out in the URL, as the reply to a request that answered does.
184
+ - Restoring a column over a secret no longer splices a short env value
185
+ into a word (`dev` in `/devices`); a row value an API request's path
186
+ holds URL-encoded is written back as its column, as in a browser URL.
187
+ - A red the agent cannot fix (an unreadable recording, a row refused, the
188
+ committed recording failing while another row's proposal waits) counts
189
+ as a failure in the report and the run summary, never as "needs
190
+ recording" with a command that would change nothing.
191
+ - The HTML report sets the reds the app did not cause apart from
192
+ failures: a stale recording in a run with `--no-agent`, a missing
193
+ environment, an agent that could not start or a run over its budget.
194
+ They get their own *not run* group under the failures (*needs
195
+ recording* when that is the only reason), a ○ on the run map, and a
196
+ *failed* count that says how many more did not run and why. Every
197
+ scenario that needs recording names the others, with one `saffron run`
198
+ command for their feature files. The run map counts them per feature
199
+ file, and the recurring failure themes leave them out (run history now
200
+ keeps the mark). They are still red, and the run still exits 1. Found in
201
+ real-world use, where one edit to a shared StepSet made 45 of 48 reds
202
+ stale recordings.
203
+ - A request group's step seeds another scenario that sends the same
204
+ requests with its own values: the same table header, request names and
205
+ rows, each cell read through `<row:column>`, and so does a group check
206
+ that reads each response's own row. Before, a group step was never
207
+ seeded, so every scenario recorded it with the agent.
208
+ - A StepSet accepted on its own (`accept --step-set`) after a scenario's
209
+ recording was made seeds the steps that come from that StepSet over the
210
+ scenario's older recording, with no `seeded from another recording:`
211
+ note; the same words written straight in a scenario keep its own.
212
+ - `saffron status --json` lists the StepSets each proposal recorded
213
+ (`stepSets`), for an editor to offer `accept --step-set`.
214
+ - `saffron eject` exports a count check as Playwright's count of shown
215
+ elements, `expect(locator.filter({ visible: true })).toHaveCount(n)`,
216
+ the number as replay reads it (`010` is 10).
217
+ - A pattern check that fails where the pattern writes an optional part
218
+ right after a value says the value is matched whole and how to write the
219
+ part instead.
220
+ - A currency code after a number counts as a price in a locator's name
221
+ only after two digits or more: "Section 4 SEK" is no price.
222
+
223
+ Editors (VS Code extension 0.8.1, JetBrains plugin 0.8.1)
224
+
225
+ - Both editors offer `saffron export` (the review bundle), a draft pull
226
+ request with the pending proposals (`export --branch`, after asking for
227
+ a branch and confirming), `saffron eject` (after asking, for the open
228
+ file or every one) and `accept --step-set` for one ticked proposal's
229
+ StepSet, from the Proposals view's menu (VS Code) or toolbar
230
+ (JetBrains).
231
+
232
+ ## 1.3.3: Concurrent API requests, and recordings that follow their data
233
+
234
+ Released 2026-11-05.
235
+
236
+ Runner
237
+
238
+ Upgrading from 1.3.2:
239
+
240
+ - Recordings with a request group (several API requests sent at once) are
241
+ schema version 19, and recordings with a count check use an action
242
+ 1.3.2 does not know: 1.3.2 refuses both rather than replaying them
243
+ wrong. Upgrade every machine that replays them, CI included.
244
+ - Two new notes hold a proposal from `accept --all`: `check weakened:`
245
+ (a `Then` recorded again with a looser check than the committed one)
246
+ and `data not followed:` (a step that names a `{data:...}` value its
247
+ recording keeps as text). Accept such a proposal by its file after
248
+ reading why.
249
+ - While recording, each browser check the agent writes is run on its page
250
+ by replay's own engine, and refused when it does not hold there. A
251
+ recording that used to fail its proof on such a check now fixes it
252
+ before the proof.
253
+ - A scenario recorded again because its recording is stale seeds the
254
+ steps it did not change from its own recording, never from another
255
+ scenario's newer one.
