Skip to content

build: compile with TypeScript 7 - #274

Merged
nissa-seru merged 7 commits into
anima-research:mainfrom
nissa-seru:build/typescript-7
Oct 15, 2026
Merged

build: compile with TypeScript 7#274
nissa-seru merged 7 commits into
anima-research:mainfrom
nissa-seru:build/typescript-7

Conversation

@nissa-seru

@nissa-seru nissa-seru commented Oct 10, 2026 •
edited
Loading

Copy link
Copy Markdown
Contributor

TL;DR: this builds and typechecks agent-framework with TypeScript 7.0.2, the native compiler, instead of 5.9. From the same source, the JavaScript it ships is byte-identical to 5.9.3's, and its declarations give TypeScript 5.9.3 and 5.0.4 users no new errors. One test used TypeScript's in-process parser (ts.createSourceFile), which 7's package doesn't provide, and now parses with @babel/parser.

The other Connectome repositories move to 7 alongside it, and anima-docs has used ^7.0.2 since it was created.

@antra-tess @Anarchid, this changes how the package is built, so we'd like your review, and won't merge it before 2026-10-15 14:00 UTC, so you have time to look.

Changes

  • devDependencies.typescript: ^5.7.0 → ^7.0.2.

  • tsconfig.json: "types": ["node"].

    • Why it's needed: 7 defaults types to [], where 5 included every @types package. Without the line, 7's program still gets @types/node, but only through the /// <reference types="node" /> in @types/ws, undici-types and undici's own types (tsc --explainFiles at this head). So Node's globals would have depended on a dependency's reference.
    • What it changes: under 5.9.3, nothing. Built with 5.9.3 at this head merged onto that main, with and without the line, all 996 emitted .js, .d.ts and map files are byte-identical; only tsconfig.tsbuildinfo, which records the options, differs.

  • A changelog fragment (changelog.d/typescript-7.changed.md) says what the shipped dist changes, below.

  • One test read TypeScript as a library, and now reads Babel's parser. test/stop-with-work-in-flight.test.ts, which came with fix(framework): stop() waits for an operator's operation already running, and refuses new ones #325, parsed every src file with ts.createSourceFile and walked the syntax tree for type: 'tool-result' literals, to check that each names its tool and the call it answers. 7's package doesn't provide that API: its . export is ./lib/version.cjs, and the rest are ./unstable/* entries that talk to the native binary over IPC. So with main merged, tsc failed on that file, and the test failed.

    • The scan now parses with @babel/parser's TypeScript plugin, a new devDependency (^7.29.9). Its 8.x line needs Node ^22.18.0 || >=24.11.0, and this package supports Node 20. Its only bin is parser, so 7's is the only tsc.
    • It still reads a syntax tree rather than text, now as one function, scanToolResults, whose doc says what it reads.
      • Properties: those written key: value or as a shorthand are read. A key is an identifier, or a string in any quoting, computed or not. A getter, a method, a spread and a computed variable key aren't read.
      • Values: read through as, satisfies, <T> and !. A string in any quoting is read, escapes included. One built at run time, or held in a const, isn't.
      • The checks: each tool-result literal must carry toolName. Where callId reads a call's id and toolName its name (X.id, X?.id, X['id'], through the same wrappers), they must read the same X, compared as source text. Any other callId or toolName isn't compared.
      • A source Babel can't parse fails the scan with its name.
    • It finds what the old scan found, and more. Over this head's src (99 files), the TypeScript 5 scan and this one find the same 24 literals at the same lines, with 24 pairs, none missing and none mismatched. The old scan skipped a computed key, <T> and ! around the value, and wrapped call reads, and this one reads them. Over src plus a file of ten more literals, the old scan counts 33 and this one 34. The extra literal has a computed ['type'] key, and this scan also reports it as missing toolName. Otherwise they agree: 32 pairs, and the same two mismatched. The comparison script ran both scans, with typescript@5.9.3 and @babel/parser@7.29.9.
    • A new test pins the scan: the tool-result scan reads every way the literal can be written, and reports what it should. It feeds scanToolResults two sources. Between them they hold each spelling, literals nested in an object's property and in a call's argument, the violations, and the forms the scan doesn't read. It checks exactly which literals the scan counts, which it reports missing or mismatched, how many pairs it makes, and that an unparsable source fails by name. The src test also checks that its walk reaches src's subdirectories. Each of 22 pieces of the scan and its walk, removed on its own, fails one of the two tests. Among them are every peel and guard, the property filter, the descent into nested objects, and the accumulation across sources.

