Skip to content

Open-source Python library for AI browser agents, plus Browser Use Cloud: stealth browsers and a hosted agent API.

Reconstruct Rust terminal command waiting results

Magnus Müller832181c39e0f
diff --git a/RUST_AGENT_INTEGRATION_FEATURES.md b/RUST_AGENT_INTEGRATION_FEATURES.md
index 956bf42bd..a5d24a780 100644
--- a/RUST_AGENT_INTEGRATION_FEATURES.md
+++ b/RUST_AGENT_INTEGRATION_FEATURES.md
@@ -764,10 +764,15 @@ Terminal core branch: `magnus/browser-use-rust-main-integration` at terminal mai
- This preserves current-main unified exec failures on Python-visible `history.errors()` and per-action `ActionResult.error` while retaining the final command output in action memory.
- Proof: `test_rust_history_surfaces_terminal_exec_command_end_failure`.
+151. Rust terminal command-waiting reconstruction
+ - Terminal `command.waiting` events now reconstruct live unified-exec process results when no generic textual `tool.output` is present.
+ - This preserves current-main long-running command state on Python-visible `ActionResult.extracted_content`, `long_term_memory`, and `action_history()` instead of falling back to synthetic tool-completion text.
+ - Proof: `test_rust_history_reconstructs_terminal_command_waiting_result`.
+
## Current Verification
- `python3 -m py_compile browser_use/agent/service.py browser_use/rust/service.py browser_use/rust/__init__.py browser_use/__init__.py tests/ci/test_rust_agent.py examples/rust_agent/basic.py examples/rust_agent/real_v8_smoke.py`
-- `uv run pytest -q tests/ci/test_rust_agent.py` (122 tests)
+- `uv run pytest -q tests/ci/test_rust_agent.py` (123 tests)
- `cargo build -q -p browser-use-cli` on terminal branch `magnus/browser-use-rust-main-integration`
- `cargo test -q -p browser-use-cli run_codex_session_command_accepts_task_id_and_model -- --nocapture`
- `CARGO_INCREMENTAL=0 cargo test -q -p browser-use-agent selected_remote_cdp_mode_allows_remote_cdp_connect -- --nocapture`
diff --git a/browser_use/rust/service.py b/browser_use/rust/service.py
index 2e6b96718..e110d53a8 100644
--- a/browser_use/rust/service.py
+++ b/browser_use/rust/service.py
@@ -1003,6 +1003,33 @@ def _exec_command_end_failure_message(payload: dict[str, Any]) -> str | None:
return f'{tool_name} failed: {detail}'
+def _command_waiting_text(payload: dict[str, Any]) -> str:
+ session_id = payload.get('session_id') or payload.get('process_id')
+ if session_id is not None and str(session_id):
+ return f'Process running with session ID {session_id}'
+ return 'Process running'
+
+
+def _command_waiting_payload(payload: dict[str, Any], previous_payload: dict[str, Any] | None) -> dict[str, Any]:
+ waiting_text = _command_waiting_text(payload)
+ previous_text = _tool_result_text(previous_payload) if previous_payload is not None else None
+ if previous_text and waiting_text not in previous_text:
+ waiting_text = f'{previous_text}\n\n{waiting_text}'
+ next_payload = dict(payload)
+ next_payload['text'] = waiting_text
+ return next_payload
+
+
+_TEXTUAL_TOOL_RESULT_EVENTS = (
+ 'tool.output',
+ 'tool.output_delta',
+ 'command.waiting',
+ 'exec_command.end',
+ 'tool.finished',
+ 'model.response.input_item',
+)
+
+
def _tool_result_key(payload: dict[str, Any]) -> tuple[str, str] | None:
call_id = payload.get('tool_call_id') or payload.get('call_id')
if call_id:
@@ -1137,6 +1164,7 @@ def _tool_results_by_call_id(events: list[dict[str, Any]]) -> dict[str, tuple[st
if event_type not in (
'tool.output',
'tool.output_delta',
+ 'command.waiting',
'exec_command.end',
'tool.failed',
'tool.aborted',
@@ -1168,6 +1196,17 @@ def _tool_results_by_call_id(events: list[dict[str, Any]]) -> dict[str, tuple[st
merged_payload['text'] = f'{previous_text if isinstance(previous_text, str) else ""}{delta_text}'
results[key] = (event_type, merged_payload)
continue
+ if event_type == 'command.waiting':
+ if previous is not None and previous[0] in (
+ 'tool.output',
+ 'tool.failed',
+ 'tool.aborted',
