PRQL compiler CLI - prqlc: baixar versions anteriors

0.13.14 has 80 commits from 7 contributors. Selected changes:
Features:
- Add a PRQL-native implementation of append by:name, which aligns columns by name rather than position and works on any dialect. Columns present in one relation but not the other are filled with NULL. Supporting this added wildcard expansion of relations via their lineage, plus tuple_uniq and tuple_reverse. (@kgutwin, #6046)
Fixes:
- Generate BETWEEN again. The optimization compared whole rq::Expr values, including their spans, so the same column referenced at two source positions never compared equal and a >= 5 && a <= 10 was left unoptimized. (@queelius, #5738)
- Don't leak a joined branch's sort into the implicit ORDER BY that take generates. A join now retains the left side's order, as documented, rather than emitting a column that is no longer in scope once the joined branch is materialized as a CTE. (@lukapeschke, #6015)
- Return a compile error instead of panicking when append is given relations with different numbers of columns, or with incompatible column types. (@prql-bot, #6025)
- Set the relation type of from_text earlier in the resolver, fixing a crash in infer_type_of_special_func. (@kgutwin, #6066)
- Give wrapper params unique names when currying through nested let-bound functions, which previously bound a curried argument to the wrong parameter. (@prql-bot, #5979)
- Escape quotes when a string contains both quote characters and one sits at each boundary, so formatted output lexes back to the original. (@prql-bot, #5987)
- Say "identifier" rather than "string literal" in the target error message. (@prql-bot, #6081)
- Sync the KSyntaxHighlighting and book highlighters with the std lib, and drop a stray sub-rule that wrongly highlighted f-strings. (@prql-bot, #6028, #6095)
Documentation:
- Numerous typo, grammar, and stale-content fixes across the book, tutorials, bindings docs, and code comments.
Integrations:
- Sync the prqlc Python type stub with the binding signatures. (@prql-bot, #6058)
- Correct ArgumentException constructor usage in the .NET binding. (@prql-bot, #5954)
- Bump the Elixir NIF's rustler crate to 0.38, matching the hex library. (@prql-bot, #6078)
Internal changes:
- Publish to crates.io with Trusted Publishing. publish-to-cargo mints a short-lived credential per run through GitHub Actions OIDC instead of reading a stored token, so no long-lived publish credential exists. (@max-sixty, #6100)
- Update the tend (Claude-powered CI) workflows; most of this release's bot-authored fixes were filed by tend.
- Update the Rust toolchain version. (#6073)

0.13.13 has 127 commits from 5 contributors. Selected changes:
Fixes:
- Correctly resolve wildcards on this and carry column aliases through natural joins, by fixing how the compiler's internal tuple_map applies redirects, folds expressions, and propagates aliases. (@kgutwin, #5875, #5877, #5980)
- Resolve identifiers up the parent module chain. (@prql-bot, #5976)
- Return a compiler error instead of panicking in lookup_cid. (@prql-bot, #5938)
- Catch a panic in the CLI's compile path so the debug log is still written. (@kgutwin, #5869)
- Set the minimum chrono version to 0.4.40 for compatibility with arrow-arith. (@prql-bot, #5841)
Documentation:
- Numerous typo, grammar, and stale-content fixes across the book, tutorials, bindings docs, and code comments.
Integrations:
- Fix several FFI defects in the .NET binding that prevented it from working correctly on 64-bit platforms or returning more than one diagnostic: map size_t as UIntPtr (was int), advance the pointer when iterating the messages array, dereference indirect string fields and Span / Location pointers, call result_destroy to free native memory, and tighten exception types and doc references. Also guard result_destroy in prqlc-c against a null messages pointer, which is set on the success path. (@prql-bot, #5848)
- Modernize the .NET binding to net10.0 using LibraryImport source generators. (@prql-bot, #5850)
- Throw a Java exception instead of panicking in the JNI bindings. (@prql-bot, #5934)
- Exit the LSP server only on the exit notification, matching the LSP spec. (@prql-bot, #6001)
Internal changes:
- Update the tend (Claude-powered CI) workflows; most of this release's bot-authored fixes were filed by tend.
- Scope release secrets to a protected GitHub environment. (@max-sixty, #5910)
- Replace the internal tuple_every helper with a more general tuple_reduce. (@kgutwin, #5878)
- Update the Rust toolchain version. (#5966)

