Create, prompt, delete and query Conversations.

Create Conversation

If you want to use a custom specification, you will need to create the conversation before prompting.

Here we show an example of creating a conversation to use a specification with the Claude 3 Opus model.

You can also assign content filters to limit the conversation to a subset of ingested content.

g create --type conversation

Prompt Conversation

You can choose to start a new conversation, or choose to continue an existing conversation.

The CLI also allows for followup prompts within the current conversation.

g prompt --type conversation

Delete Conversation

g delete --type conversation --id "{guid}"

Clear Conversations

g clear --type conversation

Get Conversation

When getting a conversation, Graphlit will format the user and assistant messages for easier reading of the conversation. You can add the --json argument to the command-line to get the full JSON response from the API.

g get --type conversation --id "{guid}"

Query Conversations

g query --type conversation

We can also search within conversations, using vector or hybrid search. Graphlit will return the entire conversation where the search 'hit' was located.

g query --type conversation --search "{text}" --search-type Vector

Last updated