If your agent is VF type we recommend directly using the Voiceflowβs API for querying the KB.
You can currently interact with VG through either the built in AI engine or through Voiceflow, the agentPlatform defines how the agent will interact and function across the whole service.
The interact API of voiceflow inspires how we interact with VF/VG agents, when you do an interact POST request the action property functions the same as that on Voiceflowβs end.
The knowledge base is a collection of documents that the agent can use to answer user queries.
You can either search the KB or directly search the vector store to reutnr the most highest similarity chunks.
By default all agents have a 1536 dimension vector store, the text inserted is also chunked into 1024 characters with overlapping of 50 chars, using the langcahin recursiveTextSplitter
All the transcripts of any channel are stored by default.
You can also insert a new conversation transcript to the agent yourself (although not recommended as its done automatically when you interact across any channel)
Popular usecase for the conversations API is to export the conversations and send them via email or in a google sheet.
The analytics API is used to get the analytics of the agent, you can get the analytics of the agent by day, week, month, or year.
You can also insert your own analytics data to the agent.
There are 2 types of analytics that we store:
Load Analytics: Analysation of your website traffic if you deploy using the popup, even if the user doesnβt interacts with the agent it still records useful insights liek geo-analytics.
Conversation Analytics: Whenever the user starts interacting with the agent the conversation analytics kicks in to gather information like interactions number or average time per conversation.