diff --git a/DESCRIPTION b/DESCRIPTION
index a816ff5..9a55081 100644
--- a/DESCRIPTION
+++ b/DESCRIPTION
@@ -36,6 +36,7 @@ Imports:
     tokenizers,
     tools,
     urltools,
+    vctrs,
     xml2,
     yaml
 Suggests: 
diff --git a/NAMESPACE b/NAMESPACE
index e04da25..bb2bf46 100644
--- a/NAMESPACE
+++ b/NAMESPACE
@@ -1,8 +1,12 @@
 # Generated by roxygen2: do not edit by hand
 
+S3method(as_gpttools_message,default)
+S3method(print,chat_list)
+S3method(print,gpttools_message)
 export(add_roxygen_addin)
 export(addin_run_scrape_pkgs)
 export(addin_run_select_pkgs)
+export(as_gpttools_message)
 export(chat)
 export(chat_with_context)
 export(chat_with_retrieval)
@@ -20,6 +24,7 @@ export(get_selection)
 export(get_transformer_model)
 export(ghost_chat)
 export(ghost_writer)
+export(ghost_writer_addin)
 export(gpt_sitrep)
 export(gpttools_index_all_scraped_data)
 export(ingest_pdf)
@@ -41,11 +46,21 @@ export(script_to_function_addin)
 export(set_user_config)
 export(suggest_unit_test_addin)
 export(transcribe_audio)
+export(vec_cast.gpttools_message.data.frame)
+export(vec_ptype_abbr.gpttools_message)
+export(vec_ptype_full.gpttools_message)
 import(cli)
 import(httr2)
 import(rlang)
+import(stringr)
 importFrom(glue,glue)
 importFrom(graphics,text)
+importFrom(jsonlite,fromJSON)
+importFrom(purrr,compact)
+importFrom(purrr,map)
+importFrom(purrr,map_chr)
+importFrom(purrr,map_dfr)
+importFrom(purrr,pluck)
 importFrom(utils,globalVariables)
 importFrom(utils,head)
 importFrom(utils,installed.packages)
diff --git a/R/addin_copilot.R b/R/addin_copilot.R
index 39c01b6..fb37481 100755
--- a/R/addin_copilot.R
+++ b/R/addin_copilot.R
@@ -3,11 +3,11 @@
 #'
 #' @export
 copilot_addin <- function() {
-  cli::cli_alert_info("Attempting to add code suggestions")
+  cli_alert_info("Attempting to add code suggestions")
   ghost_chat(
     service = getOption("gpttools.service", "openai"),
     stream  = TRUE,
     where   = "source"
   )
-  cli::cli_alert_info("Done adding code suggestion")
+  cli_alert_info("Done adding code suggestion")
 }
diff --git a/R/chat.R b/R/chat.R
index 4394dc4..a224be5 100644
--- a/R/chat.R
+++ b/R/chat.R
@@ -240,7 +240,6 @@ get_cursor_context <- function(context_lines = 20,
   } else {
     file_ext <- doc$path |> tools::file_ext()
   }
-
   list(
     above    = context_above,
     below    = context_below,
diff --git a/R/class-message.R b/R/class-message.R
new file mode 100644
index 0000000..7898759
--- /dev/null
+++ b/R/class-message.R
@@ -0,0 +1,106 @@
+new_gpttools_message <- function(role = character(),
+                                 content = character(),
+                                 datetime = character(),
+                                 model = character(),
+                                 service = character(),
+                                 temperature = double()) {
+  if (!is_character(role)) cli_abort("`role` must be a character vector.")
+  if (!is_character(content)) cli_abort("`content` must be a character vector.")
+  if (!is_character(datetime, n = 1)) {
+    cli_abort("`datetime` must be a character vector of length 1.")
+  }
+  if (!is_character(model, n = 1)) {
+    cli_abort("`model` must be a character vector of length 1.")
+  }
+  if (!is_character(service, n = 1)) {
+    cli_abort("`service` must be a character vector of length 1")
+  }
+  if (!is_double(temperature, n = 1)) {
+    cli_abort("`temperature` must be a numeric vector of length 1.")
+  }
+
+  new_rcrd(
+    fields =
+      tibble::tibble(
+        role = role,
+        content = content,
+      ),
+    datetime = datetime,
+    model = model,
+    service = service,
+    temperature = temperature,
+    class = "gpttools_message"
+  )
+}
+
+gpttools_message <- function(role = character(),
+                             content = character(),
+                             datetime = character(),
+                             model = character(),
+                             service = character(),
+                             temperature = double()) {
+  role <- vec_cast(role, character())
+  content <- vec_cast(content, character())
+  datetime <- vec_cast(datetime, character())
+  model <- vec_cast(model, character())
+  service <- vec_cast(service, character())
+  temperature <- vec_cast(temperature, double())
+
+  new_gpttools_message(role, content, datetime, model, service, temperature)
+}
+
+#' @export
+print.gpttools_message <- function(x, ...) {
+  x_valid <- which(!is.na(x))
+
+  role <- field(x, "role")[x_valid]
+  content <- field(x, "content")[x_valid]
+  datetime <- attr(x, "datetime")
+  model <- attr(x, "model")
+  service <- attr(x, "service")
+  temperature <- attr(x, "temperature")
+
+  n <- length(field(x, "role"))
+  for (i in seq_len(n)) {
+    writeLines(col_silver(rule(stringr::str_to_title(role[i]))))
+    writeLines(content[i])
+  }
+  writeLines(rule("Settings", col = "blue"))
+  writeLines(col_blue(paste0("date: ", unique(datetime))))
+  writeLines(col_blue(paste0("model: ", unique(model))))
+  writeLines(col_blue(paste0("service: ", unique(service))))
+  writeLines(col_blue(paste0("temperature: ", unique(temperature))))
+  invisible(x)
+}
+
+#' @export
+vec_ptype_full.gpttools_message <- function(x, ...) "gpttools_message"
+
+#' @export
+vec_ptype_abbr.gpttools_message <- function(x, ...) "msg"
+
+#' @export
+as_gpttools_message <- function(x, ...) {
+  UseMethod("as_gpttools_message")
+}
+
+#' @export
+as_gpttools_message.default <- function(x, ...) {
+  vec_cast(x, "gpttools_message")
+}
+
+#' @export
+vec_cast.gpttools_message.data.frame <- function(x, to, ...) {
+  if (to == "data.frame") {
+    tibble::tibble(
+      role        = field(x, "role"),
+      content     = field(x, "content"),
+      datetime    = attr(x, "datetime"),
+      model       = attr(x, "model"),
+      service     = attr(x, "service"),
+      temperature = attr(x, "temperature")
+    )
+  } else {
+    cli_abort("Can't cast gpttools_message to ", to)
+  }
+}
diff --git a/R/document_data.R b/R/document_data.R
index e1fd8ba..90baabf 100644
--- a/R/document_data.R
+++ b/R/document_data.R
@@ -7,13 +7,13 @@
 #' @export
 collect_dataframes <- function() {
   objects <- names(rlang::global_env())
-  purrr::map_chr(
+  map_chr(
     .x = objects,
     .f = \(x) {
       if (is.data.frame(get(x))) x else NA
     }
   ) |>
-    purrr::compact() |>
+    compact() |>
     unlist()
 }
 