256
+
257
+ - A downloaded file's check resolves a `{captured:...}` value, while
258
+ recording and at replay, so "the calendar file contains the booking
259
+ number `{captured:bookingNumber}`" checks this run's booking; one that
260
+ nothing captured fails saying so. Before, replay checked for the token
261
+ as text and the proof failed every file.
262
+ - A date the scenario writes as a token stays that token when the agent
263
+ types it glued to a time (`20261111T140000Z` for `{date+7:YYYYMMDD}`),
264
+ and a pattern that writes the time after it (`{date+7:YYYYMMDD}T[0-9]{6}`)
265
+ matches it. A time written as optional (`(?:T[0-9]{6})?`) still ends the
266
+ date, so the step writes it as required. A date left as text
267
+ in the format the scenario's own token writes gets `date not following
268
+ the clock:` and is held from `accept --all`.
269
+ - A short data value the step names (`{data:party.size}`, 2) is written
270
+ back as its reference where it stands alone once in a check's or fill's
271
+ value, a text pattern or an amount ("Table for 2 people", `"price" *
272
+ 2`), never in a selector, URL or path, nor a digit of regex syntax, a
273
+ time or a date. A step whose recording still holds a value it
274
+ names as text, or a list's items one by one, gets a new `data not
275
+ followed:` note and is held from `accept --all`. A price in a locator's
276
+ name (`Info Lunch SEK 150`) is a `volatile locator:`.
277
+ - The language server sees a StepSet's steps filed with `accept
278
+ --step-set` (`.saffron/step-sets`) as recorded when they change.
279
+ - The plugins' keyword tables carry each language's Feature, Background
280
+ and Rule, so a file in another Gherkin language ends a scenario at them.
281
+ - A `StepSet:` or `Fixture:` in a file written in another Gherkin
282
+ language (`# language: de`) defines its set or fixture: it was read as
283
+ the English `Scenario:`, which that language does not know, so no set
284
+ was found. A StepSet is called in a language with no `*` step too
285
+ (Slovenian), and the `# language:` line is read as Gherkin reads it:
286
+ `# language: de # Deutsch` is English to both.
287
+ - The language server reads an `As "..."` session through Background and
288
+ Rule headers in the file's own language, and ends one at a `Fixture:`,
289
+ as the VS Code extension does.
290
+ - A file the agent uploads while recording is found in the project when
291
+ its path is relative, wherever the run was started: `saffron -p app
292
+ run` from a monorepo's root refused `demo-app/sample-upload.txt` as
293
+ outside the project.
294
+ - The benchmark's recording measure counts every note held from `accept
295
+ --all`, and names the scenarios not verified on the first attempt or
296
+ held, with why.
297
+ - A browser check the agent writes is run on its page by replay's own
298
+ engine before it is recorded, with replay's values, captures and clock,
299
+ and refused with replay's reason when it does not hold: a value not
300
+ whole in its text or an element not shown fails while recording, not
301
+ at the proof.
302
+ - A `Then` recorded again under the same words with a looser check than
303
+ the committed one gets a `check weakened:` note and is held from
304
+ `accept --all`: a value it no longer reads (the same value written
305
+ another way still counts), a pattern's or name pattern's lost `^` or
306
+ `$`, a text it checks less of, a kind of check it no longer makes or
307
+ covers, a JSON field it no longer compares, an element it no longer
308
+ names. `--rerecord` keeps the recording it removes for this comparison.
309
+ - A scenario recorded again because its recording is stale seeds each
310
+ step it did not change from its own committed recording where that fits
311
+ and the page shows what it acts on, never from another scenario's newer
312
+ one (a shared StepSet's update brought those in). A step that replayed
313
+ another scenario's recording is named in a `seeded from another
314
+ recording:` note.
315
+ - A check of an element that runs out of time (shown, text, pattern,
316
+ value, style, attribute) says what replay saw last: nothing matching
317
+ the target, an element there but not shown, or the text or value it
318
+ showed, with secrets masked before it is cut.
319
+ - A count check: "3 rooms should be listed" is one check of how many
320
+ shown elements the target finds in the page (or the frame it names),
321
+ from a number, a `<column>` or a `{data:...}` value; a hidden element
322
+ never counts, and none is written as "not shown". `saffron eject` does
323
+ not export it yet, and saffron-ai 1.3.2 refuses a recording with one as
324
+ an action it does not know, never replaying it wrong.
