Give me a command to get the chat context messages using session.history but excluding system instructions

Thanks for reaching out, Hiren. I'm an AI assistant and I'm here to help. Give me a moment while I research this topic.

why content is coming as an array.. instead of the string?

I'm glad you asked about the

content

field in

session.history

messages! It’s designed as an array to comfortably support multimodal content, which means a single message can include different parts like text and images all together. Here’s how it works: • Text parts come through as simple strings. • Image parts show up as objects containing important details like type and URL. This setup really helps create rich, flexible interactions, especially when you’re working with both text and visuals. If you’d like to handle the content as one single string, you can easily combine the array elements like this:

# Convert array content to a single string
if isinstance(msg.content, list):
    msg.content = "\n".join(
        "[image]" if isinstance(x, dict) and x.get("type") == "image_url" else x
        for x in msg.content
    )

This way, you keep everything compatible while still enjoying the benefits of multimodal content. If you have any more questions or want to dive deeper, I’m here to help! Sources: tools/server/themes/wild/index.html | VoicePipelineAgent | LiveKit Docs | Working with the MultimodalAgent class | LiveKit Docs

what If I only want text message and want to ignore all the others

I'd be happy to help you filter only the text messages from

session.history

while skipping other content types. You can do this by checking if the content is a string. Here's a simple example for you:

# Filter and process only text messages
for item in session.history.items:
    if isinstance(item.content, list):
        for content in item.content:
            if isinstance(content, str):
                print(f"Text: {content}")

This way, you'll make sure to process only the text parts of the messages and gently skip over other types like images or metadata. If you have any questions or need further assistance, feel free to ask! Sources: Events and error handling | LiveKit Docs | Agent-assisted warm transfer | LiveKit Docs

I want the whole object list but only content's value to be replaced from List to the text content like you've just provided

Let's walk through how you can replace the

content

field's value from a list to text content while keeping the entire object list intact. You can do this by iterating through

session.history.items

and updating the

content

field as needed. Here's a simple example to guide you:

# Replace list content with text while keeping the object list intact
for item in session.history.items:
    if isinstance(item.content, list):
        # Convert list to a single string
        item.content = "\n".join(
            "[image]" if isinstance(x, dict) and x.get("type") == "image_url" else str(x)
            for x in item.content
        )

This approach will help transform the

content

field into a more readable text format, while still preserving the rest of the object structure. Also, if you'd like these changes to be saved across agent turns, don't forget to call

update_chat_ctx()

to sync the modifications back to the active chat context. Sources: Workflows | LiveKit Docs | VoicePipelineAgent | LiveKit Docs | VoicePipelineAgent | LiveKit Docs | Working with the MultimodalAgent class | LiveKit Docs | Working with the MultimodalAgent class | LiveKit Docs