@@ -32,7 +32,7 @@ skim_lite <- function(data) {
 }
 
 collect_column_types <- function(data) {
-  purrr::map_dfr(
+  map_dfr(
     names(data),
     ~ data.frame(
       column = .x,
diff --git a/R/embedding.R b/R/embedding.R
index 705becc..6bcbcfb 100644
--- a/R/embedding.R
+++ b/R/embedding.R
@@ -34,7 +34,7 @@ prepare_scraped_files <- function(domain) {
 
   scraped |>
     dplyr::mutate(
-      chunks = purrr::map(text, \(x) {
+      chunks = map(text, \(x) {
         chunk_with_overlap(x,
           chunk_size = 500,
           overlap_size = 50,
@@ -97,7 +97,7 @@ add_embeddings <- function(index,
     model <- get_transformer_model()
     index |>
       dplyr::mutate(
-        embeddings = purrr::map(
+        embeddings = map(
           .x = chunks,
           .f = \(x) create_text_embeddings(x, model),
           .progress = "Creating Embeddings Locally"
@@ -108,7 +108,7 @@ add_embeddings <- function(index,
   } else {
     index |>
       dplyr::mutate(
-        embeddings = purrr::map(
+        embeddings = map(
           .x = chunks,
           .f = create_openai_embedding,
           .progress = "Create Embeddings"
@@ -227,7 +227,7 @@ get_top_matches <- function(index, query_embedding, k = 5) {
   index |>
     tibble::as_tibble() |>
     dplyr::mutate(
-      similarity = purrr::map_dbl(embedding, \(x) {
+      similarity = map_dbl(embedding, \(x) {
         lsa::cosine(query_embedding, unlist(x))
       })
     ) |>
@@ -262,6 +262,6 @@ chunk_with_overlap <- function(x, chunk_size, overlap_size, doc_id, ...) {
   } else {
     names(chunks) <- NULL
   }
-  chunks <- purrr::compact(chunks)
-  purrr::map(chunks, \(x) stringr::str_c(x, collapse = " "))
+  chunks <- compact(chunks)
+  map(chunks, \(x) stringr::str_c(x, collapse = " "))
 }
diff --git a/R/gpt-query.R b/R/gpt-query.R
index 258fb64..061f8ee 100644
--- a/R/gpt-query.R
+++ b/R/gpt-query.R
@@ -40,7 +40,7 @@ gpt_chat <- function(instructions,
   cli_inform("Model: {model}")
   cli_inform("Sending query... this can take up to 3 minutes.")
   simple_prompt <- prompt |>
-    purrr::map_chr(.f = "content") |>
+    map_chr(.f = "content") |>
     paste(collapse = "\n\n")
 