0.13.12 has 81 commits from 10 contributors. Selected changes:
Language:
- Remove the deprecated std.prql_version function. Use prql.version instead, which has been the supported replacement since 0.11.1. (@prql-bot, #5806)
Features:
- Add a very early sql.oracle dialect. It currently only forces identifier quoting to accommodate Oracle's case-folding rules; most other features fall back to generic SQL. (@julien-pinchelimouroux, #5821)
- Add date.trunc function for date truncation. (@happyso, #5729)
- Add date.now function for current timestamp. (@happyso, #5721)
- Add date.diff function for date differences. (@happyso, #5726)
- Add DuckDB target to the Elixir bindings. (@prql-bot, #5817)
Fixes:
- Loosen the chrono constraint from 0.4.44 to 0.4 to avoid version conflicts with downstream crates such as polars. (@lukapeschke, #5834)
- Correct the SQLite div_i formula so integer division returns the right result when |dividend| < |divisor| (e.g., 1 // 2 now returns 0 instead of -1). (@queelius, #5736)
- Return an error rather than emitting invalid SQL when date.trunc is used against SQLite. (@prql-bot, #5733)
- Sync :duckdb into the Elixir PRQL.Native.CompileOptions target typespec so target: :duckdb actually reaches the DuckDB dialect. (@prql-bot, #5823)
- Correct the Vec::from_raw_parts element type in prqlc-c to fix undefined behavior in result_destroy. (@prql-bot, #5732)
- Prevent panic in prqlc-macros by correcting a stale macro name in the panic message. (@prql-bot, #5811)
- Align Emacs grammar keywords and regex with the lexer. (@prql-bot, #5795)
- Correct the playground's decimal regex and transforms list in syntax highlighting. (@prql-bot, #5788)
- Correct the aarch64 target path in the Java cross-compilation script. (@prql-bot, #5781)
Documentation:
- Numerous typo, grammar, and stale-content fixes across the book, tutorials, bindings docs, and code comments.
Integrations:
- Kakoune 2026.04.12 has syntax highlighting for PRQL. (@vanillajonathan)
Internal changes:
- Add tend (Claude-powered CI) workflows for autonomous PR review, issue triage, CI-failure fixes, and a nightly code-quality sweep. Most of this release's bot-authored fixes were filed by tend. (#5727)
- Update the Rust toolchain version. (#5807)
- Switch from mypy to ty for Python type checking. (#5761)
New Contributors:
- @happyso, with #5721
- @queelius, with #5736

0.13.11 has 75 commits from 7 contributors. Selected changes:
Features:
- Add support for date.to_text for BigQuery (@segv, #5712)
Fixes:
- Prevent panic on bare * in select (@max-sixty, #5696)
- Handle partial application of transforms in user-defined functions (@max-sixty, #5663; reported by @Rafferty97)
- Keep Computes with Aggregate when Filter follows (@max-sixty, #5639; reported by @matiastoro)
- Preserve sort before take in CTE (@max-sixty, #5638; reported by @lukapeschke)
- Preserve sort with empty group {} (@max-sixty, #5635)
- Return proper error for bare lambda expressions (@max-sixty, #5634)
- Prevent DISTINCT ON internal sorting from leaking past joins (@max-sixty, #5633; reported by @annashmatko)
- Fix invalid MSSQL SQL when DISTINCT and FETCH are combined (@max-sixty, #5630)
Documentation:
- Editorial tweaks (@richb-hanover, #5645)
Integrations:
- micro 2.0.15 has syntax highlighting for PRQL. (@vanillajonathan)
Internal changes:
- Use chumsky 0.12.0 for all targets (#5659)
- Update rust toolchain version (#5673)
- Fix mdbook admonition syntax for mdBook 0.5 native support (#5649)
New Contributors:
- @segv, with #5712

0.13.10 completes the npm OIDC publishing fix from 0.13.9.
Features:
- Add read_json support for DuckDB and ClickHouse (@adamnemecek, #5608)
Fixes:
- Fix text.length on Snowflake (@priithaamer, #5614)
Internal changes:
- Fix npm OIDC publishing by upgrading to Node 24 (@max-sixty, #5619)
- Fix push-web-branch workflow permissions (@max-sixty, #5620)

0.13.9 is a re-release of 0.13.8 with fixed npm publishing.
Internal changes:
- Use OIDC for npm publishing (@max-sixty, #5609)

