[lanyue-llk]

Purpose

Extend #640 with browser-client protocol, rendering and emulation support. Each target retains its own effective environment; protocol commands use the renderer and paint owners already responsible for the output.

Overall design

Full-size overview · All current PR diagrams

Implementation flow (Mermaid)

flowchart TD
  C["CDP / WebDriver clients"] --> P["Protocol validation and session routing"]
  P --> T["Live page or window target"]
  T --> E["Per-target emulation and style state"]
  T --> I["Inspector overlay, input and capture"]
  T --> N["Download and navigation lifecycle"]
  E --> S["Stylo environment and cascade"]
  S --> R["Renderer output"]
  I --> R
  R --> V["Paint: bounded vision processing"]
  V --> O["Browser-client response"]

Loading

Scope and ownership

Area	Behavior
Protocol	Complete schema, Inspector, input, capture and navigation contracts through typed dispatch.
Window bounds	Validate live windows; partial updates retain omitted dimensions.
Emulation	Apply settings to each target's effective style environment.
Stylo	Resolve preferred text scale as a CSS environment input and recompute dependent styles.
Paint	Process blurred vision in bounded tiles, including 4K captures.

Inspector geometry uses main's native node references. Worker fixtures retain the renderer stack budget.

Dependencies

main → #640 → #731 → #853. This PR targets #640; head fb2f6b397.

Stylo #1, pinned at 0db9840e674f6ad45c4df06961501b1a26bd2827, supplies preferred text scale and preserves the WebMCP selector APIs required by current main.

Reproduce affected cases

Local fixtures and expected behavior

Each case below is an existing regression test at this PR's pinned head. The fixtures create their own pages and local services; Rainbond, WebArena credentials and model calls are unnecessary. Follow the repository's Rust toolchain and install cargo-nextest before running. Initial compilation is still required; the filter runs only the listed cases.

Regression case	Trigger and expected result
schema_discovery_lists_unique_dispatchable_domains	Schema discovery returns unique domains that the typed dispatcher supports.
browser_window_bounds_readback_preserves_unspecified_dimensions	Set only part of a window rectangle; readback preserves omitted dimensions.
scroll_gesture_uses_hit_target_and_honors_wheel_cancellation	Scroll a nested target; wheel reaches that target, and preventDefault suppresses further default scrolling.
inspector_highlight_rect_is_visible_and_hide_restores_original_pixels	Draw a highlight rectangle and hide it; screenshot pixels change and then restore without adding DOM nodes.
inspector_highlight_tracks_live_node_geometry_and_rejects_stale_ids	Highlight a live node, change its geometry, then use a stale ID; rendering follows live geometry and rejects the stale reference.
os_text_scale_updates_css_environment_and_preserves_media_preferences	Set scale 1.5, 2, then clear it: env-based font/width become 24/30px, 32/40px, 16/20px; a fixed font stays 16px.
context_media_keeps_each_page_text_scale_across_reload	Reload pages with distinct effective scale; a media update preserves each page's own scale.
vision_deficiency_changes_capture_pixels_without_mutating_dom	Apply a visual filter; capture pixels change, computed background remains red, and clearing restores the original image.
blurred_vision_accepts_4k_capture	Apply blurred vision to a 4K raster within the supported paint budget.
websocket_cdp_plain_anchor_attachment_download_keeps_document	Click an attachment link; complete the download while preserving the current document.

git clone https://github.com/lexmount/moli.git moli-review-731
cd moli-review-731
git checkout --detach fb2f6b397f359f2e7b98c59dce76b0a10fa03449
cargo nextest run -p moli-protocol -p moli-paint -p moli-protocol-server \
  -E 'test(schema_discovery_lists_unique_dispatchable_domains) | test(browser_window_bounds_readback_preserves_unspecified_dimensions) | test(scroll_gesture_uses_hit_target_and_honors_wheel_cancellation) | test(inspector_highlight_rect_is_visible_and_hide_restores_original_pixels) | test(inspector_highlight_tracks_live_node_geometry_and_rejects_stale_ids) | test(os_text_scale_updates_css_environment_and_preserves_media_preferences) | test(context_media_keeps_each_page_text_scale_across_reload) | test(vision_deficiency_changes_capture_pixels_without_mutating_dom) | test(blurred_vision_accepts_4k_capture) | test(websocket_cdp_plain_anchor_attachment_download_keeps_document)'

Expected result: all selected tests pass. To run one case, replace the expression after -E with test(the_case_name). Test failures print the expected/actual assertion. These are regression case identifiers, with inputs and assertions linked to the exact source version.

