From 45719c580bbfaf054932e36ee52997d2828372bb Mon Sep 17 00:00:00 2001 From: Twirre Meulenbelt <43213592+TwirreM@users.noreply.github.com> Date: Thu, 8 Jan 2026 10:49:13 +0100 Subject: [PATCH 1/7] feat: prepend more silence before speech audio for better transcription beginnings ref: N25B-429 --- .env.example | 2 +- src/control_backend/agents/perception/vad_agent.py | 12 +++++++----- src/control_backend/core/config.py | 2 ++ 3 files changed, 10 insertions(+), 6 deletions(-) diff --git a/.env.example b/.env.example index d498054..41a382a 100644 --- a/.env.example +++ b/.env.example @@ -10,7 +10,7 @@ LLM_SETTINGS__LOCAL_LLM_URL="http://localhost:1234/v1/chat/completions" LLM_SETTINGS__LOCAL_LLM_MODEL="gpt-oss" # Number of non-speech chunks to wait before speech ended. A chunk is approximately 31 ms. Increasing this number allows longer pauses in speech, but also increases response time. -BEHAVIOUR_SETTINGS__VAD_NON_SPEECH_PATIENCE_CHUNKS=3 +BEHAVIOUR_SETTINGS__VAD_NON_SPEECH_PATIENCE_CHUNKS=15 # Timeout in milliseconds for socket polling. Increase this number if network latency/jitter is high, often the case when using Wi-Fi. Perhaps 500 ms. A symptom of this issue is transcriptions getting cut off. BEHAVIOUR_SETTINGS__SOCKET_POLLER_TIMEOUT_MS=100 diff --git a/src/control_backend/agents/perception/vad_agent.py b/src/control_backend/agents/perception/vad_agent.py index 70fa9e1..e47b27a 100644 --- a/src/control_backend/agents/perception/vad_agent.py +++ b/src/control_backend/agents/perception/vad_agent.py @@ -229,10 +229,11 @@ class VADAgent(BaseAgent): assert self.model is not None prob = self.model(torch.from_numpy(chunk), settings.vad_settings.sample_rate_hz).item() non_speech_patience = settings.behaviour_settings.vad_non_speech_patience_chunks + begin_silence_length = settings.behaviour_settings.vad_begin_silence_chunks prob_threshold = settings.behaviour_settings.vad_prob_threshold if prob > prob_threshold: - if self.i_since_speech > non_speech_patience: + if self.i_since_speech > non_speech_patience + begin_silence_length: self.logger.debug("Speech started.") self.audio_buffer = np.append(self.audio_buffer, chunk) self.i_since_speech = 0 @@ -246,11 +247,12 @@ class VADAgent(BaseAgent): continue # Speech probably ended. Make sure we have a usable amount of data. - if len(self.audio_buffer) >= 3 * len(chunk): + if len(self.audio_buffer) > begin_silence_length * len(chunk): self.logger.debug("Speech ended.") assert self.audio_out_socket is not None await self.audio_out_socket.send(self.audio_buffer[: -2 * len(chunk)].tobytes()) - # At this point, we know that the speech has ended. - # Prepend the last chunk that had no speech, for a more fluent boundary - self.audio_buffer = chunk + # At this point, we know that there is no speech. + # Prepend the last few chunks that had no speech, for a more fluent boundary. + self.audio_buffer = np.append(self.audio_buffer, chunk) + self.audio_buffer = self.audio_buffer[-begin_silence_length * len(chunk) :] diff --git a/src/control_backend/core/config.py b/src/control_backend/core/config.py index 2ed5c04..02018ee 100644 --- a/src/control_backend/core/config.py +++ b/src/control_backend/core/config.py @@ -73,6 +73,7 @@ class BehaviourSettings(BaseModel): :ivar vad_prob_threshold: Probability threshold for Voice Activity Detection. :ivar vad_initial_since_speech: Initial value for 'since speech' counter in VAD. :ivar vad_non_speech_patience_chunks: Number of non-speech chunks to wait before speech ended. + :ivar vad_begin_silence_chunks: The number of chunks of silence to prepend to speech chunks. :ivar transcription_max_concurrent_tasks: Maximum number of concurrent transcription tasks. :ivar transcription_words_per_minute: Estimated words per minute for transcription timing. :ivar transcription_words_per_token: Estimated