   answer <- chat(
diff --git a/R/gpttools-package.R b/R/gpttools-package.R
index daed30d..7f3032f 100644
--- a/R/gpttools-package.R
+++ b/R/gpttools-package.R
@@ -2,10 +2,12 @@
 "_PACKAGE"
 
 ## usethis namespace: start
-#' @import cli
 #' @import rlang
 #' @importFrom glue glue
 #' @import httr2
+#' @import cli
+#' @import stringr
+#' @importFrom purrr map_dfr map_chr pluck compact map
 #' @importFrom utils globalVariables head installed.packages old.packages
 #' packageDescription packageVersion
 #' @importFrom graphics text
diff --git a/R/harvest-docs.R b/R/harvest-docs.R
index 0894642..6e759c8 100644
--- a/R/harvest-docs.R
+++ b/R/harvest-docs.R
@@ -77,12 +77,12 @@ recursive_hyperlinks <- function(local_domain,
 
   expanded_urls <- c(expanded_urls, links)
   cli_inform(c("i" = "Total urls: {length(expanded_urls)}"))
-  links_df <- purrr::map(links, get_hyperlinks,
+  links_df <- map(links, get_hyperlinks,
     .progress = "Getting more links"
   ) |>
     dplyr::bind_rows() |>
     dplyr::filter(!stringr::str_detect(link, "^\\.$|mailto:|\\#|^\\_$")) |>
-    dplyr::mutate(link = purrr::map2_chr(
+    dplyr::mutate(link = map2_chr(
       .x = parent,
       .y = link,
       .f = \(x, y) xml2::url_absolute(y, x)
@@ -91,7 +91,7 @@ recursive_hyperlinks <- function(local_domain,
   cli_inform("Going to check {length(unique(links_df$link))} links")
 
   new_links <-
-    purrr::map(unique(links_df$link), \(x) {
+    map(unique(links_df$link), \(x) {
       if (rlang::is_true(stringr::str_detect(x, domain_pattern))) {
         validate_link(x)
       } else {
@@ -239,7 +239,7 @@ scrape_and_process <- function(url,
     unique()
   cli_inform(c("i" = "Scraping validated links"))
   scraped_data <-
-    purrr::map(links, \(x) {
+    map(links, \(x) {
       if (identical(check_url(x), 200L)) {
         tibble::tibble(
           source  = local_domain,
@@ -322,11 +322,11 @@ extract_text <- function(url, use_html_text2 = TRUE) {
   )
 
   xpath_tags <- exclude_tags |>
-    purrr::map_chr(.f = \(x) glue::glue("self::{x}")) |>
+    map_chr(.f = \(x) glue::glue("self::{x}")) |>
     stringr::str_c(collapse = " or ")
 
   xpath_attributes <- exclude_attributes |>
-    purrr::map_chr(.f = \(x) {
+    map_chr(.f = \(x) {
       glue::glue(
         "contains(concat(' ', normalize-space(@class), ' '), ' {x}')"
       )
@@ -336,7 +336,7 @@ extract_text <- function(url, use_html_text2 = TRUE) {
   # Handling general attribute selectors
   general_attributes <- c("role", "aria-", "data-", "id", "class", "style")
   xpath_general_attributes <- general_attributes |>
-    purrr::map_chr(.f = \(x) glue::glue("@{x}")) |>
+    map_chr(.f = \(x) glue::glue("@{x}")) |>
     stringr::str_c(collapse = " or ")
 
   xpath_combined <- xpath_combined <- glue::glue(
diff --git a/R/history.R b/R/history.R
index 8abfad9..35098c3 100644
--- a/R/history.R
+++ b/R/history.R
@@ -58,7 +58,7 @@ delete_history <- function(local = FALSE) {
   }
   history_files <- get_history_path(local = local)
 