Links to the original client badcases

These recorded scenario IDs came from the 12-case browser-client cohort supporting #731. They are client scenarios, independent of the 812-task WebArena dataset.

Original scenario ID	Focused regression above
pw_raw_schema_getdomains	schema_discovery_lists_unique_dispatchable_domains
v2_t2_browser_window	browser_window_bounds_readback_preserves_unspecified_dimensions
v2_t2_misc_input_scroll	scroll_gesture_uses_hit_target_and_honors_wheel_cancellation
v4_cdp_dom_highlightrect / v4_cdp_dom_hidehighlight	inspector_highlight_rect_is_visible_and_hide_restores_original_pixels
v4_cdp_dom_highlightnode	inspector_highlight_tracks_live_node_geometry_and_rejects_stale_ids
v4_cdp_emulation_setemulatedostextscale	os_text_scale_updates_css_environment_and_preserves_media_preferences
v4_cdp_emulation_setemulatedvisiondeficiency	vision_deficiency_changes_capture_pixels_without_mutating_dom
sc_dl_download_checksum__pw	websocket_cdp_plain_anchor_attachment_download_keeps_document

The vision regression above asserts grayscale pixel behavior; the original client case used protanopia. The download regression asserts attachment/navigation behavior; the original Playwright case additionally checked downloaded bytes. Both are useful checks at their stated scope.

Evaluation reports

• WebArena-Verified: 812-task trajectory replay (frozen report edition): Chromium 153 and the recorded Moli fix(browser): complete remaining protocol and rendering contracts #731 + fix(window): preserve native named-context identity #853 stack each accepted all 812 tasks. All six methods use Playwright; this is fixed trajectory replay with no model calls.
• The replay report preserves the earlier measured binaries and accepted correction ledger. It is supporting benchmark evidence for this work, not a new 812-task run of the October 8 rebased commits.
• Current-head verification is the CI report linked below. The previously completed replay cohorts were not rerun.

Current-head validation

CI report for #731: all 20 checks passed, including Nextest, protocol smoke tests, the 1,020-case frontend differential, 545-page WebMainBench, navigation soak and release comparisons.

The cumulative #640 + #731 + #853 local tree passed formatting, all-feature Clippy and 19,809 tests, with 14 ignored. This is cumulative coverage; the linked CI verifies this PR's own head.

[lanyue-llk]

已对照 #640 的审查关注点检查协议边界、默认行为与生命周期，并核对本 PR 的改动及 CI。下面两处会影响新增 Inspector 接口的实际客户端使用；其余输入、下载和绘制路径未发现足以评论的确定性回归。

[lanyue-llk]

复查最新提交：上轮两条 Inspector 评论对应的单会话行为已修复。Rust 结构整体符合仓库既有风格；以下两处是可定位的功能边界，另有一处新增状态没有生产用途。

[lanyue-llk]

复查最新提交的跨目标状态、分块模糊、Inspector、输入及下载链路。上一轮三项修复在当前代码和新增测试覆盖的场景下成立；下面这处是浏览器范围状态的后续生命周期缺口。

[github-actions]

Sequential Navigation Soak A/B

✅ HEAD completed the 200-navigation resilience and memory observation.

One browser process, one target, and one CDP session navigate CSDN → SegmentFault → Huaban → example.com repeatedly. This run issued 200 Page.navigate commands per binary.

Common ancestor 2e05488f8c → HEAD fb2f6b397f; benchmark order: head-first; workflow: success.

Session resilience

Metric	Base	HEAD
Attempted / planned	200 / 200	200 / 200
Direct observable pass	199 / 200	199 / 200
Failures / recovered	1 / 1	1 / 1
Recovery failures	0	0
Lifecycle order violations	0	0
Network order violations	0	0
Wall time	299.9 s	280.1 s

A public-page failure is reported separately from an unrecoverable session. The soak requires all navigation attempts, zero failed recovery, zero lifecycle/network ordering violations, and complete resource evidence.