words per token for transcription timing. @@ -90,6 +91,7 @@ class BehaviourSettings(BaseModel): vad_prob_threshold: float = 0.5 vad_initial_since_speech: int = 100 vad_non_speech_patience_chunks: int = 15 + vad_begin_silence_chunks: int = 3 # transcription behaviour transcription_max_concurrent_tasks: int = 3 From b88758fa7666b0b0e0c31f6c5f96908fc7acde28 Mon Sep 17 00:00:00 2001 From: Kasper Marinus Date: Thu, 8 Jan 2026 13:33:37 +0100 Subject: [PATCH 2/7] feat: phase transition independent of response ref: N25B-429 --- src/control_backend/agents/bdi/agentspeak_generator.py | 2 +- src/control_backend/agents/bdi/bdi_core_agent.py | 8 ++++++++ 2 files changed, 9 insertions(+), 1 deletion(-) diff --git a/src/control_backend/agents/bdi/agentspeak_generator.py b/src/control_backend/agents/bdi/agentspeak_generator.py index 17248a8..d02888a 100644 --- a/src/control_backend/agents/bdi/agentspeak_generator.py +++ b/src/control_backend/agents/bdi/agentspeak_generator.py @@ -171,7 +171,7 @@ class AgentSpeakGenerator: self._astify(to_phase) if to_phase else AstLiteral("phase", [AstString("end")]) ) - context = [from_phase_ast, ~AstLiteral("responded_this_turn")] + context = [from_phase_ast] if from_phase and from_phase.goals: context.append(self._astify(from_phase.goals[-1], achieved=True)) diff --git a/src/control_backend/agents/bdi/bdi_core_agent.py b/src/control_backend/agents/bdi/bdi_core_agent.py index 99bea80..74a747d 100644 --- a/src/control_backend/agents/bdi/bdi_core_agent.py +++ b/src/control_backend/agents/bdi/bdi_core_agent.py @@ -213,6 +213,14 @@ class BDICoreAgent(BaseAgent): agentspeak.runtime.Intention(), ) + # Check for transitions + self.bdi_agent.call( + agentspeak.Trigger.addition, + agentspeak.GoalType.achievement, + agentspeak.Literal("transition_phase"), + agentspeak.runtime.Intention(), + ) + self._wake_bdi_loop.set() self.logger.debug(f"Added belief {self.format_belief_string(name, args)}") From 625ef0c36543474f44a91c8641172871af255b28 Mon Sep 17 00:00:00 2001 From: Kasper Marinus Date: Thu, 8 Jan 2026 13:36:03 +0100 Subject: [PATCH 3/7] feat: phase transition waits for all goals ref: N25B-429 --- src/control_backend/agents/bdi/agentspeak_generator.py | 5 +++-- 1 file changed, 3 insertions(+), 2 deletions(-) diff --git a/src/control_backend/agents/bdi/agentspeak_generator.py b/src/control_backend/agents/bdi/agentspeak_generator.py index d02888a..22b0b8e 100644 --- a/src/control_backend/agents/bdi/agentspeak_generator.py +++ b/src/control_backend/agents/bdi/agentspeak_generator.py @@ -172,8 +172,9 @@ class AgentSpeakGenerator: ) context = [from_phase_ast] - if from_phase and from_phase.goals: - context.append(self._astify(from_phase.goals[-1], achieved=True)) + if from_phase: + for goal in from_phase.goals: + context.append(self._astify(goal, achieved=True)) body = [ AstStatement(StatementType.REMOVE_BELIEF, from_phase_ast), From 4d0ba6944300a14525c0aef5ab1be60cecc42d1d Mon Sep 17 00:00:00 2001 From: Kasper Marinus Date: Thu, 8 Jan 2026 13:44:25 +0100 Subject: [PATCH 4/7] fix: don't re-add user_said upon phase transition ref: N25B-429 --- .../agents/bdi/agentspeak_generator.py | 22 +++++++++---------- 1 file changed, 11 insertions(+), 11 deletions(-) diff --git a/src/control_backend/agents/bdi/agentspeak_generator.py b/src/control_backend/agents/bdi/agentspeak_generator.py index 22b0b8e..51d3a63 100644 --- a/src/control_backend/agents/bdi/agentspeak_generator.py +++ b/src/control_backend/agents/bdi/agentspeak_generator.py @@ -181,17 +181,17 @@ class AgentSpeakGenerator: AstStatement(StatementType.ADD_BELIEF, to_phase_ast), ] - if from_phase: - body.extend( - [ - AstStatement( - StatementType.TEST_GOAL, AstLiteral("user_said", [AstVar("Message")]) - ), - AstStatement( - StatementType.REPLACE_BELIEF, AstLiteral("user_said", [AstVar("Message")]) - ), - ] - ) + # if