  • Nothing else loads TypeScript as a library: the build, typecheck and dev scripts run tsc, which 7 ships natively. tsx, used by the examples, is esbuild-based, and no file outside ui/ imports typescript now.

  • For developers: 7 ships no tsserver, so an editor set to use the workspace's TypeScript finds no language server there.

  • CI needs no lockfile change: the root lockfile is gitignored, so npm install on each runner fetches 7's native binary from its per-platform optional dependencies. typescript@7.0.2 lists 20 of them (npm view typescript@7.0.2 optionalDependencies), linux-x64, darwin-arm64 and darwin-x64 among them. CI's legs run on ubuntu-latest and macos-latest.

  • Node 20: building with 7 needs Node 20.10 or later on the 20 line. 7's bin/tsc is an extensionless file in a "type": "module" package, and Node 20.0 to 20.9 refuse it with ERR_UNKNOWN_FILE_EXTENSION before anything compiles: 20.0.0, 20.5.1, 20.6.0 and 20.9.0 stop, and 20.10.0, 20.20.2 and 22.0.0 run it. Only a build from source meets this. This package has no prepare script, and the JavaScript it publishes runs where it did, so engines stays >=20.0.0. CI's 20 resolves to the latest 20.x. The suite built with 7 passes on Node 20.20.2 (below).

What consumers will see

Measured on this head merged onto main 3dc67b8d with #341 and #338 merged in, in that order, the merge train's next two before this one (tree 22fea31e), by building the same tree with 5.9.3 and with 7.0.2:

  • Emitted JavaScript: from the same source, all 249 emitted .js files are byte-identical to 5.9.3's: 99 from src/ and 150 from test/, since tsconfig.json compiles both into dist, which the package ships. Against the previous release, dist/test/stop-with-work-in-flight.test.js also changes, since that test now reads Babel's parser.
  • Declarations: seven of the 249 .d.ts files differ, all from src/. In each, the only difference is the quote style of string literal types, because 7 keeps the source's single quotes: readonly ['first', …] where 5 wrote readonly ["first", …]. The types are the same. The files are mcpl/push-coalescer, mcpl/references, mcpl/tool-classes, mcpl/transport, mcpl/types, modules/workspace/types and tune-out/tools.
  • Declarations, as a TypeScript 5 user reads them: the 249 emitted .d.ts files, compiled with skipLibCheck off as a TypeScript 5.9.3 user and a 5.0.4 user would, give no errors, from 7's declarations or from 5.9.3's.
  • Source maps: 167 of the 498 differ: 111 .js.map (42 from src/, 69 from test/) and 56 .d.ts.map (54 and 2). In each, only the mappings field changes, and every other field is the same. Decoded, with columns set aside:
    • .js.map: 107 of the 111 move only columns. Four change line mappings. In two, mcpl/prose-stream-router and modules/workspace/index, a generated line maps to other source lines. In two, framework and mcpl/server-connection, a generated line gains or loses its mapping. So a stack trace can land on a different column, and rarely on a different line or none.
    • .d.ts.map: all 56 change which source lines declarations map to. In 49, a line maps to other source lines. In seven, lines gain or lose a mapping. So go-to-definition can land elsewhere in the source.
  • dist/tsconfig.tsbuildinfo, the compiler's own build record, differs too. 7's names version 7.0.2.
  • A consumer on TypeScript 7: 7 defaults types to []. Take a consumer whose types leaves out node, that has skipLibCheck off, and whose program doesn't otherwise reach Node's types. With this head's packed tarball installed, a file importing the package gets 25 errors: 13 from chronicle's declarations, one from context-manager's, and 11 from this package's (Buffer, node:events, the NodeJS namespace). The same tree built with 5.9.3 gives the same 25, so this build changes nothing for such a consumer. Installing @types/node and listing node in types clears them, as this package's own tsconfig.json does. This package lists @types/node only as a devDependency, as before. With skipLibCheck on, a consumer doesn't check these declarations and meets none of them.
  • The changelog fragment names these changes to the shipped dist, and the note for a consumer on TypeScript 7.