-  purrr::map(history_files, \(x) {
+  map(history_files, \(x) {
     delete_file <- ui_yeah("Do you want to delete {basename(x)}?")
     if (delete_file) {
       file.remove(x)
@@ -246,7 +246,7 @@ chat_with_context <- function(query,
     context <-
       full_context |>
       dplyr::select(source, link, chunks) |>
-      purrr::pmap(\(source, link, chunks) {
+      pmap(\(source, link, chunks) {
         glue::glue("Source: {source}
                    Link: {link}
                    Text: {chunks}")
@@ -331,7 +331,7 @@ chat_with_context <- function(query,
   )
 
   session_history <-
-    purrr::map(session_history, \(x) {
+    map(session_history, \(x) {
       if (x$role == "system") {
         NULL
       } else if (stringr::str_detect(
@@ -344,7 +344,7 @@ chat_with_context <- function(query,
         x
       }
     }) |>
-    purrr::compact()
+    compact()
 
   prompt <- c(
     session_history,
@@ -354,7 +354,7 @@ chat_with_context <- function(query,
   )
 
   simple_prompt <- prompt |>
-    purrr::map_chr(.f = "content") |>
+    map_chr(.f = "content") |>
     paste(collapse = "\n\n")
 
   cli_alert_info("Service: {service}")
@@ -390,7 +390,7 @@ chat_with_context <- function(query,
   }
 
   if (save_history) {
-    purrr::map(prompt, \(x) {
+    map(prompt, \(x) {
       save_user_history(
         file_name = history_name,
         role      = "system",
diff --git a/R/index.R b/R/index.R
index 6ad6415..f342233 100755
--- a/R/index.R
+++ b/R/index.R
@@ -37,7 +37,7 @@ gpttools_index_all_scraped_data <- function(overwrite = FALSE,
                                             dont_ask = TRUE) {
   text_files <- list_index("text", full_path = TRUE)
 
-  purrr::walk(text_files, function(file_path) {
+  walk(text_files, function(file_path) {
     domain <- tools::file_path_sans_ext(basename(file_path))
     cli_alert_info(glue("Creating/updating index for domain {domain}..."))
     create_index(
diff --git a/R/site-index.R b/R/site-index.R
index a6336c0..b5230c4 100644
--- a/R/site-index.R
+++ b/R/site-index.R
@@ -56,7 +56,7 @@ get_pkgs_to_scrape <- function(local = TRUE,
       installed_version = Version
     ) |>
     dplyr::filter(name %in% pkgs) |>
-    dplyr::mutate(url = purrr::map_chr(name, get_pkg_doc_page)) |>
+    dplyr::mutate(url = map_chr(name, get_pkg_doc_page)) |>
     tidyr::drop_na(url) |>
     dplyr::mutate(source = urltools::domain(url)) |>
     dplyr::left_join(get_outdated_pkgs(), by = "name") |>
@@ -140,7 +140,7 @@ scrape_pkg_sites <- function(sites = get_pkgs_to_scrape(local = TRUE),
   } else {
     sites |>
       dplyr::select(url, version, name) |>
-      purrr::pmap(.f = \(url, version, name) {
+      pmap(.f = \(url, version, name) {
         crawl(
           url = url,
           index_create = index_create,
diff --git a/R/stream-anthropic.R b/R/stream-anthropic.R
index 6b69f4a..14cad1d 100644
--- a/R/stream-anthropic.R
+++ b/R/stream-anthropic.R
@@ -23,7 +23,8 @@ stream_chat_anthropic <- function(prompt,
     req_error(is_error = function(resp) FALSE) |>
     req_perform_stream(
       callback = element_callback,
-      buffer_kb = 0.01
+      buffer_kb = 0.01,
+      round = "line"
     )
 
   # error handling
@@ -31,9 +32,10 @@ stream_chat_anthropic <- function(prompt,
     status <- resp_status(response)
     description <- resp_status_desc(response)
 