0.13.8 has 47 commits from 6 contributors. Selected changes:
Features:
- Add named parameter support for DuckDB read_parquet (@max-sixty, #5563)
Fixes:
- Snowflake interval quoting style (@priithaamer, #5604)
- Return error instead of panicking when window rows/range is not a range (@max-sixty, #5603)
- Add ORDER BY fallback for Snowflake window functions (@max-sixty, #5583)
- DISTINCT ON to include wildcard when projection is empty (@max-sixty, #5562)
- Lineage traces through CTEs to source tables (@nightscape, #5581)
- Lineage works with unions (@nightscape, #5550)
- Handle lineage correctly when group pipeline simplifies to non-TransformCall (@max-sixty, #5584)
- Return error for join referencing inaccessible table (@max-sixty, #5587)
- Error when table variable used in scalar context (@max-sixty, #5585)
- Improve error messages for missing main pipeline (@max-sixty, #5565)
- Provide clear error for empty tuple/array in from (@max-sixty, #5564)
- Improve error message for negative numbers in transforms (@max-sixty, #5566)
Documentation:
- Close HTML comment in sort.md documentation (@max-sixty, #5597)
Integrations:
- Add grammar file for Raku (@vanillajonathan, #5576)
Internal changes:
- Upgrade mdbook to 0.5.0 (#5568)
- Update macOS runners to version 15 (#5596)
- Remove 2nd person from error messages (#5567)
New Contributors:
- @nightscape made their first contribution in #5550

0.13.7 has 8 commits from 6 contributors. Selected changes:
Fixes:
- Fix INTERVAL quoting in Redshift (@lukapeschke, #5545)
- Fix support for text.contains in sql.redshift target (@dimtion, #5549)
- Fix division in Redshift produces float (@priithaamer, #5546)
- Use || operator over CONCAT for std.concat in Redshift (@lukapeschke, #5540)
Web:
- Bump vite from 7.1.11 to 7.2.0 in playground (#5547)

0.13.6 has 40 commits from 6 contributors. Selected changes:
Features:
- Add support for Redshift dialect (@priithaamer, #5537)
Fixes:
- Fix append regression in 0.13.5 (@priithaamer, #5495)
- Fix s-string escaping (@priithaamer, #5497)
- Filter out attestation manifests in platform verification (@max-sixty, #5509)
Internal changes:
- Migrate devcontainer build from QEMU to native ARM64 runners (@max-sixty, #5506)
- Migrate from markdown-link-check to lychee (@max-sixty, #5519)
- Use zero-copy slice for binary/hex/octal number parsing (perf improvement, @max-sixty, #5488)
- Update rust toolchain version (#5536)
- Do not allow incompatible rust-version dependencies (@eitsupi, #5493)
- Pin typos to v1.37.2 (@max-sixty, #5510)

0.13.5 has 237 commits from 14 contributors. Selected changes:
Features:
- Support for SQL arrays as s[...] syntax (@robert Valek, #5312)
- Extract SQL column names from s-string tables when possible (@lukapeschke, #5310)
Fixes:
- Sort step before an aggregate step no longer requires its columns to avoid a group by clause error (@julien-pinchelimouroux, #5347)
- Always add quotes on identifiers for Snowflake dialect (@julien-pinchelimouroux, #5461)
- Join with table containing column named "source" now works correctly (@priit Haamer, #5468)
- Columns required by sorting are properly redirected (@lukapeschke, #5464)
- Ensure sorts are done on columns of the right table (@lukapeschke, #5338)
- Deduplicate selected items in gen_projection (@lukapeschke, #5305)
- Handle complex append cases (@Elouan Poupard-Cosquer, #5366)
- Improve requirement logic (@Elouan Poupard-Cosquer, #5357)
- Avoid type mismatch with Postgres in append (@Elouan Poupard-Cosquer, #5343)
- Apply column order on CTEs in append (@Elouan Poupard-Cosquer, #5323)
Documentation:
- Fix binary literal example (@ftsfranklin, #5475)
- Use correct table in grouping tutorial (@fnuttens, #5332)
Integrations:
- TEA 63.3.1, a Qt-based text editor has syntax highlighting for PRQL (@vanillajonathan, #5220)
- Micro text editor grammar is now upstream (@vanillajonathan, #5353)
- Add LSP stub (@vanillajonathan, #5197)
Internal changes:
- Upgrade parser and lexer to chumsky 0.11, providing a 7x performance improvement (#5223, #5476, #5477)
- Set Rust linker on win64, fix build crash (@kgutwin, #5345)
- Integration tests compile all dialects and diff (@kgutwin, #5344)
New Contributors:
- @Elouan Poupard-Cosquer, with #5366
- @priit Haamer, with #5468
- @robert Valek, with #5312
- @fnuttens, with #5332
- @ftsfranklin, with #5475
- @julien-pinchelimouroux, with #5347

