Merge pull request #59 from jcrodriguez1989/feature.jcr.allow_multipl…

…e_chat_sessions

Feature: Allow multiple chat sessions

R/ask_chatgpt.R

@@ -3,6 +3,9 @@
 #' Note: See also `reset_chat_session`.
 #'
 #' @param question The question to ask ChatGPT.
+#' @param session_id The ID of the session to be used. We can have different conversations by using
+#' different session IDs.
+#' @param openai_api_key OpenAI's API key.
 #'
 #' @examples
 #' \dontrun{
@@ -13,17 +16,19 @@
 #'
 #' @export
 #'
-ask_chatgpt <- function(question) {
+ask_chatgpt <- function(question, session_id = "1", openai_api_key = Sys.getenv("OPENAI_API_KEY")) {
  # Get the existing chat session messages, and add the new message.
- chat_session_messages <- append(get("chat_session_messages", envir = .state), list(
+ chat_session_messages <- append(get_chat_session(session_id), list(
  list(role = "user", content = question)
  ))
  # Send the query to ChatGPT.
- chat_gpt_reply <- parse_response(gpt_get_completions(question, messages = chat_session_messages))
+ chat_gpt_reply <- parse_response(
+ gpt_get_completions(question, openai_api_key, chat_session_messages)
+ )
  chat_session_messages <- append(chat_session_messages, list(
  list(role = "assistant", content = chat_gpt_reply)
  ))
  # Update the chat session messages with the new question and the reply.
- assign("chat_session_messages", chat_session_messages, .state)
+ reset_chat_session(chat_session_messages, session_id)
  chat_gpt_reply
 }

R/chatgpt-package.R

@@ -6,10 +6,6 @@
 .state <- new.env(parent = emptyenv())
 
 # Empty chat session messages at startup.
-assign(
- "chat_session_messages",
- list(list(role = "system", content = "You are a helpful assistant.")),
- envir = .state
-)
+assign("chat_session_messages", list(), envir = .state)
 
 api_url <- Sys.getenv("OPENAI_API_URL", "https://api.openai.com/v1")

R/get_chat_session.R

@@ -0,0 +1,16 @@
+#' Get Chat Session
+#'
+#' @param session_id The ID of the session to be used. If `NULL`, it will return an empty session.
+#'
+get_chat_session <- function(session_id = "1") {
+ default_session <- list(list(role = "system", content = "You are a helpful assistant."))
+ if (is.null(session_id)) {
+ return(default_session)
+ }
+ session <- get("chat_session_messages", envir = .state)[[as.character(session_id)]]
+ # If the session was not found, then it's a new (default) session.
+ if (is.null(session)) {
+ session <- default_session
+ }
+ session
+}

R/reset_chat_session.R

@@ -4,12 +4,22 @@
 #' we want to start a new conversation, we must call `reset_chat_session`.
 #'
 #' @param system_role ChatGPT's role as an AI assistant.
+#' @param session_id The ID of the session to be used. If `NULL`, this function will have no effect.
 #'
 #' @export
 #'
-reset_chat_session <- function(system_role = "You are a helpful assistant.") {
- assign(
- "chat_session_messages", list(list(role = "system", content = system_role)),
- envir = .state
- )
+reset_chat_session <- function(system_role = "You are a helpful assistant.", session_id = "1") {
+ if (is.null(session_id)) {
+ return()
+ }
+ if (is.list(system_role)) {
+ # If `system_role` is a list, then it is a ChatGPT session object.
+ session <- system_role
+ } else {
+ # Otherwise, it's a string specifying ChatGPT's role.
+ session <- list(list(role = "system", content = system_role))
+ }
+ all_sessions <- get("chat_session_messages", envir = .state)
+ all_sessions[[as.character(session_id)]] <- session
+ assign("chat_session_messages", all_sessions, envir = .state)
 }