from_phase: + # body.extend( + # [ + # AstStatement( + # StatementType.TEST_GOAL, AstLiteral("user_said", [AstVar("Message")]) + # ), + # AstStatement( + # StatementType.REPLACE_BELIEF, AstLiteral("user_said", [AstVar("Message")]) + # ), + # ] + # ) # Notify outside world about transition body.append( From 133019a92823646ace45a9cc9a3bf338c95af088 Mon Sep 17 00:00:00 2001 From: Kasper Marinus Date: Thu, 8 Jan 2026 14:04:44 +0100 Subject: [PATCH 5/7] feat: trigger name and trigger checks on belief update ref: N25B-429 --- src/control_backend/agents/bdi/bdi_core_agent.py | 8 ++++++++ src/control_backend/schemas/program.py | 1 - 2 files changed, 8 insertions(+), 1 deletion(-) diff --git a/src/control_backend/agents/bdi/bdi_core_agent.py b/src/control_backend/agents/bdi/bdi_core_agent.py index 74a747d..1658ccd 100644 --- a/src/control_backend/agents/bdi/bdi_core_agent.py +++ b/src/control_backend/agents/bdi/bdi_core_agent.py @@ -221,6 +221,14 @@ class BDICoreAgent(BaseAgent): agentspeak.runtime.Intention(), ) + # Check triggers + self.bdi_agent.call( + agentspeak.Trigger.addition, + agentspeak.GoalType.achievement, + agentspeak.Literal("check_triggers"), + agentspeak.runtime.Intention(), + ) + self._wake_bdi_loop.set() self.logger.debug(f"Added belief {self.format_belief_string(name, args)}") diff --git a/src/control_backend/schemas/program.py b/src/control_backend/schemas/program.py index 82c017e..3c8c7b4 100644 --- a/src/control_backend/schemas/program.py +++ b/src/control_backend/schemas/program.py @@ -180,7 +180,6 @@ class Trigger(ProgramElement): :ivar plan: The plan to execute. """ - name: str = "" condition: Belief plan: Plan From 866d7c4958b716c773a745d433bf4f6264bf855e Mon Sep 17 00:00:00 2001 From: Kasper Marinus Date: Thu, 8 Jan 2026 15:13:12 +0100 Subject: [PATCH 6/7] fix: end phase loop correctly notifies about user_said ref: N25B-429 --- src/control_backend/agents/bdi/agentspeak_generator.py | 5 ++++- 1 file changed, 4 insertions(+), 1 deletion(-) diff --git a/src/control_backend/agents/bdi/agentspeak_generator.py b/src/control_backend/agents/bdi/agentspeak_generator.py index 51d3a63..7c70a28 100644 --- a/src/control_backend/agents/bdi/agentspeak_generator.py +++ b/src/control_backend/agents/bdi/agentspeak_generator.py @@ -145,7 +145,10 @@ class AgentSpeakGenerator: type=TriggerType.ADDED_BELIEF, trigger_literal=AstLiteral("user_said", [AstVar("Message")]), context=[AstLiteral("phase", [AstString("end")])], - body=[AstStatement(StatementType.ACHIEVE_GOAL, AstLiteral("reply"))], + body=[ + AstStatement(StatementType.DO_ACTION, AstLiteral("notify_user_said")), + AstStatement(StatementType.ACHIEVE_GOAL, AstLiteral("reply")), + ], ) ) From 4b71981a3e7e7d70fd7cfdd1026907544d0aa700 Mon Sep 17 00:00:00 2001 From: Twirre Meulenbelt <43213592+TwirreM@users.noreply.github.com> Date: Mon, 12 Jan 2026 09:00:50 +0100 Subject: [PATCH 7/7] fix: some bugs and some tests ref: N25B-429 --- .../agents/bdi/bdi_core_agent.py | 1 - .../agents/bdi/bdi_program_manager.py | 10 +- .../agents/bdi/default_behavior.asl | 1 + .../agents/bdi/text_belief_extractor_agent.py | 14 +- .../communication/ri_communication_agent.py | 2 +- src/control_backend/core/config.py | 2 +- .../agents/bdi/test_bdi_program_manager.py | 4 +- .../agents/bdi/test_text_belief_extractor.py | 156 ++++++++++-------- 8 files changed, 103 insertions(+), 87 deletions(-) diff --git a/src/control_backend/agents/bdi/bdi_core_agent.py b/src/control_backend/agents/bdi/bdi_core_agent.py index 1658ccd..0b6fb46 100644 --- a/src/control_backend/agents/bdi/bdi_core_agent.py +++ b/src/control_backend/agents/bdi/bdi_core_agent.py @@ -101,7 +101,6 @@ class BDICoreAgent(BaseAgent): maybe_more_work = True while maybe_more_work: maybe_more_work = False - self.logger.debug("Stepping BDI.") if self.bdi_agent.step(): maybe_more_work = True diff --git a/src/control_backend/agents/bdi/bdi_program_manager.py b/src/control_backend/agents/bdi/bdi_program_manager.py index 8b8c68f..13bce95 100644 --- a/src/control_backend/agents/bdi/bdi_program_manager.py +++ b/src/control_backend/agents/bdi/bdi_program_manager.py @@ -67,14 +67,14 @@ class BDIProgramManager(BaseAgent): await self.send(msg) - def handle_message(self, msg: InternalMessage): + async def handle_message(self, msg: InternalMessage): match msg.thread: case "transition_phase": phases = json.loads(msg.body) - self._transition_phase(phases["old"], phases["new"]) + await self._transition_phase(phases["old"], phases["new"]) - def _transition_phase(self, old: str, new: str): + async def _transition_phase(self, old: str, new: str): assert old == str(self._phase.id) if new == "end": @@ -85,8 +85,8 @@ class BDIProgramManager(BaseAgent): if str(phase.id) == new: self._phase = phase - self._send_beliefs_to_semantic_belief_extractor() - self._send_goals_to_semantic_belief_extractor() + await self._send_beliefs_to_semantic_belief_extractor() + await self._send_goals_to_semantic_belief_extractor() # Notify user interaction agent msg = InternalMessage( diff --git a/src/control_backend/agents/bdi/default_behavior.asl b/src/control_backend/agents/bdi/default_behavior.asl index 249689a..f7ae16d 100644 --- a/src/control_backend/agents/bdi/default_behavior.asl +++ b/src/control_backend/agents/bdi/default_behavior.asl @@ -1,5 +1,6 @@ norms(""). +user_said(Message) : norms(Norms) <- + .notify_user_said(Message); -user_said(Message); .reply(Message, Norms). diff --git a/src/control_backend/agents/bdi/text_belief_extractor_agent.py b/src/control_backend/agents/bdi/text_belief_extractor_agent.py index feabf40..ebd9a65 100644 --- a/src/control_backend/agents/bdi/text_belief_extractor_agent.py +++ b/src/control_backend/agents/bdi/text_belief_extractor_agent.py @@ -90,7 +90,7 @@ class TextBeliefExtractorAgent(BaseAgent): self.logger.debug("Received text from LLM: %s", msg.body) self._apply_conversation_message(ChatMessage(role="assistant", content=msg.body)) case settings.agent_settings.bdi_program_manager_name: - self._handle_program_manager_message(msg) + await self._handle_program_manager_message(msg) case _: self.logger.info("Discarding message from %s", sender) return @@ -105,7 +105,7 @@ class TextBeliefExtractorAgent(BaseAgent): length_limit = settings.behaviour_settings.conversation_history_length_limit self.conversation.messages = (self.conversation.messages + [message])[-length_limit:] - def _handle_program_manager_message(self, msg: InternalMessage): + async def _handle_program_manager_message(self, msg: InternalMessage): """ Handle a message from the program manager: extract available beliefs and goals from it. @@ -114,8 +114,10 @@ class TextBeliefExtractorAgent(BaseAgent): match msg.thread: case "beliefs": self._handle_beliefs_message(msg) + await self._infer_new_beliefs() case "goals": self._handle_goals_message(msg) + await self._infer_goal_completions() case "conversation_history": if msg.body == "reset": self._reset() @@ -141,8 +143,9 @@ class TextBeliefExtractorAgent(BaseAgent): available_beliefs = [b for b in belief_list.beliefs if isinstance(b, SemanticBelief)] self.belief_inferrer.available_beliefs = available_beliefs self.logger.debug( - "Received %d semantic beliefs from the program manager.", + "Received %d semantic beliefs from the program manager: %s", len(available_beliefs), + ", ".join(b.name for b in available_beliefs), ) def _handle_goals_message(self, msg: InternalMessage): @@ -158,8 +161,9 @@ class TextBeliefExtractorAgent(BaseAgent): available_goals = [g for g in goals_list.goals if g.can_fail] self.goal_inferrer.goals = available_goals self.logger.debug( - "Received %d failable goals from the program