Process-tree memory

Metric	Base	HEAD	Δ
Boundary samples	200	200	0
Boundary RSS samples	200	200	0
Boundary PSS samples	200	200	0
Periodic peak RSS	218.0 MiB	212.5 MiB	-5.5 MiB
Periodic peak PSS	214.4 MiB	209.2 MiB	-5.2 MiB
RSS · first 10 avg	139.9 MiB	173.1 MiB	+33.2 MiB
RSS · last 10 avg	174.2 MiB	172.6 MiB	-1.6 MiB
RSS · last − first window	+34.3 MiB	-0.5 MiB	-34.8 MiB
RSS · warm slope / 100 nav	+3.5 MiB	+2.5 MiB	-1.1 MiB
PSS · first 10 avg	136.3 MiB	169.8 MiB	+33.5 MiB
PSS · last 10 avg	170.7 MiB	169.0 MiB	-1.7 MiB
PSS · last − first window	+34.4 MiB	-0.8 MiB	-35.2 MiB
PSS · warm slope / 100 nav	+3.7 MiB	+2.4 MiB	-1.2 MiB
Peak file descriptors	61	65	4
Peak threads	31	31	0

HEAD memory by 50-navigation quarter

HEAD range	Avg RSS	Final RSS	Peak RSS	Avg PSS	Final PSS	Peak FD
1–50	167.0 MiB	154.0 MiB	212.4 MiB	163.6 MiB	150.5 MiB	54
51–100	170.0 MiB	158.9 MiB	202.8 MiB	166.4 MiB	155.4 MiB	54
101–150	166.6 MiB	159.3 MiB	203.1 MiB	163.0 MiB	155.7 MiB	52
151–200	170.6 MiB	162.6 MiB	196.8 MiB	167.1 MiB	159.0 MiB	53

Raw reports and per-navigation boundary samples: workflow run and full sequential-navigation-soak-results artifact.

Public-site timing and memory are observational. Use the A/B deltas and trend shape as evidence, not as a deterministic performance threshold.

[github-actions]

Spider Bench A/B

✅ All benchmark browser service runs completed; results are informational.

Public HEAD: 184 / 240 rows; 38 / 48 sites produced rows; 10 unexpected empty sites.

✅ Deterministic fixture contract is clean.

Common ancestor 2e05488f8ce2 → HEAD fb2f6b397f35; benchmark order: head-first; workflow: success.

Public 48-site run · informational

Metric	Base	HEAD	Δ
Run status	success	success	—
Extracted rows / contract	184 / 240	184 / 240	0 (0.00%)
Contract fill	76.67%	76.67%	0.00 pp (0.00%)
Sites with rows	38 / 48	38 / 48	0 (0.00%)
Sites meeting row contract	36 / 48	36 / 48	0 (0.00%)
Unexpected empty sites	10	10	0 (0.00%)
Partial-row sites	2	2	0 (0.00%)
Site outcomes	extracted: 37; HTTP error + rows: 1; empty: 7; HTTP error + empty: 1; snapshot error + empty: 2	extracted: 36; HTTP error + rows: 2; empty: 7; HTTP error + empty: 1; snapshot error + empty: 2	—
Observed duration	45.08 s	48.24 s	+3.16 s (+7.01%)
Estimated CPU	43.93 s	46.45 s	+2.52 s (+5.75%)
Peak PSS	275.91 MiB	254.11 MiB	-21.80 MiB (-7.90%)
Peak RSS	279.91 MiB	257.92 MiB	-21.98 MiB (-7.85%)
Peak threads	27	26	-1 (-3.70%)
Mismatched sites	0	0	0
Timeout with items	0	0	0

Category rows	Base	HEAD	Δ
news	35 / 40	35 / 40	0
stocks	20 / 40	20 / 40	0
tech	35 / 40	35 / 40	0
sports	27 / 40	27 / 40	0
games	37 / 40	37 / 40	0
life	30 / 40	30 / 40	0

Public-site content, timing, and memory are noisy. This single A/B run reports evidence only; site outcome counts explain missing rows without treating them as a deterministic regression.

