Chat with Assistant

Sends messages to an assistant and receives responses with citations. This is the recommended chat interface, offering more functionality than the OpenAI-compatible interface.

Usage

assistant_chat(
  assistant_name,
  messages,
  model = NULL,
  filter = NULL,
  context_options = NULL
)

Arguments

assistant_name

Name of the assistant

messages

List of message objects. Each message should have: - role: Either "user" or "assistant" - content: The message text

model

Optional model to use (e.g., "gpt-4o"). Uses assistant default if not specified.

filter

Optional metadata filter to limit which files are searched (list)

context_options

Optional list of context options: - top_k: Number of context chunks to retrieve (default: 15) - snippet_size: Maximum size of each snippet in tokens - multimodal: Whether to include image context for PDFs (default: TRUE) - include_binary_content: Include base64 image data when multimodal is TRUE

Value

List with http response, content (chat response), and status_code. Content includes: - id: Response ID - model: Model used - message: Response message with role and content - finish_reason: Reason for completion ("stop", etc.) - citations: List of citations referencing source documents - usage: Token usage statistics

Examples

if (FALSE) { # \dontrun{
# Simple chat
assistant_chat(
  assistant_name = "my-assistant",
  messages = list(
    list(role = "user", content = "What is the main topic of the document?")
  )
)

# Chat with conversation history
assistant_chat(
  assistant_name = "my-assistant",
  messages = list(
    list(role = "user", content = "Who is the CEO?"),
    list(role = "assistant", content = "The CEO is John Smith."),
    list(role = "user", content = "When did they start?")
  )
)

# Chat with metadata filter
assistant_chat(
  assistant_name = "my-assistant",
  messages = list(list(role = "user", content = "Summarize the 2023 report")),
  filter = list(year = 2023)
)

# Chat with context options
assistant_chat(
  assistant_name = "my-assistant",
  messages = list(list(role = "user", content = "Describe the images")),
  context_options = list(
    multimodal = TRUE,
    include_binary_content = TRUE,
    top_k = 10
  )
)
} # }