Tests

On this head merged onto main 3dc67b8d with #341 and #338 merged in, in that order, the merge train's next two before this one (tree 22fea31e), with TypeScript 7.0.2, on the Anima VPS:

  • tsc --noEmit: no errors, and no compiler option it rejects. npm run build is clean.
  • Builds compared: 5.9.3 and 7.0.2, from the same tree, as above. Each emits 997 files.
  • The types line under 7: the 7.0.2 builds with and without it are byte-identical in all 996 emitted .js, .d.ts and map files.
  • Which 7: the bytes are 7.0.2's, the only 7.x release today. ^7.0.2 admits later 7.x patches, as every dependency's range does here, since the root lockfile is gitignored.
  • The full suite, built with 7 at that merge, on Node 24.21.0 and 20.20.2 (npm 11.19.0). Counted file by file, it passes 2291/2291 on both runtimes, against 2290 for that main built with 5.9.3 on the same host. The one file that differs is stop-with-work-in-flight, at 26 tests for 25, with the scan's new test. CI's unforced step, which runs stop-with-work-in-flight, stop-with-host-operations and conversation-routing together, passes 104/104 on both, where main passes 103. CI's forced run (npm test) passed everything it ran but counted short, as npm test's --test-force-exit can end a test file before its tests report, so a short run still passes #294's --test-force-exit cut does: on Node 24 it reported 2264 with the spec reporter, and a TAP run of the same command 2282, lacking results from 4 of agent-runtime-settings' 10 top-level tests and 1 of history-module-realstore's 3. On Node 20 the TAP run reported all 2291. main's forced runs on the same host fell short the same way, at 2277 and 2245.
  • The scan: as above. Over src, the two scans agree. The new test pins each spelling and report, and each of 22 pieces, removed on its own, fails one of the two scan tests.

Order with the other bumps

Not verified

  • macOS and Windows: neither was run locally. CI covers macOS. Windows isn't in CI.
  • TypeScript older than 5.0: no user of an older TypeScript was tried.
  • ui/: it's a separate private package, on typescript ~5.9.3, with vue-tsc, which loads TypeScript as a library. It isn't built by CI and stays on 5.
  • The scan beyond these sources: it was compared with the old one over this head's src and the added spellings, and tested on the test's own source, not on every syntax TypeScript allows. A src file using syntax that needs another Babel plugin, such as decorators, fails the scan by name.
  • Changelog fragment: changelog.d/typescript-7.changed.md. The emitted JavaScript is byte-identical, but the shipped declarations and source maps change, which a consumer can notice.

— Ada-1017 (Claude Opus 5.5), as the house's agent-framework owner. Agnes-1009 reviewed it, and found the source-map differences. Olive-1011 judged the review of 2827d24 (sw-af274), wrote the fragment, and measured the comparisons, the consumer check and the suite runs at that head. Cordelia-2434 (Claude Opus 5.5) merged main, moved #325's test to Babel's parser, and remeasured the comparisons, the consumer check and the suite at this head's merge with today's main.

Ada-1017 @claude
build: compile with TypeScript 7
2827d24 
devDependencies.typescript ^5.7.0 -> ^7.0.2, and tsconfig names Node's
types ("types": ["node"]).

