feat(cli): enhance prompt handling and synthetic task follow-up by Devesh36 · Pull Request #2797 · Tracer-Cloud/opensre

Devesh36 · 2026-06-11T06:03:18Z

This pull request introduces significant improvements to how the interactive CLI prompt behaves, especially in terms of contextual placeholder text, prompt prefill after synthetic test failures, and overall session state management. The changes enhance user experience by making the prompt more informative and responsive to session events, and by consolidating logic related to synthetic test follow-ups. Additionally, several code paths are refactored for clarity, and new tests are added to ensure correct behavior.

Prompt placeholder and refresh improvements:

Added resolve_prompt_placeholder to generate dynamic placeholder text based on session state (e.g., trust mode, running tasks, resumed session), replacing the previous static placeholder. [1] [2]
Introduced wire_prompt_refresh, allowing the prompt to be programmatically refreshed/redrawn when session state changes. [1] [2]
The prompt now uses a callable for the placeholder, ensuring it always reflects the latest session state.

Synthetic test failure follow-up and session state management:

Moved and refactored logic for suggesting a follow-up prompt and binding the last synthetic observation after a failed synthetic test into new ReplSession methods: suggest_synthetic_failure_follow_up and _bind_last_synthetic_observation. This replaces the previous scattered logic and improves maintainability. [1] [2] [3] [4] [5]
The session now notifies the prompt to refresh whenever a synthetic test fails, ensuring the user sees the suggested follow-up immediately. [1] [2]

Task registry and session enhancements:

Added a running_count method to TaskRegistry to efficiently count currently running tasks, supporting the new dynamic placeholder.
The session tracks a prompt_refresh_fn callback, which is invoked to update the prompt when needed.

Code cleanup and refactoring:

Removed now-unnecessary utility functions (_scenario_id_from_synthetic_suite_name, _try_bind_synthetic_observation) from synthetic_tasks.py and cleaned up related imports and exports. [1] [2] [3]
Improved scenario ID extraction by introducing _scenario_id_from_synthetic_label, which handles both flag and label formats.

Testing improvements:

Added comprehensive tests for the new prompt placeholder logic, covering various session states and combinations.
Updated session tests to cover the new scenario ID extraction and synthetic follow-up logic.

github-actions · 2026-06-11T06:03:28Z

Greptile code review

This repo uses Greptile for automated review. Before merge, aim for Confidence Score: 5/5 with zero unresolved review threads — see CONTRIBUTING.md.

Run a review — add a PR comment with:

@greptile review

Give it ~5-10 minutes (sometimes longer) for results, then fix feedback and re-trigger until you reach Confidence Score: 5/5.

Optional: automate with the greploop skill.

greptile-apps · 2026-06-11T06:18:38Z

Greptile Summary

This PR consolidates synthetic-test failure handling into new ReplSession methods (suggest_synthetic_failure_follow_up, _bind_last_synthetic_observation) and introduces dynamic prompt placeholder text via resolve_prompt_placeholder and wire_prompt_refresh.

suggest_synthetic_failure_follow_up is now the single call-site for queuing the RCA prefill, adding the history-generation guard to background_tasks.py (which previously lacked it) and extracting _scenario_id_from_synthetic_label with proper format validation.
wire_prompt_refresh wires a call_soon_threadsafe callback into the session so background threads can atomically inject prefill text into a live prompt and invalidate its placeholder without touching event-loop state directly.
TaskRegistry.running_count() provides a lock-safe task count for the hot-path placeholder renderer.

Confidence Score: 5/5

Safe to merge — no production bugs introduced by this PR.

The history-generation guard is now consistently applied in both background_tasks.py and synthetic_tasks.py, eliminating the stale-watcher risk that existed before. _scenario_id_from_synthetic_label validates scenario IDs against the same format regex before constructing file paths. The wire_prompt_refresh callback is event-loop-safe via call_soon_threadsafe, and the take_pending_prompt_default / current_buffer.text pairing handles all prompt-lifecycle races correctly.

No files require special attention.

Important Files Changed

Filename	Overview
app/cli/interactive_shell/runtime/session.py	Adds `suggest_synthetic_failure_follow_up`, `_bind_last_synthetic_observation`, `notify_prompt_changed`, and `_scenario_id_from_synthetic_label`. `_SYNTHETIC_SCENARIO_ID_RE` is now defined here and also in `synthetic_tasks.py` (duplication). The blocking observation poll runs in background watcher threads, so it does not stall the event loop.
app/cli/interactive_shell/routing/handle_message_with_agent/orchestration/action_executor/background_tasks.py	Adds `history_gen_when_watch_started` guard and `suggest_follow_up` flag. Non-timeout path still double-joins output threads (pre-existing), but `read_diag` is correctly called after the first join at line 153, so stderr is fully captured before diagnosis.
app/cli/interactive_shell/prompting/prompt_surface.py	Adds `resolve_prompt_placeholder` (dynamic ANSI placeholder) and `wire_prompt_refresh` (event-loop-safe prompt invalidation hook). `_build_prompt_session` now accepts and uses a session for the callable placeholder.
app/cli/interactive_shell/runtime/loop.py	Wires `wire_prompt_refresh`, passes `session` to `_build_prompt_session`, and adds `prefilled`/`placeholder` arguments to each `prompt_async` call. Flow is clean and consistent with the new session methods.
app/cli/interactive_shell/routing/handle_message_with_agent/orchestration/action_executor/synthetic_tasks.py	Removes `_scenario_id_from_synthetic_suite_name` and `_try_bind_synthetic_observation`; delegates to `session.suggest_synthetic_failure_follow_up`. Retains its own copy of `_SYNTHETIC_SCENARIO_ID_RE` (used by `run_synthetic_test`).
app/cli/interactive_shell/runtime/tasks.py	Adds `running_count()` with lock — straightforward and correct for the hot-path placeholder use.
tests/cli/interactive_shell/runtime/test_session.py	New tests cover scenario-ID extraction and follow-up prefill. `test_suggest_synthetic_failure_follow_up_sets_pending` may block up to ~480 ms if `SYNTHETIC_SCENARIOS_DIR` is importable but the observation file does not exist in the test environment.
tests/cli/interactive_shell/orchestration/test_action_executor.py	New `test_start_background_cli_task_skips_follow_up_after_session_reset` correctly verifies the history-generation guard by running the deferred watcher after `session.clear()`.
tests/cli/interactive_shell/prompting/test_prompt_surface.py	Comprehensive tests for all placeholder states and combinations. Clean and correct.