CPU and memory timelines

These are bounded 40-point views of the same complete process-tree samples used by the HTML report. CPU uses 100% per occupied logical core; memory lines are RSS first and PSS second when complete PSS samples are available.

Base

xychart-beta
    title "Base CPU"
    x-axis "Elapsed seconds" [1.035, 1.536, 2.537, 3.539, 4.039, 4.54, 5.039, 6.539, 7.541, 9.043, 10.044, 11.546, 12.547, 14.05, 15.052, 16.554, 17.556, 19.057, 20.057, 21.557, 22.558, 24.061, 25.062, 26.563, 27.564, 29.066, 30.067, 31.569, 32.57, 34.072, 35.072, 36.574, 37.576, 39.078, 40.079, 41.581, 42.58, 44.083]
    y-axis "CPU percent" 0 --> 200
    line [27.98, 123.8, 55.89, 133.76, 121.84, 127.99, 162.02, 148.1, 173.56, 185.64, 145.92, 139.83, 157.86, 123.7, 2, 69.98, 77.97, 0, 74, 79.91, 143.82, 135.62, 111.96, 123.77, 145.95, 149.9, 107.91, 37.95, 43.95, 113.7, 8, 137.94, 109.76, 17.96, 95.9, 167.82, 26, 117.84]

Loading

xychart-beta
    title "Base memory: RSS then PSS"
    x-axis "Elapsed seconds" [0.035, 1.035, 1.536, 2.537, 3.539, 4.039, 4.54, 5.039, 6.539, 7.541, 9.043, 10.044, 11.546, 12.547, 14.05, 15.052, 16.554, 17.556, 19.057, 20.057, 21.557, 22.558, 24.061, 25.062, 26.563, 27.564, 29.066, 30.067, 31.569, 32.57, 34.072, 35.072, 36.574, 37.576, 39.078, 40.079, 41.581, 42.58, 44.083, 45.082]
    y-axis "Memory MiB" 0 --> 500
    line [12.15, 89.14, 111.35, 119.04, 116.37, 132.19, 141.34, 157.78, 178.04, 159.48, 178.16, 175, 175.39, 154.63, 162.78, 144.2, 145.59, 149.19, 139.62, 147.79, 168.85, 169.7, 194.12, 279.91, 223.29, 160.37, 172.46, 162.4, 154.27, 153.52, 163.75, 155.23, 228.59, 193.66, 161.87, 177.56, 202.5, 157.43, 201.89, 0]
    line [8.89, 85.51, 107.64, 115.37, 112.66, 128.5, 137.64, 154.14, 174.92, 155.54, 174.44, 171.25, 171.64, 150.93, 159.01, 140.43, 141.83, 145.42, 135.85, 144.02, 165.08, 163.93, 190.38, 275.91, 219.64, 156.71, 167.94, 158.91, 150.49, 149.74, 159.98, 151.46, 224.56, 190.09, 158.09, 173.78, 199.94, 153.74, 198.13, 0]

Loading

HEAD

xychart-beta
    title "HEAD CPU"
    x-axis "Elapsed seconds" [1.036, 1.536, 2.538, 4.041, 5.544, 6.545, 8.046, 9.549, 11.05, 12.05, 13.552, 15.054, 16.056, 17.559, 19.062, 20.064, 20.565, 21.567, 23.069, 24.57, 25.573, 27.075, 27.576, 28.578, 29.579, 31.079, 32.58, 33.582, 35.083, 36.585, 37.586, 39.088, 40.589, 42.09, 43.09, 44.593, 46.095, 47.095]
    y-axis "CPU percent" 0 --> 200
    line [15.96, 121.92, 49.9, 125.79, 179.62, 149.86, 79.92, 189.77, 53.97, 87.99, 139.84, 89.96, 29.93, 73.86, 5.99, 3.99, 0, 0, 55.89, 69.95, 117.8, 119.93, 99.94, 133.76, 151.84, 151.99, 97.93, 95.75, 151.82, 69.92, 13.98, 87.94, 20.02, 7.99, 83.93, 115.84, 67.84, 112.03]