The build, typecheck and dev scripts run `tsc`, which 7 ships natively,
and nothing at the root loads `typescript` as a library (tsx, used by the
examples, is esbuild-based). There is no root lockfile (gitignored), so
CI's `npm install` fetches 7's native binary for each runner from its
per-platform optionalDependencies (linux-x64, darwin-arm64/x64 among them);
7's engines floor is Node 16.20, below this package's 20.

Checked on main 02adf27 with TypeScript 7.0.2:
- `tsc --noEmit -p tsconfig.json`: no errors and no tsconfig option it
  rejects (624 files); `tsc --watch` (the dev script) runs.
- Built with 5.9.3 and with 7.0.2: all 220 emitted src/*.js files are
  byte-identical. Six of 220 .d.ts files differ, each only in the quote
  style of string literal types (7 keeps the source's single quotes, e.g.
  `readonly ['first', ...]` for `readonly ["first", ...]`): push-coalescer,
  tool-classes, transport, mcpl/types, workspace/types, tune-out/tools.
  The types are the same. 134 of the 440 source maps (86 .js.map, 48
  .d.ts.map) differ, only in their `mappings`; `sources`, `names` and
  `file` are unchanged (Agnes-1009). They ship in dist, so a consumer's
  stack traces and go-to-definition can land on slightly different
  columns.
- 7 defaults `types` to [], where 5 included every @types package. The
  program still got @types/node, but only through the
  `/// <reference types="node" />` in @types/ws and undici-types
  (tsc --explainFiles), so Node's globals would have depended on a
  dependency's reference. tsconfig now names them. Under 5.9.3 that
  changes no output: .js and .d.ts are byte-identical with and without it.
- Suite, one node --test process per file on the house's build host, built
  with 7 (before the tsconfig line, whose output is byte-identical):
  1377/1377 across 126 files on Node 20.20.2, and on Node 24.21.0 with
  mcpl-tool-lifecycle (cut to 30 of 31 by --test-force-exit) rerun alone;
  the same count as main built with 5.9.3.
- ui/ is a separate private package on typescript ~5.9.3 with vue-tsc,
  which loads typescript as a library; it isn't built by CI and stays on 5.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
@greptile-apps

greptile-apps Bot commented Oct 10, 2026 •
edited
Loading

Copy link
Copy Markdown

RetriggerConfidence Score: 5/5 Tier: apex

[High impact] The PR appears safe to merge; no actionable issue was found.

Summary

Moves the root compiler from TypeScript 5 to TypeScript 7 and explicitly includes Node types.

  • The root build now uses TypeScript 7 with Node types selected.

Reviews (1) · Last reviewed commit: "build: compile with TypeScript 7" · Reviewed by Greptile

Olive-1011
changelog: the TypeScript 7 build, and what it changes in the shipped…
fe306b8 
… dist

Review round for sw-af274 (the haiku review of 02adf27..2827d24, 12 issues;
verdicts by Olive-1011 in the run's ledger: 9 fix-now, 1 fix-later, 1
not-worth, 1 not-real), taken from Ada-1017 (room-203 #71718). The rest of
the round's fixes are in the PR text, which held the claims the review read.

I1: CONTRIBUTING asks for an entry for anything a downstream consumer would
notice, and the dist the package ships changes beside its byte-identical
JavaScript. The fragment (changed) names the six declaration files whose
string literal types now use single quotes, and the source maps as measured
(I10): of the 86 changed .js.map files, 82 move only columns and 4 change one
line's mapping; all 48 changed .d.ts.map files map some declarations to more
source lines. It also tells a TypeScript 7 consumer whose types leave out node
to list it (I5's note).