0.13.4 is a small bugfix release.
It has 57 commits from 10 contributors. Thanks to @lovasoa & @nsiregar for their first contributions to PRQL

- Sort steps in sub-pipelines no longer cause a column lookup error (@lukapeschke, #5066)
- Dereferencing of sort columns when rendering SQL now done in context of main pipeline (@kgutwin, #5098)
- @lukapeschke, with #5066

- Add a option to the experimental documentation generator to output the docs in HTML format.
The option is given using the --format=html option. (@vanillajonathan, 4791)
- The version of the library is now read from git describe.
This doesn't affect libraries built on git tags (such as our releases), only those built when developing.
When reporting bugs, this helps identify the exact version. (@max-sixty & @m-span, #4804)
- Raw strings (r"...") are retained through prqlc fmt (@max-sixty, #4848)
- Strings containing an odd contiguous number of quotes are now delimited by an odd number of quotes when being formatted.
The previous implementation would use an even number, which is invalid PRQL. (@max-sixty, #4850)
- A few more keywords are quoted, such as user, which is a reserved keyword in PostgreSQL. (@max-sixty)

- Parentheses are always required around pipelines, even within tuples.
For example:
- prqlc compile --debug-log=log.html will generate an HTML file with a detailed log of the compilation process. (@aljazerzen, #4646)
- Added prqlc debug json-schema command to auto-generate JSON Schema representations of commonly exposed IR types such as PL and RQ. (@kgutwin, #4698)
- Add documentation comments to the output of the documentation generator. (@vanillajonathan, #4729)
- Add CLI syntax highlighting to prqlc.
You can try it as prqlc experimental highlight example.prql. (@vanillajonathan, #4755)
- Using in with an empty array pattern (e.g. expr | in []) will now output a constant false condition instead of an expr IN (), which is syntactically invalid in some SQL dialects (@Globidev, #4598)
- The Snap package previously released on the edge channel is now released on the stable channel. (@vanillajonathan, #4784)
- Major reorganization of prqlc-parser — prqlc-ast is merged into prqlc-parser, and prqlc-parser's files are rearranged, including its exports.
This is part of an effort to modularize the compiler by stage, reducing the amount of context that's required to understand a single stage.
There will likely be some further changes (more detail in the PR description). (@m-span, #4634)
- This is a breaking change for any libraries that depend on prqlc-parser (which should be fairly rare).
- Renamed prql-compiler-macros to prqlc-macros for consistency with other crates (@max-sixty, #4565)
- prql-compiler, the old name for prqlc, is removed as a facade to prqlc. It had been deprecated for a few versions and will no longer be updated. (@max-sixty)
- New benchmarks (@max-sixty, #4654)
- @Globidev, with #4598

- Add prqlc lex command to the CLI (@max-sixty)
- Add prqlc debug lineage command to the CLI, creating an expression lineage graph from a query (@kgutwin, #4533)
- Initial implementation of an experimental documentation generator that generates Markdown documentation from .prql files. (@vanillajonathan, #4152).
- Join's side parameter can take a reference that resolves to a literal (note: this is an experimental feature which may change in the future) (@kgutwin, #4499)
- Support expressions on left hand side of std.in operator. (@kgutwin, #4498)
- Prevent panic for from {} and std (@m-span, #4538)
- The browser dist files are now built with wasm-pack's web target.
As a result, they should be usable as ES Modules, through JS CDNs, and for example with Observable Framework (@srenatus, #4274).
- The syntax highlighter package for Sublime Text is now published (@vanillajonathan).
- The VSCode Great Icons icon pack extension shows a database icon for .prql files. (@EmmanuelBeziat)
- Tokei, a source lines of code counter now has support for .prql files. (@vanillajonathan)
- Add syntax highlight file for the micro text editor. (@vanillajonathan)
- @srenatus, with #4274
- @jacquayj, with #4332
- @pdelewski, with #4337
- @m-span, with #4422
- @kgutwin, with #4498