Loading

xychart-beta
    title "HEAD memory: RSS then PSS"
    x-axis "Elapsed seconds" [0.035, 1.036, 1.536, 2.538, 4.041, 5.544, 6.545, 8.046, 9.549, 11.05, 12.05, 13.552, 15.054, 16.056, 17.559, 19.062, 20.064, 20.565, 21.567, 23.069, 24.57, 25.573, 27.075, 27.576, 28.578, 29.579, 31.079, 32.58, 33.582, 35.083, 36.585, 37.586, 39.088, 40.589, 42.09, 43.09, 44.593, 46.095, 47.095, 48.242]
    y-axis "Memory MiB" 0 --> 500
    line [12.21, 76.14, 120.22, 122.82, 139.9, 164.73, 179.71, 169, 190.77, 164.31, 188.24, 193.91, 164.56, 146.95, 165.18, 149.15, 148.13, 148.13, 147.12, 155.35, 165.25, 167.58, 188.35, 257.92, 199.82, 205.39, 230.24, 145.65, 170.96, 179.2, 172.03, 167.59, 180.01, 164.78, 151.34, 166.1, 185.8, 152.58, 185.09, 0]
    line [8.89, 72.51, 116.5, 119.11, 136.12, 161.01, 175.88, 165.66, 187.11, 160.48, 184.41, 190.38, 160.71, 143.11, 161.34, 145.31, 144.29, 144.29, 143.28, 151.5, 161.41, 163.82, 184.51, 254.11, 195.98, 201.54, 226.59, 141.82, 167.29, 175.36, 168.2, 163.74, 177, 160.94, 147.49, 161.93, 183.48, 148.75, 181.26, 0]

Loading

Deterministic fixture · required contract

The fixture exercises 15 routes: 8 are expected to emit rows and 7 intentionally exercise empty, loading, or timeout behavior.

Metric	Base	HEAD	Δ
Run status	success	success	—
Extracted rows / contract	40 / 40	40 / 40	0 (0.00%)
Contract fill	100.00%	100.00%	0.00 pp (0.00%)
Sites with rows	8 / 15	8 / 15	0 (0.00%)
Sites meeting row contract	15 / 15	15 / 15	0 (0.00%)
Unexpected empty sites	0	0	0
Partial-row sites	0	0	0
Site outcomes	extracted: 7; HTTP error + rows: 1; empty: 3; timeout + empty: 4	extracted: 7; HTTP error + rows: 1; empty: 3; timeout + empty: 4	—
Observed duration	19.86 s	19.85 s	-0.01 s (-0.06%)
Estimated CPU	0.39 s	0.38 s	-0.01 s (-2.80%)
Peak PSS	65.28 MiB	65.66 MiB	+0.38 MiB (+0.58%)
Peak RSS	68.91 MiB	69.21 MiB	+0.30 MiB (+0.44%)
Peak threads	18	18	0 (0.00%)
Mismatched sites	0	0	0
Timeout with items	0	0	0

Full HTML, JSON, CSV, logs, and page snapshots: workflow run and full spider-bench-results artifact.

The public-web A/B is informational. The exact HEAD fixture contract runs as a separate required CI check; timing and public-site content deltas remain non-blocking.

[github-actions]

CI Regression Report

Source CI run · source state at render: in_progress · artifacts: 6/6

Check	Status	Signal
Release regression	✅	HEAD/base failures 0/0; raw binary +0.123316%
Frontend differential	✅	1,020/1,020 cases matched; 0 issues
Agent episodes	✅	8/8 Moli episodes passed; 0 failures
Runtime/CDP contracts	✅	26 contract cases; 0 failures
CDP smoke	✅	53/53 groups passed; 622 scenarios
WebMainBench	✅	545/545 completed; 0 unexpected failures

Release regression — ✅ HEAD/base failures 0/0; raw binary +0.123316%

Package and image size

Metric	Base	HEAD	Delta	Delta %
Raw binary	148.30 MiB	148.48 MiB	+187.27 KiB	+0.123316%
Stripped binary	102.04 MiB	102.21 MiB	+172.00 KiB	+0.164612%
gzip binary	50.87 MiB	50.94 MiB	+72.33 KiB	+0.138835%
Rootfs	155.21 MiB	155.38 MiB	+180.00 KiB	+0.113257%
gzip rootfs	59.73 MiB	59.81 MiB	+82.47 KiB	+0.134836%