+ 'exec_command.end',
+ 'model.response.input_item',
+ ):
+ continue
+ results[key] = (event_type, _command_waiting_payload(payload, previous[1] if previous is not None else None))
+ continue
if event_type == 'exec_command.end' and previous is not None and previous[0] in (
'tool.output',
'tool.failed',
@@ -1190,6 +1229,7 @@ def _unkeyed_tool_results(events: list[dict[str, Any]]) -> list[tuple[str, dict[
if event_type not in (
'tool.output',
'tool.output_delta',
+ 'command.waiting',
'exec_command.end',
'tool.failed',
'tool.aborted',
@@ -1240,6 +1280,25 @@ def _unkeyed_tool_results(events: list[dict[str, Any]]) -> list[tuple[str, dict[
if previous_delta_index is not None and payload.get('stream') is not True:
results[previous_delta_index] = (event_type, payload)
continue
+ elif event_type == 'command.waiting':
+ name = payload.get('name')
+ previous_result_index = next(
+ (
+ index
+ for index in range(len(results) - 1, -1, -1)
+ if isinstance(name, str)
+ and name
+ and results[index][1].get('name') == name
+ ),
+ None,
+ )
+ if previous_result_index is not None:
+ previous_event_type, previous_payload = results[previous_result_index]
+ if previous_event_type in ('tool.output', 'tool.failed', 'tool.aborted', 'exec_command.end', 'model.response.input_item'):
+ continue
+ results[previous_result_index] = (event_type, _command_waiting_payload(payload, previous_payload))
+ continue
+ payload = _command_waiting_payload(payload, None)
elif event_type == 'exec_command.end':
if not _tool_result_text(payload):
continue
@@ -1551,12 +1610,8 @@ def _action_results_from_tool_calls(
action_results.append(
ActionResult(
error=error,
- extracted_content=text
- if event_type in ('tool.output', 'tool.output_delta', 'exec_command.end', 'tool.finished', 'model.response.input_item')
- else None,
- long_term_memory=text
- if event_type in ('tool.output', 'tool.output_delta', 'exec_command.end', 'tool.finished', 'model.response.input_item')
- else None,
+ extracted_content=text if event_type in _TEXTUAL_TOOL_RESULT_EVENTS else None,
+ long_term_memory=text if event_type in _TEXTUAL_TOOL_RESULT_EVENTS else None,
)
)
diff --git a/tests/ci/test_rust_agent.py b/tests/ci/test_rust_agent.py
index 1d9d9affd..0804e550f 100644
--- a/tests/ci/test_rust_agent.py
+++ b/tests/ci/test_rust_agent.py
@@ -751,6 +751,59 @@ def test_rust_history_surfaces_terminal_exec_command_end_failure():
assert history.action_results()[-1].error == error
+def test_rust_history_reconstructs_terminal_command_waiting_result():
+ from browser_use.rust.service import _history_from_events
+
+ history = _history_from_events(
+ [
+ {
+ 'type': 'tool.started',
+ 'payload': {
+ 'name': 'exec_command',
+ 'tool_call_id': 'call-exec',
+ 'arguments': {'cmd': 'sleep 60 && echo done', 'yield_time_ms': 250},
+ },
+ },
+ {
+ 'type': 'tool.output_delta',
+ 'payload': {
+ 'name': 'exec_command',
+ 'tool_call_id': 'call-exec',
+ 'process_id': '1000',
+ 'session_id': 1000,
+ 'stream': 'stdout',
+ 'text': 'started\n',
+ },
+ },
+ {
+ 'type': 'command.waiting',
+ 'payload': {
+ 'name': 'exec_command',
+ 'tool_call_id': 'call-exec',
+ 'process_id': '1000',
+ 'session_id': 1000,
+ },
+ },
+ {'event_type': 'session.done', 'payload': {'result': 'command is still running'}},
+ ],
+ model='gpt-test',
+ started=1.0,
+ finished=2.0,
+ output_model_schema=None,
+ process_error=None,
+ )
+
+ first_result = history.action_results()[0]
+ waiting_result = 'started\n\nProcess running with session ID 1000'
+
+ assert history.action_names() == ['exec_command', 'done']
+ assert first_result.extracted_content == waiting_result
+ assert first_result.long_term_memory == waiting_result
+ assert first_result.error is None
+ assert history.action_history()[0][0]['result'] == waiting_result
+ assert history.final_result() == 'command is still running'
+
+
def test_rust_history_reconstructs_terminal_unkeyed_tool_results():
from browser_use.rust.service import _history_from_events