feat: add support for agent context history

Author: jpvajda

deepgram/clients/agent/__init__.py

@@ -38,6 +38,7 @@
     InjectUserMessageOptions,
     FunctionCallResponse,
     AgentKeepAlive,
+    Flags,
     # sub level
     Listen,
     Speak,
@@ -53,4 +54,8 @@
     Output,
     Audio,
     Endpoint,
+    Context,
+    HistoryConversationMessage,
+    HistoryFunctionCallsMessage,
+    FunctionCallHistory,
 )

deepgram/clients/agent/client.py

@@ -37,6 +37,7 @@
     InjectUserMessageOptions as LatestInjectUserMessageOptions,
     FunctionCallResponse as LatestFunctionCallResponse,
     AgentKeepAlive as LatestAgentKeepAlive,
+    Flags as LatestFlags,
     # sub level
     Listen as LatestListen,
     Speak as LatestSpeak,
@@ -52,6 +53,10 @@
     Output as LatestOutput,
     Audio as LatestAudio,
     Endpoint as LatestEndpoint,
+    Context as LatestContext,
+    HistoryConversationMessage as LatestHistoryConversationMessage,
+    HistoryFunctionCallsMessage as LatestHistoryFunctionCallsMessage,
+    FunctionCallHistory as LatestFunctionCallHistory,
 )
 
 
@@ -85,6 +90,7 @@
 InjectUserMessageOptions = LatestInjectUserMessageOptions
 FunctionCallResponse = LatestFunctionCallResponse
 AgentKeepAlive = LatestAgentKeepAlive
+Flags = LatestFlags
 
 Listen = LatestListen
 Speak = LatestSpeak
@@ -100,3 +106,7 @@
 Output = LatestOutput
 Audio = LatestAudio
 Endpoint = LatestEndpoint
+Context = LatestContext
+HistoryConversationMessage = LatestHistoryConversationMessage
+HistoryFunctionCallsMessage = LatestHistoryFunctionCallsMessage
+FunctionCallHistory = LatestFunctionCallHistory

deepgram/clients/agent/v1/__init__.py

@@ -42,6 +42,7 @@
     InjectUserMessageOptions,
     FunctionCallResponse,
     AgentKeepAlive,
+    Flags,
     # sub level
     Listen,
     Speak,
@@ -57,4 +58,8 @@
     Output,
     Audio,
     Endpoint,
+    Context,
+    HistoryConversationMessage,
+    HistoryFunctionCallsMessage,
+    FunctionCallHistory,
 )

deepgram/clients/agent/v1/websocket/__init__.py

@@ -33,6 +33,7 @@
     InjectUserMessageOptions,
     FunctionCallResponse,
     AgentKeepAlive,
+    Flags,
     # sub level
     Listen,
     Speak,
@@ -48,4 +49,8 @@
     Output,
     Audio,
     Endpoint,
+    Context,
+    HistoryConversationMessage,
+    HistoryFunctionCallsMessage,
+    FunctionCallHistory,
 )

deepgram/clients/agent/v1/websocket/options.py

@@ -244,6 +244,82 @@ def __getitem__(self, key):
         return _dict[key]
 
 
+# History and Context classes for Function Call Context / History feature
+
+@dataclass
+class Flags(BaseResponse):
+    """
+    This class defines configuration flags for the agent settings.
+    """
+
+    history: bool = field(default=True)
+
+
+@dataclass
+class HistoryConversationMessage(BaseResponse):
+    """
+    This class defines a conversation text message as part of the conversation history.
+    """
+
+    type: str = field(default="History")
+    role: str = field(default="")  # "user" or "assistant"
+    content: str = field(default="")
+
+
+@dataclass
+class FunctionCallHistory(BaseResponse):
+    """
+    This class defines a single function call in the history.
+    """
+
+    id: str = field(default="")
+    name: str = field(default="")
+    client_side: bool = field(default=False)
+    arguments: str = field(default="")
+    response: str = field(default="")
+
+
+@dataclass
+class HistoryFunctionCallsMessage(BaseResponse):
+    """
+    This class defines function call messages as part of the conversation history.
+    """
+
+    type: str = field(default="History")
+    function_calls: List[FunctionCallHistory] = field(default_factory=list)
+
+    def __post_init__(self):
+        """Convert dict function_calls to FunctionCallHistory objects if needed."""
+        if self.function_calls:
+            self.function_calls = [
+                FunctionCallHistory.from_dict(call) if isinstance(call, dict) else call
+                for call in self.function_calls
+            ]
+
+
+@dataclass
+class Context(BaseResponse):
+    """
+    This class defines the conversation context including the history of messages and function calls.
+    """
+
+    messages: List[Union[HistoryConversationMessage, HistoryFunctionCallsMessage]] = field(default_factory=list)
+
+    def __post_init__(self):
+        """Convert dict messages to appropriate message objects if needed."""
+        if self.messages:
+            converted_messages = []
+            for message in self.messages:
+                if isinstance(message, dict):
+                    if "function_calls" in message:
+                        converted_messages.append(HistoryFunctionCallsMessage.from_dict(message))
+                    else:
+                        converted_messages.append(HistoryConversationMessage.from_dict(message))
+                else:
+                    converted_messages.append(message)
+            self.messages = converted_messages
+
+
 @dataclass
 class Agent(BaseResponse):
     """
@@ -277,6 +353,9 @@ class Agent(BaseResponse):
     tags: Optional[List[str]] = field(
         default=None, metadata=dataclass_config(exclude=lambda f: f is None)
     )
+    context: Optional[Context] = field(
+        default=None, metadata=dataclass_config(exclude=lambda f: f is None)
+    )
 
     def __post_init__(self):
         """Handle conversion of dict/list data to proper Speak objects"""
@@ -300,6 +379,8 @@ def __getitem__(self, key):
                 _dict["speak"] = [Speak.from_dict(item) for item in _dict["speak"]]
             elif isinstance(_dict["speak"], dict):
                 _dict["speak"] = Speak.from_dict(_dict["speak"])
+        if "context" in _dict and isinstance(_dict["context"], dict):
+            _dict["context"] = Context.from_dict(_dict["context"])
         return _dict[key]
 
 
@@ -355,13 +436,19 @@ class SettingsOptions(BaseResponse):
     type: str = str(AgentWebSocketEvents.Settings)
     audio: Audio = field(default_factory=Audio)
     agent: Agent = field(default_factory=Agent)
+    flags: Optional[Flags] = field(
+        default=None, metadata=dataclass_config(exclude=lambda f: f is None)
+    )
 
     def __getitem__(self, key):
         _dict = self.to_dict()
         if "audio" in _dict and isinstance(_dict["audio"], dict):
             _dict["audio"] = Audio.from_dict(_dict["audio"])
         if "agent" in _dict and isinstance(_dict["agent"], dict):
             _dict["agent"] = Agent.from_dict(_dict["agent"])
+        if "flags" in _dict and isinstance(_dict["flags"], dict):
+            _dict["flags"] = Flags.from_dict(_dict["flags"])
+        return _dict[key]
 
     def check(self):
         """