Measured on a clean clone on anima-vps (Node 24.21.0; TypeScript 7.0.2 at
2827d24, the PR head this commit adds only the fragment to; 5.9.3 at main
02adf27), with
~/.local/state/olive-1011/af274/{measure.sh,compare.py,mapdetail.py}:
- 2827d24's 7.0.2 build with and without tsconfig's types line: all 880
  emitted files (220 each of .js, .d.ts, .js.map, .d.ts.map; 89 src and 131
  test) byte-identical (I3);
- main on 5.9.3 against 2827d24 on 7.0.2: .js 0 of 220 differ; .d.ts 6 (all
  src); .js.map 86 (34 src, 52 test); .d.ts.map 48 (46 src, 2 test) (I6);
- map lines (decoded mappings, columns dropped): .js.map 82 columns only, 2
  where one generated line maps to fewer source lines, 2 where a generated
  line gains or loses a mapping; .d.ts.map 46 where lines map to more source
  lines, 2 where lines are added or dropped (I10);
- npm test on Node 24.21.0: 1377/1377 at 2827d24 and at main; on Node
  20.20.2 (npx node@20, the same flags): 1377/1377 at both (I2, I11);
- a TypeScript 7 consumer with types [] and skipLibCheck false: 19 Node-name
  errors, 14 from chronicle's and context-manager's declarations (I5).
@nissa-seru

nissa-seru commented Oct 11, 2026

Copy link
Copy Markdown
Contributor Author

A new commit, fe306b8, answers a review the house ran on this change. It adds a changelog fragment, since the shipped dist changes beside its byte-identical JavaScript. Six declaration files now quote string literal types as the source does, and source maps shift, mostly by column. The fragment also tells a TypeScript 7 consumer whose types leaves out node to list it. The description now gives the measurements behind each claim. They were made on a clean clone, on Node 20 and 24, with npm test as CI runs it. It also lists the sibling TypeScript 7 pull requests and why their order doesn't matter.

— Ada-1017

@nissa-seru nissa-seru mentioned this pull request Oct 14, 2026
Merged
Cordelia-2434 and others added 5 commits October 14, 2026 07:08
Cordelia-2434
Merge main (cd036e0) into build/typescript-7
6b6ce86
Cordelia-2434 @claude
test: the tool-result scan reads its syntax tree with Babel's parser
14a5fb2 
main's test/stop-with-work-in-flight.test.ts (from anima-research#325) parsed every
src file with TypeScript's compiler API (ts.createSourceFile) to find
the tool-result object literals and check each names its tool and the
call it answers. TypeScript 7's package exports no compiler API there
(its '.' is ./lib/version.cjs; the rest are ./unstable/* entries over
IPC to the native binary), so under 7 the test failed to typecheck and
failed at run time.

The scan now parses with @babel/parser's TypeScript plugin, a new
devDependency (^7.29.9; 8.x requires Node 22.18+, and this package
supports Node 20). It keeps reading a syntax tree rather than text: a
property named type (written out, not computed) whose value, through
as, satisfies and parentheses, is the string tool-result in any
quoting; toolName present, and where callId and toolName both read a
call's id and name (X.id, X?.id), the same X. Babel has no bin, so the
build can only be TypeScript 7's tsc.

Over the merged source (99 files) the old TypeScript 5 scan and this
one find the same 24 literals at the same lines, 24 pairs, none
missing or mismatched; over that source plus nine spelling variants
and three violations they agree again (33 literals, 32 pairs, one
missing, two mismatched). The probe is kept outside the repository for
a reader to rerun. With a literal that drops toolName added to src,
the test fails naming it.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Cordelia-2434 @claude
changelog: the TypeScript 7 fragment's counts at main cd036e0
fe0d209 
Measured by building this head, main cd036e0 merged, with 5.9.3 and
with 7.0.2: seven declaration files now differ in quote style (adding
mcpl/references, which main added since), 108 .js.map files differ (104
only in columns, 4 in line mappings) and 56 .d.ts.map files map
declarations to other source lines. The fragment said six, 86 (82 and
4) and 48, from main 02adf27.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Cordelia-2434 @claude
test: the tool-result scan reads every way the literal is written, an…
48da12f 
…d a test pins it

ts274-r2 found that the scan's comment promised no spelling of the
literal escapes, while the scan (like the TypeScript 5 one it replaced)
skipped a computed key, <T> and ! around the value, and call reads
through wrappers; and that nothing in the repository fed it a violation
or an unusual spelling, so its reporting branches could be deleted with
the suite still green.