- Initial implementation of lutra, a query runner. (@aljazerzen, #4182, #4174, #4134)
- prqlc fmt works on projects with multiple files. (@max-sixty, #4028)
- Reduce stack memory usage (@aljazerzen, #4103)
- Add syntax highlight file for GtkSourceView. (@vanillajonathan, #4062)
- Add syntax highlight file for CotEditor. (@vanillajonathan)
- Add syntax highlight file for Sublime Text. (@vanillajonathan, #4127)
- sloc, a source lines of code counter now has support for .prql files. (@vanillajonathan)
- prql-compiler has been renamed to prqlc, and we've established a more consistent naming scheme. The existing crate will still be published, re-exporting prqlc, so no dependencies will break.
A future version will add a deprecation warning.
- The prqlc-clib crate was renamed to prqlc-c, and associated artifacts were renamed.
We're trying to make names consistent (ideally for the final time!), and have a plan to rename some other bindings. (@max-sixty, #4077)
- Add lots of whitespace items to the lexer, in preparation for the completion of prqlc fmt (@max-sixty, #4109, #4105)
- Table declarations (@aljazerzen, #4126)
- @kaspermarstal, with #4124

- Infer the type of array literals to be the union of types of its items. (@aljazerzen, #3989)
- prql module is added and the prql_version function is renamed to the prql.version function. The old prql_version function is deprecated and will be removed in the future release. (@eitsupi, #4006)
- Do not compile to DISTINCT ON when take n is used with group for the targets clickhouse, duckdb and postgres. (@PrettyWood, #3988)
- Fix take n rows for mssql dialect by switching from TOP to FETCH (@PrettyWood, #3994)

- _Breaking_: group's by columns are now excluded from the partition. (#3490)
- _Breaking_: round is now in the math module and needs to be called via math.round. (#3928)
- _Breaking_: lower and upper are now in the text module and need to be called via text.lower and text.upper. (#3913, #3973)
- The std.in function now supports a list of values (@PrettyWood, #3883)
- Most standard mathematical functions are now supported: abs, floor, ceil, pi, exp, ln, log10, log, sqrt, degrees, radians, cos, acos, sin, asin, tan, atan, pow and round.
Those functions are in the math module (@PrettyWood, #3909, #3916 & 3928)
- Most standard string functions are now supported: ltrim, rtrim, trim, length, extract, replace.
Utility functions starts_with, contains and ends_with are also available.
Those functions are in the text module (@PrettyWood, #3913, #3973)
- Formatting a date to a text is now available for Clickhouse, DuckDB, MySQL, MSSQL and Postgres.
A new date module has been added with the to_text function (@PrettyWood, #3951, #3954 & #3955)
- Fix an issue with arithmetic precedence (@max-sixty, #3846)
- + and - can be used after a cast (@PrettyWood, #3923)
- The Lezer grammar had plenty of improvements and fixes. (@vanillajonathan)
- The Playground now uses Vite. (@vanillajonathan)
- Bump prql-compiler's MSRV to 1.70.0 (@eitsupi, #3876)
- @PrettyWood, with #3883

- The std.sql.read_csv function and the std.sql.read_parquet function supports the sql.glaredb target. (@eitsupi, #3749)
- Fix the bug of compiling to DISTINCT ON when take 1 is used with group by for the targets sql.clickhouse, sql.duckdb and sql.postgres. (@aljazerzen, #3792)
- Enable integration tests for GlareDB. (@eitsupi, #3749)
- trapd00r/LS_COLORS, a collection of LS_COLORS definitions colorizes .prql files. (@vanillajonathan)
- vivid, a themeable LS_COLORS generator colorizes .prql files. (@vanillajonathan)
- colorls, displays .prql files with a database icon. (@vanillajonathan)
- Emoji File Icons, a VS Code extension displays .prql files with a database emoji icon. (@vanillajonathan)
- eza, a modern ls replacement colorizes .prql files. (@vanillajonathan)
- lsd, next gen ls command displays .prql files with a database icon. (@vanillajonathan)