manager.", + "Received %d failable goals from the program manager: %s", len(available_goals), + ", ".join(g.name for g in available_goals), ) async def _user_said(self, text: str): @@ -183,6 +187,7 @@ class TextBeliefExtractorAgent(BaseAgent): new_beliefs = conversation_beliefs - self._current_beliefs if not new_beliefs: + self.logger.debug("No new beliefs detected.") return self._current_beliefs |= new_beliefs @@ -217,6 +222,7 @@ class TextBeliefExtractorAgent(BaseAgent): self._current_goal_completions[goal] = achieved if not new_achieved and not new_not_achieved: + self.logger.debug("No goal achievement changes detected.") return belief_changes = BeliefMessage( diff --git a/src/control_backend/agents/communication/ri_communication_agent.py b/src/control_backend/agents/communication/ri_communication_agent.py index 443e87c..b12bac6 100644 --- a/src/control_backend/agents/communication/ri_communication_agent.py +++ b/src/control_backend/agents/communication/ri_communication_agent.py @@ -248,7 +248,7 @@ class RICommunicationAgent(BaseAgent): self._req_socket.recv_json(), timeout=seconds_to_wait_total / 2 ) - if message["endpoint"] and message["endpoint"] != "ping": + if "endpoint" in message and message["endpoint"] != "ping": self.logger.debug(f'Received message "{message}" from RI.') if "endpoint" not in message: self.logger.warning("No received endpoint in message, expected ping endpoint.") diff --git a/src/control_backend/core/config.py b/src/control_backend/core/config.py index 02018ee..6deb1b8 100644 --- a/src/control_backend/core/config.py +++ b/src/control_backend/core/config.py @@ -91,7 +91,7 @@ class BehaviourSettings(BaseModel): vad_prob_threshold: float = 0.5 vad_initial_since_speech: int = 100 vad_non_speech_patience_chunks: int = 15 - vad_begin_silence_chunks: int = 3 + vad_begin_silence_chunks: int = 6 # transcription behaviour transcription_max_concurrent_tasks: int = 3 diff --git a/test/unit/agents/bdi/test_bdi_program_manager.py b/test/unit/agents/bdi/test_bdi_program_manager.py index 50dc4ed..217e814 100644 --- a/test/unit/agents/bdi/test_bdi_program_manager.py +++ b/test/unit/agents/bdi/test_bdi_program_manager.py @@ -108,8 +108,8 @@ async def test_send_clear_llm_history(mock_settings): await manager._send_clear_llm_history() - assert manager.send.await_count == 1 - msg: InternalMessage = manager.send.await_args[0][0] + assert manager.send.await_count == 2 + msg: InternalMessage = manager.send.await_args_list[0][0][0] # Verify the content and recipient assert msg.body == "clear_history" diff --git a/test/unit/agents/bdi/test_text_belief_extractor.py b/test/unit/agents/bdi/test_text_belief_extractor.py index 176afd2..6782ba1 100644 --- a/test/unit/agents/bdi/test_text_belief_extractor.py +++ b/test/unit/agents/bdi/test_text_belief_extractor.py @@ -6,10 +6,13 @@ import httpx import pytest from control_backend.agents.bdi import TextBeliefExtractorAgent +from control_backend.agents.bdi.text_belief_extractor_agent import BeliefState from control_backend.core.agent_system import InternalMessage from control_backend.core.config import settings from control_backend.schemas.belief_list import BeliefList +from control_backend.schemas.belief_message import Belief as InternalBelief from control_backend.schemas.belief_message import BeliefMessage +from control_backend.schemas.chat_history import ChatHistory, ChatMessage from control_backend.schemas.program import ( ConditionalNorm, KeywordBelief, @@ -23,11 +26,21 @@ from control_backend.schemas.program import ( @pytest.fixture -def agent(): - agent = TextBeliefExtractorAgent("text_belief_agent") - agent.send = AsyncMock() - agent._query_llm = AsyncMock() - return agent +def llm(): + llm = TextBeliefExtractorAgent.LLM(MagicMock(), 4) + llm._query_llm = AsyncMock() + return llm + + +@pytest.fixture +def agent(llm): + with