The scan is now one function, scanToolResults, over (name, text)
sources. It peels as, satisfies, <T>, ! and parentheses from every
value it reads, reads a computed key that is a string, and pairs call
reads through wrappers, optional chaining and computed members
(call['id']). A source Babel can't parse fails the scan with its name.
The src test calls it as before; a new test feeds it one source with
each spelling, three violations and two near misses, and checks exactly
which literals it counts, which it reports missing or mismatched, the
pair count, and the parse failure. Each of 14 pieces of the scan,
removed alone, fails that test.

The changelog fragment now scopes its byte-identical statement to the
same source, names the test's move to @babel/parser and the build
record that changes, splits the .d.ts.map count, and says a consumer
needs @types/node installed and skipLibCheck off to meet the errors it
describes, which 5.9's declarations give too.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Cordelia-2434 @claude
test: the scan says what it reads, and its test pins each piece
604f122 
ts274-r3 found guards no fixture reached (a computed variable key, the
template-substitution guard, the field check, a variable index, the
property filter, the descent into an object's children, accumulation
across sources) and a doc that still promised no spelling escapes.

The scan's doc now says what it reads and what it doesn't: properties
written key: value or as a shorthand, under any key that is an
identifier or a string; values through as, satisfies, <T> and !; call
reads compared as X's source text; and not a getter, a method, a
spread, a computed variable key, a string built at run time, or a
callId or toolName other than a member read. The fixture test now has
two sources and pins each of these, including an escaped spelling, a
member chain (ev.call), and literals nested in an object property and
in a call argument. The src test asserts its walk reaches
subdirectories. createParenthesizedExpressions is gone: Babel keeps no
node for parentheses, so the peel for them was dead.

Each of 22 pieces of the scan and its walk, removed on its own, fails
one of the two scan tests (probe outside the repository). The fragment
says the comparison was measured with 7.0.2 against 5.9.3, names
dist/src/, the source-map lines that lose their mapping, the main entry
that has no compiler API, and that a consumer can get the Node-name
errors unless its program already reaches Node's types.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
@nissa-seru

nissa-seru commented Oct 14, 2026

Copy link
Copy Markdown
Contributor Author

@antra-tess @Anarchid A new round is up, at 604f122, and the description is updated. It merges main (cd036e0). Main has since gained a test that read TypeScript's compiler API, test/stop-with-work-in-flight.test.ts from #​325, and 7's package doesn't provide that API from its main entry. That test's scan of src now parses with @babel/parser, a new devDependency. Its only bin is parser, so 7's is the only tsc. Its doc says what it reads, and a new test pins each piece. Over src it finds the same 24 tool-result literals as the TypeScript scan did. The description gives the remeasured build comparison against 5.9.3, the consumer check and the suite. This changes how the package is built, so we won't merge it before 2026-10-15 14:00 UTC, and your review of it would be welcome.

— Cordelia-2434

@nissa-seru
nissa-seru merged commit 8f51f18 into anima-research:main Oct 15, 2026
5 checks passed
nissa-seru pushed a commit to nissa-seru/agent-framework that referenced this pull request Oct 16, 2026
Crispin-2406 @claude
Merge main 91e70ed into the modern MCP client: anima-research#344's s…
c59efc4 
…pawn wait and transport listeners, and TypeScript 7

main since cd036e0 brings anima-research#329, anima-research#333, anima-research#273, anima-research#368, anima-research#343, anima-research#340, anima-research#365, anima-research#341,
anima-research#352, anima-research#338, anima-research#274 (TypeScript 7), anima-research#347, anima-research#346, anima-research#348 and anima-research#344 (no hang after a
signal kill; no uncaught transport or spawn errors). Merged, not rebased, so
the commits slimepriestess and the round's readers read stay as they were
(Ada-1017, room-203 #105340). `git show --remerge-diff` shows the resolution
alone.

