feat(kiro_cli): add full TUI mode support with --legacy-ui fallback (#159) by haofeif · Pull Request #163 · awslabs/cli-agent-orchestrator

haofeif · 2026-04-08T05:25:41Z

Title

feat(kiro_cli): add full TUI mode support with --legacy-ui fallback (#159)

Summary

Remove hardcoded --legacy-ui from the Kiro CLI launch command — the provider now launches in TUI mode by default and auto-detects the output format
Add ▸ Credits: completion marker detection for TUI mode (TUI_CREDITS_PATTERN) alongside existing green arrow detection for legacy mode
Add separator-based message extraction (_extract_tui_message()) as fallback when no green arrows are found in pure TUI output
Add --legacy-ui retry fallback in initialize() — if TUI init times out, the provider sends /exit and retries with --legacy-ui automatically
Add "aren't available" to e2e REFUSAL_KEYWORDS to fix a Claude Code test false positive

What changed

`src/cli_agent_orchestrator/providers/kiro_cli.py`

Change	Details
New patterns	`TUI_SEPARATOR_PATTERN = r"^[─]{4,}$"`, `TUI_CREDITS_PATTERN = r"▸\sCredits:\s[\d.]+"`
`initialize()`	TUI-first launch → `/exit` + `--legacy-ui` fallback on timeout
`get_status()`	Credits + idle prompt after it → `COMPLETED`; Credits without idle → `PROCESSING`
`_extract_tui_message()`	New method: finds last separator before `▸ Credits:`, skips user message paragraph, extracts agent response
`extract_last_message_from_script()`	Falls back to `_extract_tui_message()` when no green arrows found

Status detection flow (updated)

No output           → ERROR
No idle prompt      → PROCESSING
Error indicators    → ERROR
Permission prompt   → WAITING_USER_ANSWER
Green arrow + idle  → COMPLETED  (legacy path)
Credits + idle      → COMPLETED  (TUI path)
Credits, no idle    → PROCESSING (TUI still rendering)
Idle only           → IDLE

`test/providers/test_kiro_cli_unit.py`

Updated test_initialize_success — asserts kiro-cli chat --agent developer (no --legacy-ui)
Updated test_initialize_kiro_cli_timeout — expects "both TUI and --legacy-ui failed"
Added test_initialize_legacy_ui_fallback — verifies TUI fail → /exit → --legacy-ui retry
Added TestKiroCliTuiMode class (11 tests):
- Idle, completed, processing detection from TUI fixtures
- Message extraction (basic + complex with code blocks)
- Error cases (no Credits, no separator)
- Credits without idle = PROCESSING
- Status-extraction synchronization guarantee

`test/providers/fixtures/` (4 new files)

kiro_cli_tui_idle_output.txt — separator + header + idle prompt
kiro_cli_tui_completed_output.txt — user msg + agent response + Credits + idle
kiro_cli_tui_processing_output.txt — no idle prompt visible
kiro_cli_tui_complex_response.txt — code blocks, multiple paragraphs + Credits

`docs/kiro-cli.md`

Updated launch command (no --legacy-ui by default)
Documented dual-path extraction (green arrows first, TUI separator/Credits fallback)
Documented auto-detection and backward compatibility with --legacy-ui wrappers

`test/e2e/test_allowed_tools.py`

Added "aren't available" to REFUSAL_KEYWORDS (fixes Claude Code test_reviewer_cannot_write false positive)

Test plan

Unit tests: 69/69 passed (including 12 new tests)
Full non-e2e suite: 1106/1106 passed
Kiro CLI e2e: 11/11 passed in TUI mode (11m19s)
Verify --legacy-ui fallback triggers correctly when TUI is unavailable (e.g., older kiro-cli version)

Closes #159

…159) Remove hardcoded --legacy-ui from launch command and add TUI-native status detection and message extraction. The provider now: - Launches in TUI mode by default, falls back to --legacy-ui on timeout - Detects COMPLETED via ▸ Credits: marker + idle prompt (TUI path) - Extracts messages using separator (────) boundaries when no green arrows - Retains full backward compatibility with legacy UI patterns Also adds "aren't available" to e2e REFUSAL_KEYWORDS for Claude Code test fix. Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>

jwalaQ · 2026-04-08T09:37:29Z

+             Ask a question or describe a task
+
+        Strategy:
+            1. Find the last Credits line (response end marker)


Could this be simplified by detecting the ghost text "Kiro is working" present in the user prompt space? Unless waiting for a tool permission, the "Kiro is working" text is always present when the agent is not idle. When a permission request appears the text changes to Yes, No and Always Allow for this session.

Ghost text

Tool permission

Implemented — added TUI_PROCESSING_PATTERN = r'Kiro is working' as a positive PROCESSING signal (checked first in get_status()), and TUI_PERMISSION_PATTERN = r'Yes\s+No\s+Always [Aa]llow' for TUI permission prompt detection that triggers WAITING_USER_ANSWER. Both are in the latest commits.

jwalaQ · 2026-04-08T09:42:01Z

            return TerminalStatus.PROCESSING

+        # Check 5: TUI completion — Credits marker + idle prompt after it.
+        # In pure TUI mode, there are no green arrows. Completion is indicated


Kiro in TUI mode allows queuing messages while the agent is working. The prompt remains in the queue until the current loop completes. It is similar to the current behaviour. Maybe exact checks for agent completion can be relaxed in the future and leverage the in-built functionality.