patch( + "control_backend.agents.bdi.text_belief_extractor_agent.TextBeliefExtractorAgent.LLM", + return_value=llm, + ): + agent = TextBeliefExtractorAgent("text_belief_agent") + agent.send = AsyncMock() + return agent @pytest.fixture @@ -102,24 +115,12 @@ async def test_handle_message_from_transcriber(agent, mock_settings): agent.send.assert_awaited_once() # noqa # `agent.send` has no such property, but we mock it. sent: InternalMessage = agent.send.call_args.args[0] # noqa - assert sent.to == mock_settings.agent_settings.bdi_belief_collector_name + assert sent.to == mock_settings.agent_settings.bdi_core_name assert sent.thread == "beliefs" - parsed = json.loads(sent.body) - assert parsed == {"beliefs": {"user_said": [transcription]}, "type": "belief_extraction_text"} - - -@pytest.mark.asyncio -async def test_process_user_said(agent, mock_settings): - transcription = "this is a test" - - await agent._user_said(transcription) - - agent.send.assert_awaited_once() # noqa # `agent.send` has no such property, but we mock it. - sent: InternalMessage = agent.send.call_args.args[0] # noqa - assert sent.to == mock_settings.agent_settings.bdi_belief_collector_name - assert sent.thread == "beliefs" - parsed = json.loads(sent.body) - assert parsed["beliefs"]["user_said"] == [transcription] + parsed = BeliefMessage.model_validate_json(sent.body) + replaced_last = parsed.replace.pop() + assert replaced_last.name == "user_said" + assert replaced_last.arguments == [transcription] @pytest.mark.asyncio @@ -144,46 +145,46 @@ async def test_query_llm(): "control_backend.agents.bdi.text_belief_extractor_agent.httpx.AsyncClient", return_value=mock_async_client, ): - agent = TextBeliefExtractorAgent("text_belief_agent") + llm = TextBeliefExtractorAgent.LLM(MagicMock(), 4) - res = await agent._query_llm("hello world", {"type": "null"}) + res = await llm._query_llm("hello world", {"type": "null"}) # Response content was set as "null", so should be deserialized as None assert res is None @pytest.mark.asyncio -async def test_retry_query_llm_success(agent): - agent._query_llm.return_value = None - res = await agent._retry_query_llm("hello world", {"type": "null"}) +async def test_retry_query_llm_success(llm): + llm._query_llm.return_value = None + res = await llm.query("hello world", {"type": "null"}) - agent._query_llm.assert_called_once() + llm._query_llm.assert_called_once() assert res is None @pytest.mark.asyncio -async def test_retry_query_llm_success_after_failure(agent): - agent._query_llm.side_effect = [KeyError(), "real value"] - res = await agent._retry_query_llm("hello world", {"type": "string"}) +async def test_retry_query_llm_success_after_failure(llm): + llm._query_llm.side_effect = [KeyError(), "real value"] + res = await llm.query("hello world", {"type": "string"}) - assert agent._query_llm.call_count == 2 + assert llm._query_llm.call_count == 2 assert res == "real value" @pytest.mark.asyncio -async def test_retry_query_llm_failures(agent): - agent._query_llm.side_effect = [KeyError(), KeyError(), KeyError(), "real value"] - res = await agent._retry_query_llm("hello world", {"type": "string"}) +async def test_retry_query_llm_failures(llm): + llm._query_llm.side_effect = [KeyError(), KeyError(), KeyError(), "real value"] + res = await llm.query("hello world", {"type": "string"}) - assert agent._query_llm.call_count == 3 + assert llm._query_llm.call_count == 3 assert res is None @pytest.mark.asyncio -async def test_retry_query_llm_fail_immediately(agent): - agent._query_llm.side_effect = [KeyError(), "real value"] - res = await agent._retry_query_llm("hello world", {"type": "string"}, tries=1) +async def test_retry_query_llm_fail_immediately(llm): + llm._query_llm.side_effect = [KeyError(), "real value"] + res = await llm.query("hello world", {"type": "string"}, tries=1) - assert