-    cli::cli_abort(message = c(
+    cli_abort(message = c(
       "x" = "Anthropic API request failed. Error {status} - {description}",
       "i" = "Visit the Anthropic API documentation for more details"
     ))
   }
+  invisible(response)
 }
diff --git a/R/stream-azure-openai.R b/R/stream-azure-openai.R
index ab9ec1e..a34af48 100755
--- a/R/stream-azure-openai.R
+++ b/R/stream-azure-openai.R
@@ -38,7 +38,8 @@ stream_chat_azure_openai <- function(prompt = NULL,
     req_error(is_error = function(resp) FALSE) |>
     req_perform_stream(
       callback = element_callback,
-      buffer_kb = 0.01
+      buffer_kb = 0.01,
+      round = "line"
     )
 
   invisible(response)
diff --git a/R/stream-chat.R b/R/stream-chat.R
index b4de647..8035057 100644
--- a/R/stream-chat.R
+++ b/R/stream-chat.R
@@ -41,15 +41,16 @@ stream_chat <- function(prompt,
       )
     }
   )
+  invisible(response)
 }
 
 create_handler <- function(service = "openai",
                            r = NULL,
                            output_id = "streaming",
-                           where = "console") {
-  env <- rlang::env()
+                           where = "console",
+                           env = caller_env()) {
   env$resp <- NULL
-  env$full_resp <- NULL
+  env$full_resp <- ""
 
   stream_details <- get_stream_pattern(service)
   new_pattern <- stream_details$pattern
@@ -73,7 +74,7 @@ create_handler <- function(service = "openai",
     if (stringr::str_detect(env$resp, pattern)) {
       parsed <- stringr::str_extract(env$resp, pattern) |>
         jsonlite::fromJSON() |>
-        purrr::pluck(!!!new_pluck)
+        pluck(!!!new_pluck)
       env$full_resp <- paste0(env$full_resp, parsed)
 
       if (where == "shiny") {
@@ -121,8 +122,8 @@ get_stream_pattern <- function(service) {
       pluck <- "text"
     },
     "ollama" = {
-      pattern <- '\\{"model":.*"done":false\\}'
-      pluck <- "response"
+      pattern <- '\\{"id":.*?\\}\\]\\}'
+      pluck <- c("choices", "delta", "content")
     },
     "azure_openai" = {
       pattern <- '\\{"id":.*?\\}\\]\\}'
diff --git a/R/stream-cohere.R b/R/stream-cohere.R
index 4b5cd5f..48b8ee8 100644
--- a/R/stream-cohere.R
+++ b/R/stream-cohere.R
@@ -1,5 +1,5 @@
 stream_chat_cohere <- function(prompt,
-                               model = getOption("gpttools.model", "command"),
+                               model = "command",
                                element_callback = create_handler("cohere"),
                                key = Sys.getenv("COHERE_API_KEY")) {
   request_body <- list(
@@ -15,19 +15,23 @@ stream_chat_cohere <- function(prompt,
       `Authorization` = paste("Bearer", key),
       `content-type` = "application/json"
     ) |>
-    req_method("POST") |>
     req_body_json(data = request_body) |>
     req_retry(max_tries = 3) |>
     req_error(is_error = function(resp) FALSE) |>
-    req_perform_stream(callback = element_callback, buffer_kb = 0.01)
+    req_perform_stream(
+      callback = element_callback,
+      buffer_kb = 0.01,
+      round = "line"
+    )
 
   if (resp_is_error(response)) {
     status <- resp_status(response)
     description <- resp_status_desc(response)
 
-    cli::cli_abort(message = c(
-      "x" = glue::glue("Cohere API request failed. Error {status} - {description}"),
+    cli_abort(c(
+      "x" = glue("Cohere API request failed. Error {status} - {description}"),
       "i" = "Visit the Cohere API documentation for more details"
     ))
   }
+  invisible(response)
 }
diff --git a/R/stream-copilot.R b/R/stream-copilot.R
new file mode 100644
index 0000000..cb7ac64
--- /dev/null
+++ b/R/stream-copilot.R
@@ -0,0 +1,29 @@
+get_copilot_oauth <- function(dir = "~/.config/github-copilot") {
+  hosts <- jsonlite::read_json(file.path(dir, "hosts.json"))
+  hosts[[1]]$oauth_token
+}
+
+chat_copilot <- function() {
+  token <- request("https://api.github.com/copilot_internal/v2/token") |>
+    req_auth_bearer_token(oauth_token) |>
+    req_perform() |>
+    resp_body_json() |>
+    pluck("token")
+
+  request("https://api.githubcopilot.com/chat/completions") |>
+    req_auth_bearer_token(token = token) |>
+    req_headers("Editor-Version" = "vscode/9.9.9") |>
+    req_body_json(
+      data = list(
+        messages = list(
+          list(
+            role = "user",
+            content = "Tell me a joke about the R language."
+          )
+        ),
+        model = "copilot"
+      )
+    ) |>
+    req_perform() |>
+    resp_chat_openai(stream = FALSE)
+}
diff --git a/R/stream-ollama.R b/R/stream-ollama.R
index 3fce7ec..d28b2e7 100644
--- a/R/stream-ollama.R
+++ b/R/stream-ollama.R
@@ -15,17 +15,22 @@ stream_chat_ollama <- function(prompt,
     req_url_path_append("api") |>
     req_url_path_append("generate") |>
     req_body_json(data = body) |>
-    req_perform_stream(callback = element_callback, buffer_kb = 0.01)
+    req_perform_stream(
+      callback = element_callback,
+      buffer_kb = 0.01,
+      round = "line"
+    )
 
