PR #66067: fix(tui): preserve spaces between thinking fragments

• Repository: openclaw/openclaw
• Author: Rohan5commit
• State: open | merged: False
• Link: https://github.com/openclaw/openclaw/pull/66067

Description (problem / solution / changelog)

Summary

• Problem: streamed thinking fragments were joined with single newlines, and the TUI markdown renderer collapsed those boundaries so adjacent sentences appeared smushed together
• Why it matters: reasoning output in the terminal became hard to read even though the underlying content arrived correctly
• What changed: normalize single-line thinking boundaries into spaces while preserving paragraph breaks, and add formatter/stream-assembler regressions
• What did NOT change (scope boundary): assistant answer text rendering and non-thinking text block joining remain unchanged

Change Type (select all)

• Bug fix
• Feature
• Refactor required for the fix
• Docs
• Security hardening
• Chore/infra

Scope (select all touched areas)

• Gateway / orchestration
• Skills / tool execution
• Auth / tokens
• Memory / storage
• Integrations
• API / contracts
• UI / DX
• CI/CD / infra

Linked Issue/PR

• Closes #66008
• Related #
• This PR fixes a bug or regression

Root Cause (if applicable)

• Root cause: the TUI formatter preserved single newlines between thinking fragments, but the markdown renderer did not render those boundaries as readable spaces
• Missing detection / guardrail: there was no regression test covering multiple thinking blocks rendered in a single streamed assistant update
• Contributing context (if known): reasoning/thinking often arrives in sentence-sized chunks instead of one complete paragraph

Regression Test Plan (if applicable)

• Coverage level that should have caught this:
  • Unit test
  • Seam / integration test
  • End-to-end test
  • Existing coverage already sufficient
• Target test or file: src/tui/tui-formatters.test.ts and src/tui/tui-stream-assembler.test.ts
• Scenario the test should lock in: consecutive thinking fragments stay readable with spaces while explicit paragraph breaks survive
• Why this is the smallest reliable guardrail: it exercises the exact formatter and streamed-assembly surfaces that build terminal assistant output
• Existing test that already covers this (if any): prior tests covered thinking visibility but not fragment boundary spacing
• If no new test is added, why not:

User-visible / Behavior Changes

• Streamed thinking text in the TUI now keeps sentence spacing instead of smushing adjacent fragments together.

Diagram (if applicable)

Before:
[thinking chunk A]\n[thinking chunk B] -> markdown render -> "chunk Achunk B"

After:
[thinking chunk A]\n[thinking chunk B] -> whitespace normalization -> "chunk A chunk B"

Security Impact (required)

• New permissions/capabilities? (No)
• Secrets/tokens handling changed? (No)
• New/changed network calls? (No)
• Command/tool execution surface changed? (No)
• Data access scope changed? (No)
• If any Yes, explain risk + mitigation:

Repro + Verification

Environment

• OS: macOS
• Runtime/container: local pnpm/vitest
• Model/provider: N/A
• Integration/channel (if any): TUI
• Relevant config (redacted): showThinking=true

Steps

1. Render an assistant message containing multiple thinking fragments.
2. Show thinking in the TUI.
3. Inspect the rendered output.

Expected

• Adjacent thinking fragments remain readable with spaces between sentences.

Actual

• Before this fix, sentence boundaries could collapse and appear as one long smushed block.

Evidence

• Failing test/log before + passing after
• Trace/log snippets
• Screenshot/recording
• Perf numbers (if relevant)

Human Verification (required)

• Verified scenarios: targeted vitest run for src/tui/tui-formatters.test.ts and src/tui/tui-stream-assembler.test.ts
• Edge cases checked: consecutive thinking fragments and explicit paragraph breaks
• What you did not verify: full interactive terminal reproduction against a live model stream

Review Conversations

• I replied to or resolved every bot review conversation I addressed in this PR.
• I left unresolved only the conversations that still need reviewer or maintainer judgment.

Compatibility / Migration

• Backward compatible? (Yes)
• Config/env changes? (No)
• Migration needed? (No)
• If yes, exact upgrade steps:

Risks and Mitigations

• Risk: some intentionally line-broken thinking text could become single-paragraph prose
  • Mitigation: explicit blank-line paragraph breaks are preserved, and the fix only touches thinking text in the TUI path

AI-assisted: Yes (Codex). Testing: targeted vitest coverage listed above. I attempted codex review --base origin/main, but the local Codex CLI is currently rate-limited.

Changed files

• src/tui/tui-formatters.test.ts (modified, +22/-1)
• src/tui/tui-formatters.ts (modified, +12/-2)
• src/tui/tui-stream-assembler.test.ts (modified, +14/-0)