agent._query_llm.call_count == 1 + assert llm._query_llm.call_count == 1 assert res is None @@ -192,7 +193,7 @@ async def test_extracting_semantic_beliefs(agent): """ The Program Manager sends beliefs to this agent. Test whether the agent handles them correctly. """ - assert len(agent.available_beliefs) == 0 + assert len(agent.belief_inferrer.available_beliefs) == 0 beliefs = BeliefList( beliefs=[ KeywordBelief( @@ -213,26 +214,28 @@ async def test_extracting_semantic_beliefs(agent): to=settings.agent_settings.text_belief_extractor_name, sender=settings.agent_settings.bdi_program_manager_name, body=beliefs.model_dump_json(), + thread="beliefs", ), ) - assert len(agent.available_beliefs) == 2 + assert len(agent.belief_inferrer.available_beliefs) == 2 @pytest.mark.asyncio -async def test_handle_invalid_program(agent, sample_program): - agent.available_beliefs.append(sample_program.phases[0].norms[0].condition) - agent.available_beliefs.append(sample_program.phases[0].triggers[0].condition) - assert len(agent.available_beliefs) == 2 +async def test_handle_invalid_beliefs(agent, sample_program): + agent.belief_inferrer.available_beliefs.append(sample_program.phases[0].norms[0].condition) + agent.belief_inferrer.available_beliefs.append(sample_program.phases[0].triggers[0].condition) + assert len(agent.belief_inferrer.available_beliefs) == 2 await agent.handle_message( InternalMessage( to=settings.agent_settings.text_belief_extractor_name, sender=settings.agent_settings.bdi_program_manager_name, body=json.dumps({"phases": "Invalid"}), + thread="beliefs", ), ) - assert len(agent.available_beliefs) == 2 + assert len(agent.belief_inferrer.available_beliefs) == 2 @pytest.mark.asyncio @@ -254,13 +257,13 @@ async def test_handle_robot_response(agent): @pytest.mark.asyncio -async def test_simulated_real_turn_with_beliefs(agent, sample_program): +async def test_simulated_real_turn_with_beliefs(agent, llm, sample_program): """Test sending user message to extract beliefs from.""" - agent.available_beliefs.append(sample_program.phases[0].norms[0].condition) - agent.available_beliefs.append(sample_program.phases[0].triggers[0].condition) + agent.belief_inferrer.available_beliefs.append(sample_program.phases[0].norms[0].condition) + agent.belief_inferrer.available_beliefs.append(sample_program.phases[0].triggers[0].condition) # Send a user message with the belief that there's no more booze - agent._query_llm.return_value = {"is_pirate": None, "no_more_booze": True} + llm._query_llm.return_value = {"is_pirate": None, "no_more_booze": True} assert len(agent.conversation.messages) == 0 await agent.handle_message( InternalMessage( @@ -275,20 +278,20 @@ async def test_simulated_real_turn_with_beliefs(agent, sample_program): assert agent.send.call_count == 2 # First should be the beliefs message - message: InternalMessage = agent.send.call_args_list[0].args[0] + message: InternalMessage = agent.send.call_args_list[1].args[0] beliefs = BeliefMessage.model_validate_json(message.body) assert len(beliefs.create) == 1 assert beliefs.create[0].name == "no_more_booze" @pytest.mark.asyncio -async def test_simulated_real_turn_no_beliefs(agent, sample_program): +async def test_simulated_real_turn_no_beliefs(agent, llm, sample_program): """Test a user message to extract beliefs from, but no beliefs are formed.""" - agent.available_beliefs.append(sample_program.phases[0].norms[0].condition) - agent.available_beliefs.append(sample_program.phases[0].triggers[0].condition) + agent.belief_inferrer.available_beliefs.append(sample_program.phases[0].norms[0].condition) + agent.belief_inferrer.available_beliefs.append(sample_program.phases[0].triggers[0].condition) # Send a user message with no new beliefs - agent._query_llm.return_value = {"is_pirate": None, "no_more_booze": None} + llm._query_llm.return_value = {"is_pirate": None, "no_more_booze": None} await agent.handle_message( InternalMessage( to=settings.agent_settings.text_belief_extractor_name, @@ -302,17 +305,17 @@ async def test_simulated_real_turn_no_beliefs(agent, sample_program): @pytest.mark.asyncio -async def test_simulated_real_turn_no_new_beliefs(agent, sample_program): +async def test_simulated_real_turn_no_new_beliefs(agent, llm, sample_program): """ Test a user message to extract beliefs from, but no new beliefs are formed because they already existed. """ - agent.available_beliefs.append(sample_program.phases[0].norms[0].condition) - agent.available_beliefs.append(sample_program.phases[0].triggers[0].condition) - agent.beliefs["is_pirate"] = True + agent.belief_inferrer.available_beliefs.append(sample_program.phases[0].norms[0].condition) + agent.belief_inferrer.available_beliefs.append(sample_program.phases[0].triggers[0].condition) + agent._current_beliefs = BeliefState(true={InternalBelief(name="is_pirate", arguments=None)}) # Send a user message with the belief the user is a pirate, still - agent._query_llm.return_value = {"is_pirate": True, "no_more_booze": None} + llm._query_llm.return_value = {"is_pirate": True, "no_more_booze": None} await agent.handle_message( InternalMessage( to=settings.agent_settings.text_belief_extractor_name, @@ -326,17 +329,19 @@ async def test_simulated_real_turn_no_new_beliefs(agent, sample_program): @pytest.mark.asyncio -async def test_simulated_real_turn_remove_belief(agent, sample_program): +async def test_simulated_real_turn_remove_belief(agent, llm, sample_program): """ Test a user message to extract beliefs from, but an existing belief is determined no longer to hold. """ - agent.available_beliefs.append(sample_program.phases[0].norms[0].condition) - agent.available_beliefs.append(sample_program.phases[0].triggers[0].condition) - agent.beliefs["no_more_booze"] = True + agent.belief_inferrer.available_beliefs.append(sample_program.phases[0].norms[0].condition) + agent.belief_inferrer.available_beliefs.append(sample_program.phases[0].triggers[0].condition) + agent._current_beliefs = BeliefState( + true={InternalBelief(name="no_more_booze", arguments=None)}, + ) # Send a user message with the belief the user is a pirate, still - agent._query_llm.return_value = {"is_pirate": None, "no_more_booze": False} + llm._query_llm.return_value = {"is_pirate": None, "no_more_booze": False} await agent.handle_message( InternalMessage( to=settings.agent_settings.text_belief_extractor_name, @@ -349,18 +354,23 @@ async def test_simulated_real_turn_remove_belief(agent, sample_program): assert agent.send.call_count == 2 # Agent's current beliefs should've changed - assert not agent.beliefs["no_more_booze"] + assert any(b.name == "no_more_booze" for b in agent._current_beliefs.false) @pytest.mark.asyncio -async def test_llm_failure_handling(agent, sample_program): +async def test_llm_failure_handling(agent, llm, sample_program): """ Check that the agent handles failures gracefully without crashing. """ - agent._query_llm.side_effect = httpx.HTTPError("") - agent.available_beliefs.append(sample_program.phases[0].norms[0].condition) - agent.available_beliefs.append(sample_program.phases[0].triggers[0].condition) + llm._query_llm.side_effect = httpx.HTTPError("") + agent.belief_inferrer.available_beliefs.append(sample_program.phases[0].norms[0].condition) + agent.belief_inferrer.available_beliefs.append(sample_program.phases[0].triggers[0].condition) - belief_changes = await agent._infer_turn() + belief_changes = await agent.belief_inferrer.infer_from_conversation( + ChatHistory( + messages=[ChatMessage(role="user", content="Good day!")], + ), + ) - assert len(belief_changes) == 0 + assert len(belief_changes.true) == 0 + assert len(belief_changes.false) == 0