   if (resp_is_error(response)) {
     status <- resp_status(response)
     description <- resp_status_desc(response)
 
-    cli::cli_abort(message = c(
+    cli_abort(message = c(
       "x" = glue::glue("Ollama API request failed. Error {status} - {description}"),
       "i" = "Visit the Ollama API documentation for more details"
     ))
   }
+  invisible(response)
 }
 
 ollama_is_available <- function(verbose = FALSE) {
@@ -39,14 +44,14 @@ ollama_is_available <- function(verbose = FALSE) {
       response <- req_perform(request) |>
         resp_body_string()
 
-      if (verbose) cli::cli_alert_success(response)
+      if (verbose) cli_alert_success(response)
       check_value <- TRUE
     },
     error = function(cnd) {
       if (inherits(cnd, "httr2_failure")) {
-        if (verbose) cli::cli_alert_danger("Couldn't connect to Ollama in {.url {ollama_api_url()}}. Is it running there?")
+        if (verbose) cli_alert_danger("Couldn't connect to Ollama in {.url {ollama_api_url()}}. Is it running there?")
       } else {
-        if (verbose) cli::cli_alert_danger(cnd)
+        if (verbose) cli_alert_danger(cnd)
       }
       check_value <- FALSE
     }
diff --git a/R/stream-openai.R b/R/stream-openai.R
index 231bc6e..66c7d0c 100644
--- a/R/stream-openai.R
+++ b/R/stream-openai.R
@@ -1,32 +1,171 @@
-stream_chat_openai <- function(prompt = NULL,
-                               element_callback = create_handler("openai"),
-                               model = getOption("gpttools.model", "gpt-4-turbo-preview"),
-                               openai_api_key = Sys.getenv("OPENAI_API_KEY"),
-                               shiny = FALSE) {
-  messages <- list(
+chat_openai <- function(prompt = "Tell me a joke about the R language.",
+                        model = "gpt-3.5-turbo",
+                        history = NULL,
+                        temperature = NULL,
+                        stream = FALSE) {
+  prompt <- prompt |> add_history(history)
+
+  response <-
+    req_chat_openai(
+      prompt = prompt,
+      model = model,
+      temperature = temperature,
+      stream = is_true(stream)
+    ) |>
+    resp_chat_openai(stream = is_true(stream))
+
+  response <- c(
+    prompt,
+    list(list(role = response$role, content = response$content))
+  )
+
+  class(response) <- c("chat_list", class(response))
+
+  response
+}
+
+#' @export
+print.chat_list <- function(x, ...) {
+  n <- length(x)
+
+  writeLines("\n")
+
+  for (i in seq_len(n)) {
+    print_role <- rule(stringr::str_to_title(x[[i]]$role))
+    print_role <-
+      switch(x[[i]]$role,
+        "assistant" = col_green(print_role),
+        "system"    = col_silver(print_role),
+        "user"      = col_blue(print_role)
+      )
+    writeLines(print_role)
+    writeLines(x[[i]]$content)
+  }
+
+  writeLines("\n")
+  invisible(x)
+}
+
+
+# Make API Request --------------------------------------------------------
+
+req_base_openai <- function(
+    url = getOption("gpttools.url", "https://api.openai.com/")) {
+  request(url) |>
+    req_url_path_append("v1", "chat", "completions")
+}
+
+req_auth_openai <- function(request) {
+  request |> req_auth_bearer_token(token = Sys.getenv("OPENAI_API_KEY"))
+}
+
+req_body_openai <- function(request,
+                            prompt = "Tell me a joke about the R language.",
+                            model = "gpt-4-turbo-preview",
+                            history = NULL,
+                            temperature = 0.7,
+                            stream = FALSE) {
+  body <-
     list(
-      role = "user",
-      content = prompt
+      model = model,
+      messages = prompt,
+      temperature = temperature,
+      stream = is_true(stream)
     )
-  )
 