- _Breaking:_ Case syntax now uses brackets [] rather than braces {}.
To convert previous PRQL queries to this new syntax simply change case { ... } to case [ ... ]. (@AaronMoat, #3517)
- _Breaking_: The std.sql.read_csv function is now compiled to read_csv by default.
Please set the target sql.duckdb to use the DuckDB's read_csv_auto function as previously. (@eitsupi, #3599)
- The std.sql.read_csv function and the std.sql.read_parquet function supports the sql.clickhouse target. (@eitsupi, #1533)
- Add std.prql_version function to return PRQL version (@hulxv, #3533)
- Add support for hex escape sequences in strings.
Example "Hello \x51". (@vanillajonathan, #3568)
- Add support for long Unicode escape sequences.
Example "Hello \u{01F422}". (@vanillajonathan, #3569)
- Add support for binary numerical notation.
Example filter status == 0b1111000011110000. (@vanillajonathan, #3661)
- Add support for hexadecimal numerical notation.
Example filter status == 0xff. (@vanillajonathan, #3654)
- Add support for octal numerical notation.
Example filter status == 0o777. (@vanillajonathan, #3672)
- New compile target sql.glaredb for GlareDB and integration tests for it (However, there is a bug in the test and it is currently not running). (@universalmind303, @scsmithr, @eitsupi, #3669)
- Allow cmd-/ (Mac) or ctrl-/ (Windows) to toggle comments in the playground editor (@AaronMoat, #3522)
- Limit maximum height of the playground editor's error panel to avoid taking over whole screen (@AaronMoat, #3524)
- The playground now uses Vite (@vanillajonathan).
- Add a CLI command prqlc collect to collect a project's modules into a single file (@aljazerzen, #3739)
- Add a CLI command prqlc debug expand-pl to parse & and expand into PL without resolving (@aljazerzen, #3739)
- Bump prqlc's MSRV to 1.70.0 (@eitsupi, #3521)
- Pygments, a syntax highlighting library now has syntax highlighting for PRQL. (@vanillajonathan, #3564)
- chroma, a syntax highlighting library written in Go and used by the static website generator Hugo. (@vanillajonathan, #3597)
- scc, a source lines of code counter now has support for .prql files. (@vanillajonathan)
- gcloc a source lines of code counter now has support for .prql files. (@vanillajonathan)
- cloc a source lines of code counter now has support for .prql files. (@AlDanial)
- gocloc a source lines of code counter now has support for .prql files. (@vanillajonathan)
- The Quarto VS Code extension supports editing PRQL code blocks (prqlr is required to render Quarto Markdown with PRQL code blocks). (@jjallaire)
- @hulxv, with #3533
- @AaronMoat, with #3522
- @jangorecki, with #3634

- A new line-wrapping character, for lines that are long and we want to break up into multiple physical lines. This is slightly different from from many languages — it's on the subsequent line:
- Fix stack overflow on very long queries in Windows debug builds (@max-sixty, #2908)
- Fix panic when unresolved lineage appears in group or window (@davidot, #3266)
- Fix a corner-case in handling precedence, and remove unneeded parentheses in some outputs (@max-sixty, #3472)
- Compiler panics are now printed to the console (@max-sixty, #3446)
- Ace, the JavaScript code editor now has syntax highlighting for PRQL. (@vanillajonathan, #3493)
- Simplify & speed up lexer (@max-sixty, #3426, #3418)
- @davidot, with #3450

- Strings can be delimited with any odd number of quote characters.
The logic for lexing quotes is now simpler and slightly faster.
Escapes in single-quote-delimited strings escape single-quotes rather than double-quotes. (@max-sixty, #3274)
- S-strings within double braces now parse correctly (@max-sixty, #3265)
- New docs for strings (@max-sixty, #3281)
- Improve syntax highlighting for numbers in the book & website (@max-sixty, #3261)
- Add ClickHouse integration to docs (@max-sixty, #3251)
- prqlc no longer displays a prompt when piping a query into its stdin (@max-sixty, #3248).
- Add a minimal example for use prql-lib with Zig (@vanillajonathan, #3372)
- Overhaul our CI to run a cohesive set of tests depending on the specific changes in the PR, and elide all others.
This cuts CI latency to less than three minutes for most changes, and enables GitHub's auto-merge to wait for all relevant tests.
It also reduces the CI time on merging to main, by moving some tests to only run on specific path changes or on our nightly run.
- Auto-merge PRs for backports or pre-commit updates (@max-sixty, #3246)
- Add a workflow to create an issue when the scheduled nightly workflow fails (@max-sixty, #3304)
- @FinnRG, with #3292
- @sitiom, with #3353