Startup latency and PSS

Case	Base p50	HEAD p50	Delta	Base PSS	HEAD PSS	Delta
Serve ready	27.18 ms	27.21 ms	+0.02 ms	4.89 MiB	3.56 MiB	−1.33 MiB
CDP first page	48.56 ms	48.44 ms	−0.12 ms	4.32 MiB	3.42 MiB	−919.00 KiB
CDP warm page	11.18 ms	11.35 ms	+0.17 ms	50.01 MiB	48.85 MiB	−1.16 MiB
CLI about:blank	14.27 ms	13.84 ms	−0.43 ms	4.64 MiB	3.54 MiB	−1.09 MiB
CLI local JS	17.93 ms	17.21 ms	−0.71 ms	2.18 MiB	4.38 MiB	+2.20 MiB

Concurrency matrix

Concurrency	Base p50	HEAD p50	Delta	Base PSS	HEAD PSS	Delta	HEAD failures	Unstable B/H
1	16.45 ms	16.45 ms	−0.00 ms	720.00 KiB	768.00 KiB	+48.00 KiB	0	0/0
5	20.10 ms	20.12 ms	+0.02 ms	22.70 MiB	12.86 MiB	−9.84 MiB	0	0/0
10	19.26 ms	20.08 ms	+0.82 ms	22.25 MiB	12.79 MiB	−9.46 MiB	0	0/0
25	19.29 ms	20.18 ms	+0.89 ms	12.48 MiB	27.84 MiB	+15.36 MiB	0	0/0
100	19.49 ms	19.85 ms	+0.36 ms	12.62 MiB	39.93 MiB	+27.31 MiB	0	0/0

Frontend differential — ✅ 1,020/1,020 cases matched; 0 issues

Match	DOM mismatch	Diagnostic mismatch	Moli error	Reference error	Infrastructure error	Mismatched frames	Duration
1,020	0	0	0	0	0	0	6m 30.96s

Agent episodes · Moli vs Chromium — ✅ 8/8 Moli episodes passed; 0 failures

Fast contract	Episodes	Passed	Assertions	Ready p50	Episode p50
Moli	8	8	47/47	5.95 ms	30.72 ms
Chromium	8	8	47/47	278.83 ms	120.84 ms

Operation p50	Chromium	Moli	Delta	Delta %
navigate	53.98 ms	7.92 ms	−46.06 ms	−85.33%
observe	2.07 ms	1.58 ms	−0.48 ms	−23.34%
fill	5.71 ms	4.31 ms	−1.40 ms	−24.50%
click	9.00 ms	4.93 ms	−4.07 ms	−45.26%

Idle-resource episode	Peak RSS	Peak PSS	Average CPU	Peak CPU	Peak processes	Sampler
Moli	65.52 MiB	62.08 MiB	0.65%	8.06%	1	✅
Chromium	1292.09 MiB	338.97 MiB	2.20%	88.14%	13	⚠️ PSS partial

Runtime and CDP session contracts — ✅ 26 contract cases; 0 failures

Suite	Cases	Failures	Median case p50	Median case p95	Median peak PSS p50
Synthetic fetch	13	0	16.67 ms	17.76 ms	1016.00 KiB
Long-lived CDP session	13	0	18.09 ms	18.64 ms	—

CDP smoke — ✅ 53/53 groups passed; 622 scenarios

Workers: 4 · cumulative group time: 194.57 s · failed groups: 0

Slowest group	Scenarios	Duration	Status
scrollbar-visibility	22	30.58 s	✅
target-lifecycle	24	17.10 s	✅
dom-input	27	12.85 s	✅
chromium-cdp	56	11.15 s	✅
multi-page	45	10.82 s	✅

WebMainBench · 545 pages — ✅ 545/545 completed; 0 unexpected failures

Completed	Successful	Expected DNS failure	Unexpected failures	Panics	Crashes	Timeouts	Empty output
545/545	544	1	0	0	0	0	0

Frozen HTML · Linux · --wait done · 45 seconds per page · no retries. Content quality is evaluated separately.

Diagnostics: webmainbench-results — every page's Markdown, stderr, and exit status. Source run and artifacts

All artifact fields are parsed by the trusted default-branch renderer; missing or invalid inputs remain visible as unavailable.

[lanyue-llk]

依赖 stylo pr 1 的合入