-  # Set the request body
-  body <- list(
-    model = model,
-    stream = TRUE,
-    messages = messages
-  )
+  request |>
+    req_body_json(data = body)
+}
 
-  response <-
-    request("https://api.openai.com/v1/chat/completions") |>
-    req_auth_bearer_token(token = openai_api_key) |>
-    req_body_json(data = body) |>
+
+req_chat_openai <- function(prompt, model, temperature, stream = FALSE) {
+  req <-
+    req_base_openai() |>
+    req_auth_openai() |>
+    req_body_openai(
+      prompt = prompt,
+      model = model,
+      temperature = temperature,
+      stream = is_true(stream)
+    ) |>
     req_retry(max_tries = 3) |>
-    req_error(is_error = function(resp) FALSE) |>
-    req_perform_stream(
-      callback = element_callback,
-      buffer_kb = 0.01
-    )
+    req_error(is_error = function(resp) FALSE)
+
+  if (is_true(stream)) {
+    env <- caller_env()
+    req |>
+      req_perform_stream(
+        callback = \(x) stream_callback_openai(x, env),
+        buffer_kb = 0.01,
+        round = "line"
+      )
+    tibble::tibble(role = "assistant", content = env$response)
+  } else {
+    req |>
+      req_perform()
+  }
+}
+
+#' @importFrom jsonlite fromJSON
+stream_callback_openai <- function(x, env) {
+  txt <- rawToChar(x)
 
-  invisible(response)
+  lines <- str_split(txt, "\n")[[1]]
+  lines <- lines[lines != ""]
+  lines <- str_replace_all(lines, "^data: ", "")
+  lines <- lines[!str_detect(lines, "\"finish_reason\":\"stop\"")]
+  lines <- lines[lines != "[DONE]"]
+
+  tokens <- map_chr(lines, \(line) {
+    chunk <- jsonlite::fromJSON(line)
+    chunk$choices$delta$content
+  })
+
+  env$response <- paste0(env$response, tokens)
+
+  cat(tokens)
+
+  TRUE
+}
+
+# Process API Response ----------------------------------------------------
+
+resp_chat_openai <- function(response, stream) {
+  resp <- response
+
+  if (is_true(stream)) {
+    resp
+  } else {
+    resp |>
+      resp_chat_error_openai() |>
+      resp_body_json(simplifyVector = TRUE) |>
+      pluck("choices", "message")
+  }
+}
+
+resp_chat_error_openai <- function(response) {
+  if (resp_is_error(response)) {
+    status <- resp_status(response)
+    description <- resp_status_desc(response)
+
+    cli_abort(c(
+      "x" = glue("OpenAI API request failed. Error {status} - {description}"),
+      "i" = "Visit the OpenAI API documentation for more details"
+    ))
+  } else {
+    invisible(response)
+  }
+}
+
+add_history <- function(prompt, history = NULL) {
+  c(
+    history,
+    list(
+      list(
+        role = "user",
+        content = prompt
+      )
+    )
+  ) |>
+    purrr::compact()
 }
diff --git a/R/stream-perplexity.R b/R/stream-perplexity.R
index 8364a3c..8a96fe2 100644
--- a/R/stream-perplexity.R
+++ b/R/stream-perplexity.R
@@ -1,6 +1,6 @@
 stream_chat_perplexity <- function(prompt,
                                    element_callback = create_handler("perplexity"),
-                                   model = getOption("gpttools.model", "sonar-small-chat"),
+                                   model = "sonar-small-chat",
                                    api_key = Sys.getenv("PERPLEXITY_API_KEY")) {
   request_body <- list(
     model = model,
@@ -19,11 +19,19 @@ stream_chat_perplexity <- function(prompt,
     ) |>
     req_body_json(data = request_body) |>
     req_retry(max_tries = 3) |>
-    req_perform_stream(callback = element_callback, buffer_kb = 0.01)
+    req_perform_stream(
+      callback = element_callback,
+      buffer_kb = 0.01,
+      round = "line"
+    )
 