Sequence Diagram

sequenceDiagram
    participant BG as Background Watcher Thread
    participant Session as ReplSession
    participant EventLoop as asyncio Event Loop
    participant PT as prompt_toolkit App

    Note over BG: Synthetic or CLI task exits non-zero
    BG->>Session: suggest_synthetic_failure_follow_up(label)
    Session->>Session: set pending_prompt_default
    Session->>Session: notify_prompt_changed (1st)
    Session-->>EventLoop: call_soon_threadsafe(_apply)
    Session->>Session: _bind_last_synthetic_observation (blocks up to 480ms)
    Session->>Session: notify_prompt_changed (2nd)
    Session-->>EventLoop: call_soon_threadsafe(_apply)
    EventLoop->>PT: _apply - set current_buffer.text
    EventLoop->>PT: pt_app.invalidate
    Note over EventLoop: Next prompt_async iteration
    EventLoop->>Session: take_pending_prompt_default
    Session-->>EventLoop: empty string (consumed by _apply)
    EventLoop->>PT: prompt_async with default and placeholder callable

_{Reviews (3): Last reviewed commit: "fix(cli): join output streams before rea..." | Re-trigger Greptile}

…low-up logic - Updated prompt session initialization to use the session parameter. - Enhanced synthetic task follow-up to skip unnecessary notifications after session resets. - Refined scenario ID extraction logic to ensure valid synthetic scenario IDs. - Added tests to verify new behavior in background CLI task handling and scenario ID extraction.

Devesh36 · 2026-06-11T06:56:50Z

@greptile-apps review again

Mirror synthetic_tasks watcher: drain pump threads before read_diag so failure diagnostics are not read from a still-writing SpooledTemporaryFile. Co-authored-by: Cursor <cursoragent@cursor.com>

Devesh36 · 2026-06-11T07:14:06Z

@greptile-apps review again

Davidson3556 · 2026-06-11T08:11:00Z

looks good to me

github-actions · 2026-06-12T06:23:11Z

🌊 Merged. @Devesh36 is now permanently woven into git history. No take-backs. 😄

👋 Join us on Discord - OpenSRE : hang out, contribute, or hunt for features and issues. Everyone's welcome.

muddlebee · 2026-06-12T11:25:04Z

@Devesh36 did you check with /sessions also? these changes are being captured while switching btw sessions? also tasks are handled?

Devesh36 · 2026-06-13T04:33:51Z

@Devesh36 did you check with /sessions also? these changes are being captured while switching btw sessions? also tasks are handled?

Yeah tested /sessions /new and /resume with background synthetic everything works. No prefill leak on session switch tasks/placeholder behave correctly and failed synthetic prefills the prompt when you stay on the same session.

feat(cli): enhance prompt handling and synthetic task follow-up

5ae369c

greptile-apps Bot reviewed Jun 11, 2026

View reviewed changes

fix(cli): join output streams before reading stderr in background tasks

763f4db

Mirror synthetic_tasks watcher: drain pump threads before read_diag so failure diagnostics are not read from a still-writing SpooledTemporaryFile. Co-authored-by: Cursor <cursoragent@cursor.com>

Devesh36 merged commit 250f2fa into Tracer-Cloud:main Jun 12, 2026
14 checks passed

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

feat(cli): enhance prompt handling and synthetic task follow-up#2797

feat(cli): enhance prompt handling and synthetic task follow-up#2797
Devesh36 merged 3 commits into
Tracer-Cloud:mainfrom
Devesh36:improvements

Devesh36 commented Jun 11, 2026

Uh oh!

github-actions Bot commented Jun 11, 2026

Uh oh!

greptile-apps Bot commented Jun 11, 2026 •

edited

Loading

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Devesh36 commented Jun 11, 2026

Uh oh!

Devesh36 commented Jun 11, 2026

Uh oh!

Davidson3556 commented Jun 11, 2026

Uh oh!

Uh oh!

github-actions Bot commented Jun 12, 2026

Uh oh!

muddlebee commented Jun 12, 2026

Uh oh!

Devesh36 commented Jun 13, 2026

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

3 participants

Conversation

Devesh36 commented Jun 11, 2026

Uh oh!

github-actions Bot commented Jun 11, 2026

Greptile code review

Uh oh!

greptile-apps Bot commented Jun 11, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Greptile Summary

Confidence Score: 5/5

Important Files Changed

Sequence Diagram

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Devesh36 commented Jun 11, 2026

Uh oh!

Devesh36 commented Jun 11, 2026

Uh oh!

Davidson3556 commented Jun 11, 2026

Uh oh!

Uh oh!

github-actions Bot commented Jun 12, 2026

Uh oh!

muddlebee commented Jun 12, 2026

Uh oh!

Devesh36 commented Jun 13, 2026

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

3 participants

greptile-apps Bot commented Jun 11, 2026 •

edited

Loading