Skip to content

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

Populate Rust Laminar spans with messages

Magnus Müller920af58bfab2
diff --git a/RUST_AGENT_INTEGRATION_FEATURES.md b/RUST_AGENT_INTEGRATION_FEATURES.md
index 14467fe35..d66e2665f 100644
--- a/RUST_AGENT_INTEGRATION_FEATURES.md
+++ b/RUST_AGENT_INTEGRATION_FEATURES.md
@@ -1412,8 +1412,16 @@ Terminal core branch: `magnus/browser-use-rust-main-integration` at terminal mai
- This makes main Rust-core LLM calls visible in eval traces alongside the existing executor/evaluation/judge spans and sub-agent calls.
- Proof: `.venv/bin/python -m pytest -q tests/ci/test_rust_agent.py -k 'laminar_run_summary_populates_current_span or terminal_usage_prices_token_count_events or reconstructs_terminal_token_count_usage or reconstructs_terminal_nested_model_usage'` and `.venv/bin/python -m py_compile browser_use/rust/service.py tests/ci/test_rust_agent.py`.
+271. Rust terminal Laminar LLM message payloads
+ - Rust terminal `model.turn.request` events now include a bounded, provider-neutral `llm_input` payload with system blocks and model-visible messages from the assembled per-turn request.
+ - The payload redacts common secret fields and inline media data, caps message/text volume, and preserves roles/content/tool-call structure so the Python Laminar bridge can show real LLM inputs instead of only tool-name summaries.
+ - The Python `browser_use.rust.Agent` Laminar replay now passes those messages as the `LLM` span input and adds GenAI-style `gen_ai.input.messages`, `gen_ai.output.messages`, `gen_ai.system_instructions`, and `gen_ai.usage.*` attributes.
+ - Proof: terminal `cargo test -p browser-use-agent turn_request_event_carries_sanitized_llm_input_messages -- --nocapture`, browser-use `.venv/bin/python -m pytest -q tests/ci/test_rust_agent.py -k 'laminar_run_summary_populates_current_span'`, and browser-use `.venv/bin/python -m py_compile browser_use/rust/service.py tests/ci/test_rust_agent.py`.
+
## Current Verification
+- terminal `cargo test -p browser-use-agent turn_request_event_carries_sanitized_llm_input_messages -- --nocapture`
+- `.venv/bin/python -m pytest -q tests/ci/test_rust_agent.py -k 'laminar_run_summary_populates_current_span'`
- `.venv/bin/python -m pytest -q tests/ci/test_rust_agent.py -k 'laminar_run_summary_populates_current_span or terminal_usage_prices_token_count_events or reconstructs_terminal_token_count_usage or reconstructs_terminal_nested_model_usage'`
- `.venv/bin/python -m py_compile browser_use/rust/service.py tests/ci/test_rust_agent.py`
- `.venv/bin/python -m pytest -q tests/ci/test_rust_agent.py -k 'terminal_usage_prices_token_count_events or reconstructs_terminal_token_count_usage or reconstructs_terminal_nested_model_usage'`
diff --git a/browser_use/rust/service.py b/browser_use/rust/service.py
index d3f4ca684..fa43174a9 100644
--- a/browser_use/rust/service.py
+++ b/browser_use/rust/service.py
@@ -2508,6 +2508,15 @@ def _terminal_laminar_turn_input(
for tool in tools[:30]:
if isinstance(tool, dict) and isinstance(tool.get('name'), str):
tool_names.append(tool['name'])
+ llm_input = request_payload.get('llm_input')
+ if not isinstance(llm_input, dict):
+ llm_input = {}
+ messages = llm_input.get('messages')
+ if not isinstance(messages, list):
+ messages = []
+ system = llm_input.get('system')
+ if not isinstance(system, list):
+ system = []
return {
'model': request_payload.get('model') or default_model,
'provider': request_payload.get('provider') or default_provider,
@@ -2516,6 +2525,11 @@ def _terminal_laminar_turn_input(
'system_prompt_tokens': composition.get('system_prompt_tokens'),
'tools_count': len(tools),
'tool_names': tool_names,
+ 'system': system,
+ 'messages': messages,
+ 'message_count': _int_value(llm_input.get('message_count')),
+ 'omitted_earlier_messages': _int_value(llm_input.get('omitted_earlier_messages')),
+ 'truncated': bool(llm_input.get('truncated')),
}
@@ -2527,15 +2541,59 @@ def _terminal_laminar_turn_output(events: list[dict[str, Any]], raw_usage: dict[
response_item_extractor=_response_output_item_reasoning_text,
)
tool_calls = _tool_started_calls(events)
+ assistant_preview = _laminar_preview(assistant_text, limit=2000)
+ thinking_preview = _laminar_preview(thinking_text, limit=1200)
+ output_messages = []
+ if assistant_preview:
+ output_messages.append({'role': 'assistant', 'content': [{'type': 'text', 'text': assistant_preview}]})
+ if tool_calls:
+ output_messages.append(
+ {
+ 'role': 'assistant',
+ 'tool_calls': [
+ {
