Makes select_speaker more robust by checking for mentions anywhere. (m…

…icrosoft#669)

* Makes select_speaker more robust by checking for agents mentioned anywhere in the selection string. Addresses 663.

* Added test coverage for group chat mentions. Refactored mention counter to own function.

* Fixed pre-commit formatting.

autogen/agentchat/groupchat.py

@@ -2,7 +2,7 @@
 import sys
 from dataclasses import dataclass
 from typing import Dict, List, Optional, Union
-
+import re
 from .agent import Agent
 from .conversable_agent import ConversableAgent
 
@@ -101,6 +101,13 @@ def select_speaker(self, last_speaker: Agent, selector: ConversableAgent):
         if not final:
             # i = self._random.randint(0, len(self._agent_names) - 1)  # randomly pick an id
             return self.next_agent(last_speaker, agents)
+
+        # If exactly one agent is mentioned, use it. Otherwise, leave the OAI response unmodified
+        mentions = self._mentioned_agents(name, agents)
+        if len(mentions) == 1:
+            name = next(iter(mentions))
+
+        # Return the result
         try:
             return self.agent_by_name(name)
         except ValueError:
@@ -119,6 +126,22 @@ def _participant_roles(self):
             roles.append(f"{agent.name}: {agent.system_message}")
         return "\n".join(roles)
 
+    def _mentioned_agents(self, message_content: str, agents: List[Agent]) -> Dict:
+        """
+        Finds and counts agent mentions in the string message_content, taking word boundaries into account.
+
+        Returns: A dictionary mapping agent names to mention counts (to be included, at least one mention must occur)
+        """
+        mentions = dict()
+        for agent in agents:
+            regex = (
+                r"(?<=\W)" + re.escape(agent.name) + r"(?=\W)"
+            )  # Finds agent mentions, taking word boundaries into account
+            count = len(re.findall(regex, " " + message_content + " "))  # Pad the message to help with matching
+            if count > 0:
+                mentions[agent.name] = count
+        return mentions
+
 
 class GroupChatManager(ConversableAgent):
     """(In preview) A chat manager agent that can manage a group chat of multiple agents."""

test/agentchat/test_groupchat.py

@@ -1,5 +1,6 @@
 import pytest
 import autogen
+import json
 
 
 def test_func_call_groupchat():
@@ -112,8 +113,80 @@ def test_plugin():
     assert len(groupchat.messages) == 2
 
 
+def test_agent_mentions():
+    agent1 = autogen.ConversableAgent(
+        "alice",
+        max_consecutive_auto_reply=2,
+        human_input_mode="NEVER",
+        llm_config=False,
+        default_auto_reply="This is alice sepaking.",
+    )
+    agent2 = autogen.ConversableAgent(
+        "bob",
+        max_consecutive_auto_reply=2,
+        human_input_mode="NEVER",
+        llm_config=False,
+        default_auto_reply="This is bob speaking.",
+    )
+    agent3 = autogen.ConversableAgent(
+        "sam",
+        max_consecutive_auto_reply=2,
+        human_input_mode="NEVER",
+        llm_config=False,
+        default_auto_reply="This is sam speaking.",
+    )
+    groupchat = autogen.GroupChat(agents=[agent1, agent2, agent3], messages=[], max_round=2)
+
+    # Basic counting
+    assert json.dumps(groupchat._mentioned_agents("", [agent1, agent2, agent3]), sort_keys=True) == "{}"
+    assert json.dumps(groupchat._mentioned_agents("alice", [agent1, agent2, agent3]), sort_keys=True) == '{"alice": 1}'
+    assert (
+        json.dumps(groupchat._mentioned_agents("alice bob alice", [agent1, agent2, agent3]), sort_keys=True)
+        == '{"alice": 2, "bob": 1}'
+    )
+    assert (
+        json.dumps(groupchat._mentioned_agents("alice bob alice sam", [agent1, agent2, agent3]), sort_keys=True)
+        == '{"alice": 2, "bob": 1, "sam": 1}'
+    )
+    assert (
+        json.dumps(groupchat._mentioned_agents("alice bob alice sam robert", [agent1, agent2, agent3]), sort_keys=True)
+        == '{"alice": 2, "bob": 1, "sam": 1}'
+    )
+
+    # Substring
+    assert (
+        json.dumps(groupchat._mentioned_agents("sam samantha basam asami", [agent1, agent2, agent3]), sort_keys=True)
+        == '{"sam": 1}'
+    )
+
+    # Word boundaries
+    assert (
+        json.dumps(groupchat._mentioned_agents("alice! .alice. .alice", [agent1, agent2, agent3]), sort_keys=True)
+        == '{"alice": 3}'
+    )
+
+    # Special characters in agent names
+    agent4 = autogen.ConversableAgent(
+        ".*",
+        max_consecutive_auto_reply=2,
+        human_input_mode="NEVER",
+        llm_config=False,
+        default_auto_reply="Match everything.",
+    )
+
+    groupchat = autogen.GroupChat(agents=[agent1, agent2, agent3, agent4], messages=[], max_round=2)
+    assert (
+        json.dumps(
+            groupchat._mentioned_agents("alice bob alice sam robert .*", [agent1, agent2, agent3, agent4]),
+            sort_keys=True,
+        )
+        == '{".*": 1, "alice": 2, "bob": 1, "sam": 1}'
+    )
+
+
 if __name__ == "__main__":
     test_func_call_groupchat()
     # test_broadcast()
     test_chat_manager()
     # test_plugin()
+    # test_agent_mentions()