   if (resp_is_error(response)) {
     status <- resp_status(response)
     description <- resp_status_desc(response)
-    stop("Perplexity API request failed with error ", status, ": ", description, call. = FALSE)
+
+    cli_abort(message = c(
+      "x" = glue::glue("Perplexity API request failed. Error {status} - {description}"),
+      "i" = "Visit the Perplexity API documentation for more details"
+    ))
   }
 }
diff --git a/R/transcribe.R b/R/transcribe.R
index 89bbf9e..826c697 100644
--- a/R/transcribe.R
+++ b/R/transcribe.R
@@ -83,7 +83,7 @@ transcribe_audio <- function(file_path,
                              link = NA,
                              prompt = NA, chunk_size = 120) {
   audio_chunks <- split_audio(file_path = file_path, duration_secs = chunk_size)
-  purrr::map(audio_chunks, \(x) {
+  map(audio_chunks, \(x) {
     tibble::tibble(
       source = source,
       text = transcribe_audio_chunk(audio_file = x, prompt = prompt),
@@ -135,7 +135,7 @@ create_index_from_audio <- function(file_path,
 #' @export
 create_transcript <- function(file_path, prompt = NULL, chunk_size = 120) {
   split_audio(file_path = file_path, duration_secs = chunk_size) |>
-    purrr::map(\(x) {
+    map(\(x) {
       transcribed_text <- transcribe_audio_chunk(
         audio_file = x,
         prompt = prompt
diff --git a/inst/retriever/app.R b/inst/retriever/app.R
index 6a29803..7992a03 100644
--- a/inst/retriever/app.R
+++ b/inst/retriever/app.R
@@ -12,6 +12,7 @@ library(bslib)
 library(bsicons)
 library(waiter)
 library(reprex)
+library(purrr)
 
 window_height_ui <- function(id) {
   ns <- NS(id)
@@ -45,14 +46,14 @@ window_height_server <- function(id) {
 
 make_chat_history <- function(chats) {
   history <-
-    purrr::discard(chats, \(x) x$role == "system") |>
-    purrr::map(\(x) {
+    discard(chats, \(x) x$role == "system") |>
+    map(\(x) {
       list(
         strong(stringr::str_to_title(x$role)),
         markdown(x$content)
       )
     }) |>
-    purrr::list_flatten()
+    list_flatten()
   history
 }
 
@@ -61,7 +62,7 @@ api_services <-
   stringr::str_remove(
     pattern = "gptstudio_request_perform.gptstudio_request_"
   ) |>
-  purrr::discard(~ .x == "gptstudio_request_perform.default")
+  discard(~ .x == "gptstudio_request_perform.default")
 
 ui <- page_fillable(
   useWaiter(),
@@ -255,7 +256,7 @@ server <- function(input, output, session) {
       if ("All" %in% input$source) {
         load_index(domain = "All", local_embeddings = TRUE)
       } else {
-        purrr::map(input$source, \(x) {
+        map(input$source, \(x) {
           load_index(x, local_embeddings = TRUE) |>
             tibble::as_tibble()
         }) |>
@@ -264,7 +265,7 @@ server <- function(input, output, session) {
     } else if ("All" %in% input$source) {
       load_index(domain = "All", local_embeddings = FALSE)
     } else {
-      purrr::map(input$source, \(x) {
+      map(input$source, \(x) {
         load_index(x, local_embeddings = FALSE) |>
           tibble::as_tibble()
       }) |>
diff --git a/inst/scripts/scrape_other_docs.R b/inst/scripts/scrape_other_docs.R
index 626010f..289601e 100644
--- a/inst/scripts/scrape_other_docs.R
+++ b/inst/scripts/scrape_other_docs.R
@@ -51,7 +51,7 @@ scrape_resources <- function(resources) {
 
   sites |>
     dplyr::select(url, name) |>
-    purrr::pmap(.f = \(url, name) {
+    pmap(.f = \(url, name) {
       crawl(
         url = url,
         index_create = TRUE,
diff --git a/inst/settings/app.R b/inst/settings/app.R
index 218896f..7d86167 100755
--- a/inst/settings/app.R
+++ b/inst/settings/app.R
@@ -9,7 +9,7 @@ api_services <-
   stringr::str_remove(
     pattern = "gptstudio_request_perform.gptstudio_request_"
   ) |>
-  purrr::discard(~ .x == "gptstudio_request_perform.default")
+  discard(~ .x == "gptstudio_request_perform.default")
 
 # Define UI
 ui <- page_fillable(
diff --git a/man/ghost_writer_addin.Rd b/man/ghost_writer_addin.Rd
new file mode 100644
index 0000000..2c206e9
--- /dev/null
+++ b/man/ghost_writer_addin.Rd
@@ -0,0 +1,11 @@
+% Generated by roxygen2: do not edit by hand
+% Please edit documentation in R/addin_ghost_writer.R
+\name{ghost_writer_addin}
+\alias{ghost_writer_addin}
+\title{Writing suggestions}
+\usage{
+ghost_writer_addin()
+}
+\description{
+Writing suggestions
+}