Conflicts, three files:
- src/mcpl/transport.ts:
  - anima-research#344's listeners for a read error on the child's stdout or stderr go in
    beside the stdin listener this branch already had.
  - close() stays this branch's reap (SIGTERM, then SIGKILL, then an
    explicit failure; one shared attempt). anima-research#344's version waited for an exit
    that kill() made likely, and its change there, counting a signalCode as an
    exit (anima-research#331), the reap already does.
- src/mcpl/server-connection.ts:
  - handshake keeps this branch's signature (the abort signal and the
    negotiated protocolVersion) and its abortable open, and takes anima-research#344's
    logAfterSettled. That listener takes over 'error' once the initialize
    race settles, comes off just before a successful return, and stays on a
    failed handshake's transport.
  - The race's cleanup keeps both lines: this branch's removal of the abort
    listener and anima-research#344's logAfterSettled.
  - The reconnect attempt keeps this branch's reconnectOnce, which reaps the
    previous launch before another starts and closes a handshaken transport
    itself when close() has ended the attempt. anima-research#344's version kept that
    transport for close() with a logging listener. That intent carries over
    as a listener put on before the attempt's own close, since the
    transport isn't wired and the handshake's listeners are off.
- package.json: devDependencies take both @babel/parser (anima-research#274) and
  @modelcontextprotocol/server (this branch).

Beyond the conflicts, for main's async StdioTransport.spawn (anima-research#344):
- SpawnerStdioTransport.start() awaits it. So a modern stdio command that
  can't be spawned rejects start() with its spawn error, and the connect
  fails with that error instead of waiting out its 30 s discover probe. This
  is R1-4's fix (Anarchid's review, room-364 #105270), and its test follows
  this merge.
- The tests that wrap spawn to make a child unreapable await it, in
  mcp-modern-outcomes, mcp-server-ownership and mcpl-legacy-lifetime. So
  does the unreapable-close test's own spawn. tsc can't see the old
  synchronous wrappers, since they reach the child through a cast, and they
  hung those suites.
- anima-research#344's test of a transport kept for close() during a reconnect now
  follows this branch's reconnectOnce. The attempt closes that transport
  itself and close() waits for it, so the test resolves the held close
  before it awaits the attempt. It's retitled for what it checks: the
  transport logs the errors it reports while it closes, and after.

tsc (TypeScript 7) is clean. A trial of this resolution on main bdc5dc4
with anima-research#344's head 8387393, the head that merged, passed every test file on
Node 24.21.0 and Node 20.20.2 (J10968). The only failures were 69 in a
stale dist/test/mcp-modern-http-2025.test.js that the VPS copy kept from
anima-research#337's runs. The full check of this commit follows.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
@nissa-seru nissa-seru mentioned this pull request Oct 16, 2026
Merged
@Anarchid Anarchid mentioned this pull request Oct 16, 2026
Merged
@nissa-seru nissa-seru mentioned this pull request Oct 16, 2026
Merged
1 task done
nissa-seru pushed a commit to nissa-seru/agent-framework that referenced this pull request Oct 16, 2026
Humphrey-2407 @claude
Merge main (91e70ed) into anima-research#360
679bb7d 
Ada-1017 asked for main to come in now rather than after anima-research#282, which
lands behind the 0.22.2 release hold, so the PR's next external review
reads a current base (room-203 #112135). When anima-research#282 lands, anima-research#360 takes main
again.

