Decompose the mega files: lib.rs (18.7k), config.rs (12.3k), client.rs (11.1k), runtime_threads.rs (9.3k)

[Hmbown]

User ask for the 0.9.12 cleanup lane: the 10k+-line files keep causing pain. Measured today (prod code, excluding the 20k-line test files which should split alongside their subjects):

• crates/tui/src/lib.rs — 18,747
• crates/tui/src/config.rs — 12,346
• crates/tui/src/client.rs — 11,122
• crates/tui/src/runtime_threads.rs — 9,322
• crates/tui/src/core/engine.rs — 6,891, core/engine/turn_loop.rs — 6,668, runtime_api.rs — 6,542, tui/ui/event_loop.rs — 5,721 (event_loop already tracked in Streamline TUI monoliths: split event_loop.rs match into per-domain reducers, slim app.rs #5580)

Approach (same discipline as #5580): mechanical module extraction, one cohesive region per commit, no behavior change, moves verified by cargo check + the file's existing test slice. Suggested first cuts: client.rs wire-parsing (parse_usage/SSE chunk parsing + their tests) into client/wire.rs; config.rs merge/persistence into config/merge.rs; lib.rs engine-config assembly into its own module. Declared one-way per AGENTS.md once started: new code lands in the extracted modules, touched regions migrate.

Tracker: #5573. Related: #5580, #5581.

[Hmbown]

Release-blocking workflow decomposition slice B is now on PR #5576 in be3a201.

This is a file-boundary-only extraction: the existing nested tools::workflow::journal module moved from workflow/mod.rs into workflow/journal.rs with the same module path, visibility, journal format, imports, and tests. No signatures or behavior changed.

Current integration-head evidence before commit:

• cargo fmt --all -- --check: clean
• cargo check --workspace: clean
• cargo clippy --workspace --all-targets: clean, zero warnings
• focused codewhale-tui workflow nextest slice: 187/187 passed
• staged diff check: clean

Slice C, the existing workflow usage types and summation helper, is now in progress. Slice D remains a hard pre-release blocker after C.

Written and posted by ChatGPT 5.6 Sol while operating the maintainer-approved v0.9.12 release lane.

[Hmbown]

Slice D (report module extraction) is now complete on the integration branch.

Landed: 64545aeb9 on codex/v0912-integration-20260823 (refactor(workflow): extract report module), closing out the B/C/D workflow extraction set: B journal be3a201f2, C usage a4a788cd3, D report 64545aeb9.

Move shape (semantic-free): 208-line report block removed from crates/tui/src/tools/workflow/mod.rs with mod report; plus three imports added; crates/tui/src/tools/workflow/report.rs adds 210 lines. Total 212 insertions / 206 deletions; normalized moved-content SHA-256 9a9e2e66f666a87ec2dcb8c03db191854962ff9771abc8234914b58055aa84d8; no renames, signature changes, or opportunistic cleanup.

Evidence:

• Isolated slice-D worktree (worktrees/v0912-workflow-report-slice-d, base a4a788cd3): fmt green; cargo check --workspace green; clippy --workspace --all-targets green; workspace test(workflow) wall 218/218 green (13,206 skipped).
• Current-head integration checkout with D staged (now committed): fmt green; workspace check green; workspace/all-target clippy green; TUI workflow slice 187/187 green (10,910 skipped) at the merged head, after clearing the macOS com.apple.provenance loader xattr from the freshly linked test binary (the earlier discovery stall was environmental, not a product failure).

Slice B evidence was posted earlier: #5586 (comment)

Remaining for #5586: the broader mega-file decomposition checklist (lib.rs 18.7k, config.rs 12.3k, client.rs 11.1k, runtime_threads.rs 9.3k) is still open; only the workflow-module extraction set (1a rename + B/C/D) is complete.

Comment written and posted by ChatGPT 5.6 Sol on Hunter's behalf.