+ 'name': call.get('name'),
+ 'id': call.get('tool_call_id'),
+ 'arguments': call.get('arguments'),
+ }
+ for call in tool_calls[:20]
+ ],
+ }
+ )
return {
- 'assistant_output_preview': _laminar_preview(assistant_text, limit=2000),
- 'thinking_preview': _laminar_preview(thinking_text, limit=1200),
+ 'messages': output_messages,
+ 'assistant_output_preview': assistant_preview,
+ 'thinking_preview': thinking_preview,
'tool_calls_count': len(tool_calls),
'tool_call_names': [call['name'] for call in tool_calls[:20] if isinstance(call.get('name'), str)],
'usage': _terminal_laminar_usage_summary(raw_usage),
}
+def _terminal_laminar_gen_ai_attributes(
+ span_input: dict[str, Any],
+ span_output: dict[str, Any],
+ usage: dict[str, int | float] | None,
+) -> dict[str, Any]:
+ attributes: dict[str, Any] = {
+ 'gen_ai.operation.name': 'chat',
+ 'gen_ai.request.model': str(span_input.get('model') or ''),
+ 'gen_ai.system': str(span_input.get('provider') or ''),
+ 'gen_ai.input.messages': _laminar_preview(span_input.get('messages') or [], limit=60_000),
+ 'gen_ai.system_instructions': _laminar_preview(span_input.get('system') or [], limit=30_000),
+ 'gen_ai.output.messages': _laminar_preview(span_output.get('messages') or [], limit=40_000),
+ }
+ if usage:
+ attributes.update(
+ {
+ 'gen_ai.usage.input_tokens': usage.get('input_tokens'),
+ 'gen_ai.usage.output_tokens': usage.get('output_tokens'),
+ 'gen_ai.usage.total_tokens': usage.get('total_tokens'),
+ }
+ )
+ return attributes
+
+
def _record_laminar_terminal_llm_spans(
events: list[dict[str, Any]],
*,
@@ -2563,6 +2621,7 @@ def _record_laminar_terminal_llm_spans(
start_time = _event_time_seconds(turn_events[0], 0.0)
end_time = _event_time_seconds(turn_events[-1], start_time)
with _laminar_start_span('rust_core.llm', input=span_input, span_type='LLM'):
+ span_output = _terminal_laminar_turn_output(turn_events, raw_usage)
_laminar_set_span_attributes(
{
'runtime': 'browser_use.rust',
@@ -2578,7 +2637,8 @@ def _record_laminar_terminal_llm_spans(
'duration_seconds': max(0.0, end_time - start_time),
}
)
- _laminar_set_span_output(_terminal_laminar_turn_output(turn_events, raw_usage))
+ _laminar_set_span_attributes(_terminal_laminar_gen_ai_attributes(span_input, span_output, usage))
+ _laminar_set_span_output(span_output)
if len(ranges) > max_spans:
_laminar_event(
'rust_core.llm_spans_truncated',
diff --git a/tests/ci/test_rust_agent.py b/tests/ci/test_rust_agent.py
index 1a1887b6d..0b4dd96d8 100644
--- a/tests/ci/test_rust_agent.py
+++ b/tests/ci/test_rust_agent.py
@@ -7305,6 +7305,17 @@ def test_rust_agent_laminar_run_summary_populates_current_span(monkeypatch):
'system_prompt_tokens': 100,
'tools': [{'name': 'browser'}, {'name': 'done'}],
},
+ 'llm_input': {
+ 'system': [{'text': 'System prompt'}],
+ 'messages': [
+ {
+ 'role': 'user',
+ 'content': [{'type': 'text', 'text': 'Find the title on example.com'}],
+ }
+ ],
+ 'message_count': 1,
+ 'omitted_earlier_messages': 0,
+ },
},
},
{'event_type': 'model.stream_delta', 'ts_ms': 1500, 'payload': {'text': 'laminar answer'}},
@@ -7350,10 +7361,21 @@ def test_rust_agent_laminar_run_summary_populates_current_span(monkeypatch):
assert FakeLaminar.spans[0]['name'] == 'rust_core.llm'
assert FakeLaminar.spans[0]['span_type'] == 'LLM'
assert FakeLaminar.spans[0]['input']['tools_count'] == 2
- assert FakeLaminar.spans[0]['attributes'][-1]['input_tokens'] == 100
- assert FakeLaminar.spans[0]['attributes'][-1]['output_tokens'] == 20
- assert FakeLaminar.spans[0]['attributes'][-1]['total_tokens'] == 120
+ assert FakeLaminar.spans[0]['input']['messages'][0]['content'][0]['text'] == 'Find the title on example.com'
+ span_attrs = {}
+ for attributes in FakeLaminar.spans[0]['attributes']:
+ span_attrs.update(attributes)
+ assert span_attrs['input_tokens'] == 100
+ assert span_attrs['output_tokens'] == 20
+ assert span_attrs['total_tokens'] == 120
+ assert span_attrs['gen_ai.operation.name'] == 'chat'
+ assert span_attrs['gen_ai.request.model'] == 'gpt-test'
+ assert 'Find the title on example.com' in span_attrs['gen_ai.input.messages']
+ assert span_attrs['gen_ai.usage.input_tokens'] == 100
+ assert span_attrs['gen_ai.usage.output_tokens'] == 20
+ assert span_attrs['gen_ai.usage.total_tokens'] == 120
assert FakeLaminar.spans[0]['outputs'][-1]['assistant_output_preview'] == 'laminar answer'
+ assert FakeLaminar.spans[0]['outputs'][-1]['messages'][0]['content'][0]['text'] == 'laminar answer'
async def test_rust_agent_authenticate_cloud_sync_logs_browser_use_warning(monkeypatch):