forked from MichelNivard/gptstudio
-
Notifications
You must be signed in to change notification settings - Fork 0
Commit
This commit does not belong to any branch on this repository, and may belong to a fork outside of the repository.
Merge pull request MichelNivard#91 from calderonsamuel/httr2
Migrate internals to httr2 and handle chat completion stream
- Loading branch information
Showing
49 changed files
with
1,324 additions
and
402 deletions.
There are no files selected for viewing
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
| Original file line number | Diff line number | Diff line change |
|---|---|---|
| @@ -0,0 +1,162 @@ | ||
| #' Stream handler for chat completions | ||
| #' | ||
| #' R6 class that allows to handle chat completions chunk by chunk. | ||
| #' It also adds methods to retrieve relevant data. This class DOES NOT make the request. | ||
| #' | ||
| #' Because `curl::curl_fetch_stream` blocks the R console until the stream finishes, | ||
| #' this class can take a shiny session object to handle communication with JS | ||
| #' without recurring to a `shiny::observe` inside a module server. | ||
| #' | ||
| #' @param session The shiny session it will send the message to (optional). | ||
| #' @param user_prompt The prompt for the chat completion. Only to be displayed in an HTML tag containing the prompt. (Optional). | ||
| #' @importFrom rlang %||% | ||
| #' @importFrom magrittr %>% | ||
| #' @importFrom R6 R6Class | ||
| #' @importFrom stringr str_remove str_split_1 | ||
| #' @importFrom purrr map_chr map | ||
| #' @importFrom jsonlite fromJSON | ||
| StreamHandler <- R6::R6Class( | ||
| classname = "StreamHandler", | ||
| public = list( | ||
|
|
||
| #' @field current_value The content of the stream. It updates constantly until the stream ends. | ||
| current_value = NULL, | ||
|
|
||
| #' @field chunks The list of chunks streamed. It updates constantly until the stream ends. | ||
| chunks = list(), | ||
|
|
||
| #' @field shinySession Holds the `session` provided at initialization | ||
| shinySession = NULL, | ||
|
|
||
| #' @field user_message The `user_prompt` provided at initialization after being formatted with markdown. | ||
| user_message = NULL, | ||
|
|
||
| #' @description Start a StreamHandler. Recommended to be assigned to the `stream_handler` name. | ||
| initialize = function(session = NULL, user_prompt = NULL) { | ||
| self$current_value <- "" | ||
| self$shinySession <- session | ||
| self$user_message <- shiny::markdown(user_prompt) | ||
| }, | ||
|
|
||
| #' @description The main reason this class exists. It reduces to stream to chunks and its current value. If the object finds a shiny session will send a `render-stream` message to JS. | ||
| #' @param x The streamed element. Preferably after conversion from raw. | ||
| handle_streamed_element = function(x) { | ||
| translated <- private$translate_element(x) | ||
| self$chunks <- c(self$chunks, translated) | ||
| self$current_value <- private$convert_chunks_into_response_str() | ||
|
|
||
| if (!is.null(self$shinySession)) { | ||
| # any communication with JS should be handled here!! | ||
| self$shinySession$sendCustomMessage( | ||
| type = "render-stream", | ||
| message = list( | ||
| user = self$user_message, | ||
| assistant = shiny::markdown(self$current_value) | ||
| ) | ||
| ) | ||
| } | ||
| }, | ||
|
|
||
| #' @description Extract the message content as a message ready to be styled or appended to the chat history. Useful after the stream ends. | ||
| extract_message = function() { | ||
| list( | ||
| role = "assistant", | ||
| content = self$current_value | ||
| ) | ||
| } | ||
| ), | ||
| private = list( | ||
| # Translates a streamed element and converts it to chunk. | ||