325
+ - The report page's headline counts and cost are this run's own after a
326
+ run of named scenarios, with the number of rows kept from earlier runs
327
+ beside them; before, it added up the kept rows' results and AI calls.
328
+
329
+ - Several API requests at once, each with its own values, and checks over
330
+ all their responses: an idempotency test that registers one charge from
331
+ several requests at the same moment. A step's table names the requests,
332
+ a header row and a row per request (`When I send these registrations
333
+ concurrently as "registrations":`); the request is recorded once and
334
+ sent once per row, every row before any answers, each row's cell
335
+ written as `<row:column>`, so editing a cell needs no recording again;
336
+ a `<row:column>` naming no column of the table is refused rather than
337
+ sent as text, and the group's name may come from an Examples column.
338
+ Every response stays with its row whatever order it arrives in, a
339
+ request that gets none is kept and fails the step once all have
340
+ settled, and a group is never sent again. `Then` steps check every
341
+ response, those of one status, a named one, at least some, or all of
342
+ them sharing a field's value (`all 200 responses should have the same
343
+ "/paymentReference"`); a response's own cell in a check is its row's.
344
+ The common checks are exact phrases recorded with no agent. A failure
345
+ lists each request with its status and why it failed. "The response"
346
+ after a group fails instead of reading one of them. Recordings with a
347
+ group are schema version 19: saffron-ai 1.3.2 and earlier refuse them.
348
+ `saffron eject` does not export a group yet.
349
+
350
+ - `saffron accept --propagate` fixes a healed API request everywhere it
351
+ is sent, as it already did for a locator. Scenarios that share a step's
352
+ wording hold copies of its request; when a heal rewrites that request
353
+ (a field the API now requires, a renamed path), accepting it with
354
+ `--propagate` rewrites every other cache that sends the same request
355
+ under the same step text, so one heal fixes them all instead of each
356
+ scenario healing on its own run. `--propagate` goes on the accept itself:
357
+ without it, accept names the caches it left as they are, which heal on
358
+ their own run. Either way it prints what the request was and is now. A
359
+ fix is not copied when it writes a value of the healed scenario's own (a
360
+ table cell, a doc string, its Outline row), adds a `<column>` or
361
+ `{captured:...}` value, drops a value the request extracted, sends the
362
+ request to another API, or when the heal kept the request as it was in
363
+ one place and rewrote it in another; accept says which. A step whose
364
+ heal added or removed an action is not copied either.
365
+ - `saffron accept` without `--propagate` no longer tells you to re-run it
366
+ with `--propagate`: the proposal is accepted by then, and a re-run had
367
+ nothing left to fix. It says the other caches were left to heal on
368
+ their own run.
369
+
370
+ Editors (VS Code extension 0.8.0, JetBrains plugin 0.8.0)
371
+
372
+ - Both editors mark a proposal with a `check weakened:` or `data not
373
+ followed:` note, or one that fails an Examples row, as held from Accept
374
+ All, as `saffron accept --all` skips it. JetBrains asks before Accept
375
+ All runs, as VS Code does.
376
+ - In a file written in another Gherkin language (`# language: de`),
377
+ Right-click Run reads that language's Feature, Background and Rule
378
+ headers (JetBrains), and VS Code's completion, recording badges, hovers
379
+ and wording hints read its step keywords and headers, a keyword with no
380
+ space of its own ("假如") included.
381
+ - A `Fixture:` block is a block of its own: it ends the StepSet above it
382
+ (VS Code), and Right-click Run inside it runs no scenario (JetBrains).
383
+ - VS Code shows the steps a StepSet's proposal filed for reuse
384
+ (`.saffron/step-sets`) as recorded, and no longer marks `Examples:
385
+ {data: roles}` as an error when a description line and then a table
386
+ follow it.
387
+ - JetBrains passes a run's files after `--`, so a file whose name starts
388
+ with a dash is a file, never an option.
389
+
8
390
  ## 1.3.2: Exact checks for amounts, URLs and numbers, and secrets kept masked
9
391
 
10
392
  Released 2026-11-04.