how many chat context we can give to LiveKit agent pipeline along with initial context?

I'm not able to find any data sources to answer your question. You might want to specify if you're referring to the number of messages or the size of the context in terms of data. If this doesn't help, please reach out to the Livekit support team for further assistance. 👍 0 button 👎 0 button 💡 Hint: Mention @runllm in the thread for followups.

can I change the context during the call in voice ai agent pipeline @millions-hair-24956

Yes, you can change the chat context during a call in the VoicePipelineAgent. According to LiveKit documentation, you can modify the chat context before it is sent to the LLM using the

before_llm_cb

callback. This allows you to adjust the context dynamically during the conversation. Would you like more details on how to implement this? Sources: - LiveKit documentation 👍 0 button 👎 0 button 💡 Hint: Mention @runllm in the thread for followups.

how can I conditionally add tools @millions-hair-24956

It seems likely that you can conditionally add tools in the LiveKit VoicePipelineAgent by updating the

FunctionContext

dynamically. According to LiveKit documentation, you can modify the

FunctionContext

to include different tools or functions as needed during the conversation. This can be done by setting

voice_assistant.fnc_ctx

to a new

FunctionContext

instance with the desired tools. Would you like more detailed guidance on implementing this? Sources: - LiveKit documentation 👍 0 button 👎 0 button 💡 Hint: Mention @runllm in the thread for followups.

can you show me with an example @millions-hair-24956

Certainly! Here's an example of how you can conditionally add tools to a LiveKit VoicePipelineAgent by updating the

FunctionContext

.

from livekit.agents import llm
from livekit.agents.pipeline import VoicePipelineAgent

# Define a custom FunctionContext with tools
class CustomFunctionContext(llm.FunctionContext):
    @llm.ai_callable()
    async def tool_one(self):
        """Tool one description."""
        return "Tool one executed"

    @llm.ai_callable()
    async def tool_two(self):
        """Tool two description."""
        return "Tool two executed"

# Initialize the agent with an initial FunctionContext
initial_fnc_ctx = CustomFunctionContext()

agent = VoicePipelineAgent(
    fnc_ctx=initial_fnc_ctx,
    # other parameters...
)

# Function to update the FunctionContext conditionally
def update_function_context(agent, condition):
    if condition:
        new_fnc_ctx = CustomFunctionContext()
        # Add or modify tools based on the condition
        agent.fnc_ctx = new_fnc_ctx

# Example usage
condition = True  # This could be any condition based on your logic
update_function_context(agent, condition)

In this example,

CustomFunctionContext

is defined with two tools,

tool_one

and

tool_two

. The

update_function_context

function updates the agent's

FunctionContext

based on a condition. Would you like more information on this? Sources: - https://docs.livekit.io/sip/examples/sip-participant-attributes/ - https://livekit-users.slack.com/archives/C07FY8WHGPM/p1731463251.597929 👍 0 button 👎 0 button 💡 Hint: Mention @runllm in the thread for followups.

can we integrate langchain with LiveKit?

Yes, you can integrate LangChain with LiveKit. According to a Slack conversation, LiveKit is working on making it possible to transition between states, which would allow for integration with frameworks like LangChain. You can use LangChain to manage complex workflows and state transitions within your LiveKit agents. Would you like more detailed guidance on how to set this up? Sources: - Slack conversation 👍 0 button 👎 0 button 💡 Hint: Mention @runllm in the thread for followups.