| # Also handles the case of multiple elements in a single stream. | ||
| translate_element = function(x) { | ||
| x %>% | ||
| stringr::str_remove("^data: ") %>% # handle first element | ||
| stringr::str_remove("(\n\ndata: \\[DONE\\])?\n\n$") %>% # handle last element | ||
| stringr::str_split_1("\n\ndata: ") %>% | ||
| purrr::map(\(x) jsonlite::fromJSON(x, simplifyVector = FALSE)) | ||
| }, | ||
| # Reduces the chuks into just the message content. | ||
| convert_chunks_into_response_str = function() { | ||
| self$chunks %>% | ||
| purrr::map_chr(~ .x$choices[[1]]$delta$content %||% "") %>% | ||
| paste0(collapse = "") | ||
| } | ||
| ) | ||
| ) | ||
|
|
||
| #' Stream Chat Completion | ||
| #' | ||
| #' This function sends a prompt to the OpenAI API for chat-based completion and retrieves the streamed response. | ||
| #' | ||
| #' @param prompt The user's message or prompt. | ||
| #' @param history A list of previous messages in the conversation (optional). | ||
| #' @param element_callback A callback function to handle each element of the streamed response (optional). | ||
| #' @param style The style of the chat conversation (optional). Default is retrieved from the "gptstudio.code_style" option. | ||
| #' @param skill The skill to use for the chat conversation (optional). Default is retrieved from the "gptstudio.skill" option. | ||
| #' @param model The model to use for chat completion (optional). Default is "gpt-3.5-turbo". | ||
| #' @param openai_api_key The OpenAI API key (optional). By default, it is fetched from the "OPENAI_API_KEY" environment variable. | ||
| #' | ||
| #' @return the same as `curl::curl_fetch_stream` | ||
| #' | ||
| stream_chat_completion <- | ||
| function(prompt, | ||
| history = NULL, | ||
| element_callback = cat, | ||
| style = getOption("gptstudio.code_style"), | ||
| skill = getOption("gptstudio.skill"), | ||
| model = "gpt-3.5-turbo", | ||
| openai_api_key = Sys.getenv("OPENAI_API_KEY")) { | ||
| # Set the API endpoint URL | ||
| url <- "https://api.openai.com/v1/chat/completions" | ||
|
|
||
| # Set the request headers | ||
| headers <- list( | ||
| "Content-Type" = "application/json", | ||
| "Authorization" = paste0("Bearer ", openai_api_key) | ||
| ) | ||
|
|
||
| # Set the new chat history so the system prompt depends | ||
| # on the current parameters and not in previous ones | ||
| instructions <- list( | ||
| list( | ||
| role = "system", | ||
| content = chat_create_system_prompt(style, skill, in_source = FALSE) | ||
| ), | ||
| list( | ||
| role = "user", | ||
| content = prompt | ||
| ) | ||
| ) | ||
|
|
||
| history <- purrr::discard(history, ~ .x$role == "system") | ||
|
|
||
| messages <- c(history, instructions) | ||
|
|
||
| # Set the request body | ||
| body <- list( | ||
| "model" = model, | ||
| "stream" = TRUE, | ||
| "messages" = messages | ||
| ) | ||
|
|
||
| # Create a new curl handle object | ||
| handle <- curl::new_handle() %>% | ||
| curl::handle_setheaders(.list = headers) %>% | ||
| curl::handle_setopt(postfields = jsonlite::toJSON(body, auto_unbox = TRUE)) # request body | ||
|
|
||
|
|
||
| # Make the streaming request using curl_fetch_stream() | ||
| curl::curl_fetch_stream( | ||
| url = url, | ||
| fun = \(x) { | ||
| element <- rawToChar(x) | ||
| element_callback(element) # Do whatever element_callback does | ||
| }, | ||
| handle = handle | ||
| ) | ||
| } | ||
|
|
||
| # stream_handler <- StreamHandler$new() | ||
|
|
||
| # stream_chat_completion(messages = "Count from 1 to 10") | ||
| # stream_chat_completion(messages = "Count from 1 to 10", element_callback = stream_handler$handle_streamed_element) |
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Oops, something went wrong.