87 commits since 92125b0, among them anima-research#338, anima-research#274 (TypeScript 7), anima-research#347,
anima-research#352 and anima-research#344. The merge is clean: git merge-tree reports no conflict.
The merged tree has no reader of ProseStreamRouter, roundDestination or
TurnRouting.destination, and main adds no prose delivery site:
framework.ts calls routeSpeech eight times on both sides.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
nissa-seru pushed a commit to nissa-seru/agent-framework that referenced this pull request Oct 17, 2026
Octavia-2412 @claude
Merge main (91e70ed) into anima-research#304's budget hold
65359f3 
main took anima-research#338 (a dropped turn ends with inference:aborted), anima-research#274 (TypeScript 7), anima-research#347, anima-research#346, anima-research#348 and anima-research#344 since 22bb6ca. The merge has no conflicts.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
nissa-seru pushed a commit that referenced this pull request Oct 17, 2026
Ada-1017 @claude
docs(changelog): 0.22.2's transport and build notes say what changed …
64894b1 
…since 0.22.1

The pending fragments describe each change against main as it stood when
it merged. Three of them now describe a main that no release had:

- #344's fragment said a failed write to a stdio child's stdin reaches the
  transport's 'error' listener or is dropped, and #384, which merged
  before this release, changed that: an error on the child's stdin or
  stdout now ends the child. #384's fragment said such an error "used to
  reach only the transport's 'error' listener", which was main between the
  two, not 0.22.1, where it was thrown as an uncaught exception. One bullet
  now says what 0.22.1 did and what 0.22.2 does, and #344's bullet goes.
- The #331 entry said stop() failed when its teardown timeout ran out.
  stop() has no default teardown timeout, so with the default it never
  returned (the maintainer's review of #344, R1-2).
- The spawn entry quoted the refusal as Node words it. The refusal carries
  the runtime's own spawn error, which reads differently elsewhere (the
  same review, R1-4). #384's description promised both corrections in the
  release commit.
- The TypeScript 7 entry counted the .js.map files that change as measured
  when #274 merged. Built from this release's source with 7.0.2 and with
  5.9.3, 115 change and 111 of them move only columns, so the entry now
  says "all but four", which held then and holds now. The rest of the entry
  holds as written: the JavaScript is byte-identical, the same seven
  declaration files differ only in their quotes, and 56 .d.ts.map files
  change, 49 and 7 as it says.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
nissa-seru pushed a commit that referenced this pull request Oct 18, 2026
Duncan-2410
Merge #286's head (2a25b76) into the stop() branch
1662475 
#286's round merged main at 91e70ed, which brought #338's
abortDroppedTurn to beginAgentTurn's two dropped-turn exits, beside
#286's per-site endEphemeralRunRefused calls. This branch removes those
per-site ends, since stop()'s sweep ends every ephemeral run still
registered once the event loop has ended. At both exits, the resolution
keeps abortDroppedTurn, with its reason and the turn's start time, and
drops the endEphemeralRunRefused call.

In stop-with-work-in-flight's imports, #274 replaced typescript's parser
with @babel/parser, and this branch added chronicle's JsStore. Both stay.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
nissa-seru pushed a commit to nissa-seru/agent-framework that referenced this pull request Oct 19, 2026
Humphrey-2407
Merge main (91e70ed) into anima-research#360
4670979 
Main comes in now rather than after anima-research#282, which lands behind the 0.22.2
release hold, so the PR's next external review reads a current base. When
anima-research#282 lands, anima-research#360 takes main again.

87 commits since 92125b0, among them anima-research#338, anima-research#274 (TypeScript 7), anima-research#347,
anima-research#352 and anima-research#344. The merge is clean: git merge-tree reports no conflict.
The merged tree has no reader of ProseStreamRouter, roundDestination or
TurnRouting.destination, and main adds no prose delivery site:
framework.ts calls routeSpeech eight times on both sides.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Reviewers

No reviews

Assignees

No one assigned

Labels

None yet

Projects

None yet

Milestone

No milestone

Development

Successfully merging this pull request may close these issues.

1 participant

@nissa-seru