Add embedding_model as input param to Agent and ConversationalAgent

bondai/agents/agent.py

@@ -6,10 +6,15 @@
 from typing import Dict, List, Tuple, Callable
 from bondai.util import EventMixin, Runnable, load_local_resource
 from bondai.tools import Tool, ResponseQueryTool
-from bondai.models import LLM
+from bondai.models import LLM, EmbeddingModel
 from bondai.memory import MemoryManager
 from bondai.prompt import JinjaPromptBuilder
-from bondai.models.openai import OpenAILLM, OpenAIModelNames, get_total_cost
+from bondai.models.openai import (
+    OpenAILLM,
+    OpenAIEmbeddingModel,
+    OpenAIModelNames,
+    get_total_cost,
+)
 from .conversation_member import ConversationMember
 from .messages import AgentMessage, AgentMessageList, SystemMessage, ToolUsageMessage
 from .compression import summarize_conversation, summarize_messages
@@ -56,6 +61,7 @@ class Agent(EventMixin, Runnable):
     def __init__(
         self,
         llm: LLM | None = None,
+        embedding_model: EmbeddingModel | None = None,
         tools: List[Tool] | None = None,
         quiet: bool = True,
         allowed_events: List[str] | None = None,
@@ -85,6 +91,10 @@ def __init__(
 
         if llm is None:
             llm = OpenAILLM(OpenAIModelNames.GPT4_0613)
+        if embedding_model is None:
+            embedding_model = OpenAIEmbeddingModel(
+                OpenAIModelNames.TEXT_EMBEDDING_ADA_002
+            )
         if tools is None:
             tools = []
         if system_prompt_sections is None:
@@ -96,6 +106,7 @@ def __init__(
         self._status: AgentStatus = AgentStatus.IDLE
         self._messages = AgentMessageList(messages=messages)
         self._llm: LLM = llm
+        self._embedding_model: EmbeddingModel = embedding_model
         self._tools: List[Tool] = tools
         self._quiet: bool = quiet
         self._system_prompt_sections: List[Callable[[], str]] = system_prompt_sections
@@ -317,7 +328,9 @@ def _run_tool_loop(
         last_error_message = None
         local_messages = []
         self._force_stop = False
-        response_query_tool = ResponseQueryTool()
+        response_query_tool = ResponseQueryTool(
+            llm=self._llm, embedding_model=self._embedding_model
+        )
 
         def append_message(message):
             if isinstance(message, SystemMessage):

bondai/agents/conversational_agent.py

@@ -13,7 +13,7 @@
     ExitConversationTool,
 )
 from bondai.prompt import JinjaPromptBuilder
-from bondai.models.llm import LLM
+from bondai.models import LLM, EmbeddingModel
 from bondai.models.openai import OpenAILLM, OpenAIModelNames, get_total_cost
 from .agent import (
     Agent,
@@ -47,6 +47,7 @@ class ConversationalAgent(Agent, ConversationMember):
     def __init__(
         self,
         llm: LLM | None = None,
+        embedding_model: EmbeddingModel | None = None,
         tools: List[Tool] | None = None,
         messages: List[AgentMessage] | None = None,
         name: str = DEFAULT_AGENT_NAME,
@@ -82,6 +83,7 @@ def __init__(
         Agent.__init__(
             self,
             llm=llm,
+            embedding_model=embedding_model,
             quiet=quiet,
             tools=tools,
             messages=messages,
@@ -251,7 +253,7 @@ def validate_recipient(recipient_name: str):
                     "instructions": self.instructions,
                     "conversation_enabled": self._enable_conversation_tools
                     or self._enable_conversational_content_responses,
-                    "allow_exit": self._enable_exit_conversation,
+                    "enable_exit_conversation": self._enable_exit_conversation,
                 }
 
                 tool_result = self._run_tool_loop(
@@ -320,9 +322,9 @@ def to_dict(self) -> Dict:
         data["persona"] = self._persona
         data["persona_summary"] = self._persona_summary
         data["instructions"] = self.instructions
-        data["allow_exit"] = self._enable_exit_conversation
         data["quiet"] = self._quiet
         data["enable_conversation_tools"] = self._enable_conversation_tools
+        data["enable_exit_conversation"] = self._enable_exit_conversation
         data[
             "enable_conversational_content_responses"
         ] = self._enable_conversational_content_responses
@@ -355,7 +357,7 @@ def from_dict(
             persona=data["persona"],
             persona_summary=data["persona_summary"],
             instructions=data["instructions"],
-            allow_exit=data["allow_exit"],
+            enable_exit_conversation=data["enable_exit_conversation"],
             quiet=data["quiet"],
             enable_conversation_tools=data["enable_conversation_tools"],
             enable_conversational_content_responses=data[

bondai/cli/cli.py

@@ -105,7 +105,7 @@ def build_agents(llm: LLM) -> GroupConversation:
         tools=[AgentTool(task_execution_agent)],
         enable_conversation_tools=False,
         enable_conversational_content_responses=True,
-        allow_exit=False,
+        enable_exit_conversation=False,
         memory_manager=MemoryManager(
             core_memory_datasource=PersistentCoreMemoryDataSource(
                 file_path="./.memory/user_liason_core_memory.json",

website/docs/api-spec/create-agent.md

@@ -12,7 +12,7 @@ This API will create a Conversational Agent.
 
 ```json
 {
-   "allow_exit":true,
+   "enable_exit_conversation":true,
    "enable_conversation_tools":false,
    "enable_conversational_content_responses":true,
    "id":"34c2262b-1a9b-4ace-9b74-54e892ea59a2",

website/docs/api-spec/get-agent.md

@@ -12,7 +12,7 @@ This API returns the current state of an Agent.
 
 ```json
 {
-   "allow_exit":true,
+   "enable_exit_conversation":true,
    "enable_conversation_tools":false,
    "enable_conversational_content_responses":true,
    "id":"34c2262b-1a9b-4ace-9b74-54e892ea59a2",

website/docs/api-spec/list-agents.md

@@ -13,7 +13,7 @@ This API returns a list of all active Agents.
 ```json
 [
     {
-        "allow_exit":true,
+        "enable_exit_conversation":true,
         "enable_conversation_tools":false,
         "enable_conversational_content_responses":true,
         "id":"34c2262b-1a9b-4ace-9b74-54e892ea59a2",