spike: cc-chat as a Scala.js custom element — core split, ADR 0002, measured bundles #60

Merged

mph merged 8 commits from spike/cc-chat-scalajs into main 2026-08-12 14:47:55 +00:00

Conversation 0 Commits 8 Files changed 37 +709 -16

mph commented 2026-08-12 14:25:36 +00:00

Owner

Copy link

Summary

Validates the agent-chat UI direction decided with ADR 0002: the durable asset is a versioned session-over-HTTP contract, and its reference client is a self-contained Scala.js custom element (<cc-chat>) sharing the SDK's model and fold logic through a cross-built core. This branch carries the enabling core split, the ADR, the spike that proved the approach, and the bundle-size CI gate.

What lands

• core split for cross-building: platform-free sources (model, parsing, CLIError, conversation-log model/parser) move to core/shared/src, compiled by both the JVM core and the new core.js Scala.js module (claude-code-query-core_sjs1_3). JVM-bound code stays in core/src.
• Timestamps on Moment (released iw-support-time:0.2.0, resolved from the forge registry): removes scala-java-time at the root — java.time is unreachable from cross-built sources, cutting ~158 KB gzip from browser bundles. ConversationLogEntry.timestamp: Option[Moment], parsed with Moment.parseIsoOption (timestamps stay best-effort on vendor JSONL).
• ADR 0002: the contract-first design, Scala.js decision and its accepted costs, asset-first delivery from the jar, Laminar adoption (27 KB delta, under the agreed 40 KB rule), deferred decisions with triggers.
• CI bundle-size gate: 250 KB gzip, just above the measured 217.9 KB — tight enough that a scala-java-time-class regression breaks the build instead of eating the 400 KB design ceiling silently.
• Spike code (uispike/*, throwaway by declaration): bare/rawdom/laminar element variants and a toy host proving hashed classpath-resource delivery, history + SSE wiring with shared circe decoders, and correct fold semantics (verified in Chrome, incl. a real 1,128-line transcript → 122 rows).

Measured (fullLinkJS, gzip; ceiling 400 KB)

target	size
bare element	7.9 KB
+ shared model, circe, parsers	190 KB
+ Laminar	217 KB

Not in scope

The real ui module (transcript viewer + Tapir endpoints, built test-first) is the next slice per ADR 0002 and supersedes uispike/*. The spike deliberately skipped TDD, as recorded in the ADR.

🤖 Generated with Claude Code

## Summary Validates the agent-chat UI direction decided with ADR 0002: the durable asset is a versioned session-over-HTTP contract, and its reference client is a self-contained Scala.js custom element (`<cc-chat>`) sharing the SDK's model and fold logic through a cross-built core. This branch carries the enabling core split, the ADR, the spike that proved the approach, and the bundle-size CI gate. ## What lands - **core split for cross-building**: platform-free sources (model, parsing, `CLIError`, conversation-log model/parser) move to `core/shared/src`, compiled by both the JVM `core` and the new `core.js` Scala.js module (`claude-code-query-core_sjs1_3`). JVM-bound code stays in `core/src`. - **Timestamps on `Moment`** (released `iw-support-time:0.2.0`, resolved from the forge registry): removes `scala-java-time` at the root — `java.time` is unreachable from cross-built sources, cutting ~158 KB gzip from browser bundles. `ConversationLogEntry.timestamp: Option[Moment]`, parsed with `Moment.parseIsoOption` (timestamps stay best-effort on vendor JSONL). - **ADR 0002**: the contract-first design, Scala.js decision and its accepted costs, asset-first delivery from the jar, Laminar adoption (27 KB delta, under the agreed 40 KB rule), deferred decisions with triggers. - **CI bundle-size gate**: 250 KB gzip, just above the measured 217.9 KB — tight enough that a scala-java-time-class regression breaks the build instead of eating the 400 KB design ceiling silently. - **Spike code (`uispike/*`, throwaway by declaration)**: bare/rawdom/laminar element variants and a toy host proving hashed classpath-resource delivery, history + SSE wiring with shared circe decoders, and correct fold semantics (verified in Chrome, incl. a real 1,128-line transcript → 122 rows). ## Measured (fullLinkJS, gzip; ceiling 400 KB) | target | size | |---|---| | bare element | 7.9 KB | | + shared model, circe, parsers | 190 KB | | + Laminar | 217 KB | ## Not in scope The real `ui` module (transcript viewer + Tapir endpoints, built test-first) is the next slice per ADR 0002 and supersedes `uispike/*`. The spike deliberately skipped TDD, as recorded in the ADR. 🤖 Generated with [Claude Code](https://claude.com/claude-code)

mph added 8 commits 2026-08-12 14:25:36 +00:00

core: split platform-free sources into shared/src and cross-build core.js ... 6d5ebd35b9

The message model, parsers, CLIError, and the conversation log entry
model/parser have no JVM dependency. They move to core/shared/src,
compiled by both the JVM core module and a new core.js Scala.js module,
so browser consumers can share the wire model and codecs with the
server. JVM-bound code (os-lib, file IO) stays in core/src.

Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>

spike: cc-chat custom element in Scala.js, three measured link targets ... db8027fb54

bare (element floor) 7.9 KB gzip; rawdom (+ shared model, circe, both
parsers) 347 KB; laminar 375 KB — Laminar delta 27 KB. scala-java-time,
dragged in by Instant in ConversationLogEntry, accounts for ~178 KB:
serving the browser normalized history instead of raw vendor JSONL
would nearly halve the bundle.

Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>

spike: toy host proving hashed classpath-resource delivery and the wire contract ... f24a49a59e

Serves the linked element bundle from the jar-shaped resource layout
under a content-hashed immutable path, a history page (real transcript
via CC_CHAT_TRANSCRIPT or a bundled sample), and canned SSE frames
exercising the fold: tool status, assistant reply, non-duplicating
result, and a fallback-only result. Verified in Chrome: correct
rendering on canned and real data, clean console, element survives DOM
moves without re-wiring.

Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>

docs: ADR 0002 — session-over-HTTP contract with a Scala.js chat element ... 8b0424a491

Records the two design sessions and the spike results: the versioned
contract as the durable asset, the self-contained cc-chat element as
its reference client, Scala.js sharing the core model, Laminar (27 KB
delta), asset-first delivery from the jar, and the normalized-history
decision that removes scala-java-time (~178 KB) from the browser.

Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>

core: represent log timestamps as Moment, dropping scala-java-time from JS ... b20b21e255

ConversationLogEntry.timestamp is works.iterative.core.Moment (from the
dependency-free iw-support-time module) instead of java.time.Instant,
and the parser goes through Moment.parseIso. java.time is no longer
reachable from the cross-built sources at all, so the creep risk is
gone, not avoided: the JS classpath has no scala-java-time to pull in.
Spike bundles drop ~158 KB gzip: rawdom 347→190 KB, laminar 375→217 KB.

BREAKING: consumers reading ConversationLogEntry.timestamp get a Moment;
on the JVM, works.iterative.core.MomentInterop.toInstant converts.

Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>

ci,docs: bundle-size gate at 250 KB; ADR 0002 amended for the Moment re-basing ... 62be631089

The gate sits just above the measured 217 KB rather than at the 400 KB
design ceiling, so a scala-java-time-class regression (~158 KB) breaks
the build instead of silently consuming headroom. ADR records the root
fix, the re-measured numbers, the iw-support-time dependency and its
release sequencing, and moves the normalized-history question to open.

Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>

chore(core): adopt Moment.parseIsoOption for optional timestamp parsing ... 866912bee6

iw-support-time's parseIso now returns Either[Moment.ParseError, Moment];
the log parser keeps Option semantics via parseIsoOption. Laminar bundle:
217,852 B gzip (gate 256,000).

Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>

build: depend on released iw-support-time 0.2.0 from the forge registry ... a381de915d

Adds the iterative-works Forgejo Maven registry to every module that
reaches iw-support-time (directly or through core.js), so CI resolves
the release without a local ivy copy.

Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>

mph force-pushed spike/cc-chat-scalajs from a381de915d to e86bc729b6 2026-08-12 14:32:33 +00:00 Compare

mph merged commit e1d69b38c9 into main 2026-08-12 14:47:55 +00:00

mph deleted branch spike/cc-chat-scalajs 2026-08-12 14:47:55 +00:00

mph referenced this pull request from a commit 2026-08-12 14:47:57 +00:00

Merge pull request 'spike: cc-chat as a Scala.js custom element — core split, ADR 0002, measured bundles' (#60) from spike/cc-chat-scalajs into main

mph referenced this pull request from a commit 2026-08-13 13:03:25 +00:00

workflow(sl): correct cc-chat worklog — PR #60 already merged

Sign in to join this conversation.

Reviewers

No reviewers

Labels

Clear labels   

bug

Something isn't working

  

documentation

Improvements or additions to documentation

  

duplicate

This issue or pull request already exists

  

enhancement

New feature or request

  

good first issue

Good for newcomers

  

help wanted

Extra attention is needed

  

invalid

This doesn't seem right

  

question

Further information is requested

  

wontfix

This will not be worked on

No labels

bug

documentation

duplicate

enhancement

good first issue

help wanted

invalid

question

wontfix

Milestone

Clear milestone

No items

No milestone

Projects

Clear projects

No items

No project

Assignees

Clear assignees

No assignees

1 participant

Notifications

Subscribe

Due date

The due date is invalid or out of range. Please use the format "yyyy-mm-dd".

No due date set.

Dependencies

No dependencies set

Reference

iterative-works/claude-code-query!60

No description provided.