@jwalaQ Acknowledged — thanks for the insight on native queuing. CAO currently requires a strict send → wait for COMPLETED → extract response cycle across all providers (including Claude Code, which also supports native queuing). This is because CAO needs to extract each response to return it via the API, route it between agents (handoffs/assigns), and store conversation history. Relaxing completion checks would require a new CAO-level API contract (e.g., fire-and-forget with async response collection), which is a broader architectural change beyond this PR. Noted as a potential future optimization.

Also cao at the moment do not support "queuing message" for any providers, including Claude code, which by itself supports queuing messages long ago.

- Raise separator minimum from 4 to 20 chars to avoid matching short markdown separators in agent output (jwalaQ comment #4) - Remove redundant ANSI cleanup in _extract_tui_message — input is already ANSI-stripped by caller (jwalaQ comment #5) - Improve timeout error message wording (jwalaQ comment #3) Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>

- Add "Kiro is working" ghost text as positive PROCESSING signal, checked before idle prompt absence (jwalaQ comment #1) - Add TUI permission pattern "Yes No Always Allow" alongside legacy [y/n/t] format, requires all three options to avoid false positives on bare "Yes"/"No" in agent output (jwalaQ comment #2) Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>

…output Verified against real kiro-cli v1.29.1 TUI output via tmux capture-pane: - Idle prompt is "Ask a question or describe a task" (capital A, no comma) - Pattern now accepts both old (lowercase, comma) and new formats - Updated fixtures to use real TUI output format - Updated inline test strings to match v1.29+ output Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>

haofeif · 2026-04-09T01:49:34Z

@jwalaQ thanks very much for you looking into this, can you please review again, i believe your comments are addressed.

…utput separators (#159) Changed _extract_tui_message() to find the first separator after the previous turn's Credits line instead of the last separator before the current Credits. This prevents false matches when agent output contains box-drawing separator characters. Also updated docs for launch command. Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>

jwalaQ · 2026-04-10T08:00:49Z

+────────────────────────────────────────────────────
+  What files are in this directory?
+
+  Analyzing the directory structure...


Add the "Kiro is working" ghost text here as well?

Added @jwalaQ thanks for your review

…#159) Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>

jwalaQ · 2026-04-10T10:14:52Z

Thanks for this PR!

haofeif requested a review from a team April 8, 2026 05:25

style: apply black formatting to kiro_cli provider and tests

0d23224

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>

haofeif added the enhancement New feature or request label Apr 8, 2026

jwalaQ reviewed Apr 8, 2026

View reviewed changes

haofeif and others added 3 commits April 9, 2026 10:01

haofeif requested a review from jwalaQ April 9, 2026 01:48

haofeif and others added 2 commits April 9, 2026 18:11

Merge branch 'main' into fix/kiro-tui-mode

5a3c8d4

jwalaQ approved these changes Apr 10, 2026

View reviewed changes

fix(test): add 'Kiro is working' ghost text to TUI processing fixture (…

5e92070

…#159) Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>

haofeif merged commit a9fdfdf into main Apr 10, 2026
4 checks passed

haofeif deleted the fix/kiro-tui-mode branch April 10, 2026 08:40

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

feat(kiro_cli): add full TUI mode support with --legacy-ui fallback (#159)#163

feat(kiro_cli): add full TUI mode support with --legacy-ui fallback (#159)#163
haofeif merged 8 commits into
mainfrom
fix/kiro-tui-mode

haofeif commented Apr 8, 2026

Uh oh!

jwalaQ Apr 8, 2026

Uh oh!

haofeif Apr 9, 2026

Uh oh!

jwalaQ Apr 8, 2026

Uh oh!

haofeif Apr 9, 2026

Uh oh!

Uh oh!

Uh oh!

Uh oh!

haofeif commented Apr 9, 2026

Uh oh!

jwalaQ Apr 10, 2026

Uh oh!

haofeif Apr 10, 2026

Uh oh!

Uh oh!

jwalaQ commented Apr 10, 2026

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

2 participants

Conversation

haofeif commented Apr 8, 2026

Title

Summary

What changed

src/cli_agent_orchestrator/providers/kiro_cli.py

Status detection flow (updated)

test/providers/test_kiro_cli_unit.py

test/providers/fixtures/ (4 new files)

docs/kiro-cli.md

test/e2e/test_allowed_tools.py

Test plan

Uh oh!

jwalaQ Apr 8, 2026

Choose a reason for hiding this comment

Ghost text

Tool permission

Uh oh!

haofeif Apr 9, 2026

Choose a reason for hiding this comment

Uh oh!

jwalaQ Apr 8, 2026

Choose a reason for hiding this comment

Uh oh!

haofeif Apr 9, 2026

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Uh oh!

Uh oh!

haofeif commented Apr 9, 2026

Uh oh!

jwalaQ Apr 10, 2026

Choose a reason for hiding this comment

Uh oh!

haofeif Apr 10, 2026

Choose a reason for hiding this comment

Uh oh!

Uh oh!

jwalaQ commented Apr 10, 2026

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

2 participants

`src/cli_agent_orchestrator/providers/kiro_cli.py`

`test/providers/test_kiro_cli_unit.py`

`test/providers/fixtures/` (4 new files)

`docs/kiro-cli.md`

`test/e2e/test_allowed_tools.py`