runCodeInResponse and associated functions in

DESCRIPTION

@@ -42,5 +42,6 @@ Collate:
     'fileHeaderPrompt.R'
     'test-helper-test_argument_validation.R'
     'promptContext.R'
-RoxygenNote: 7.2.3
+    'runCodeInResponse.R'
+RoxygenNote: 7.3.0
 Config/testthat/edition: 3

NAMESPACE

@@ -8,6 +8,8 @@ export(extractFilenames)
 export(extractInstallPkg)
 export(fileHeaderPrompt)
 export(promptContext)
+export(runCodeInResponse)
+export(sampleResponse)
 export(selfcorrect)
 export(sendPrompt)
 export(setupAgent)

NEWS.md

@@ -4,6 +4,11 @@
     agent name as "generic" in setupAgent() function
   * new feature: clean_code_blocks() now automatically runs when extractCode() is called
   * new feature: sendPrompt() now can send previous message correspondence with LLM APIs via "previous.msgs" argument
+  * new feature: selfcorrect() has a new argument called responseWithErrors (default:NULL), via this arguments
+    users can input problematic responses to the selfcorrect() function instead of letting the getting the initial response and potentially failing.
+  * new function: runCodeInResponse() can take LLM response and directly run code, and tries
+    to fix if there are errors.
+  * new function: sampleResponse() function tries to fix code errors by repatedly sampling new solutions to the existing prompts
   * fix: typos in setupAgent fixed
   * fix: sendPrompt fixed which was broken due to openai API changes
   * fix: extractInstallPkg can now deal with require() calls in code blocks

R/runCodeInResponse.R

@@ -0,0 +1,130 @@
+#' Executes the code in the received response from the agent
+#'
+#' The function extracts and executes the code in the response. If required it can
+#' try to correct errors in the code via different strategies.
+#'
+#' @param response response to be parsed for code and executed
+#' @param prompt prompt for the response, if correction="none" it is not needed
+#' @param agent  AI agent, if correction="none" it is not needed
+#' @param context  context for the prompt, if correction="none" it is not needed
+#' @param correction "none" no code correction is needed. "selfcorrect" feedback the errors to LLM and ask for fix.
+#' "sampleMore" sample responses for the #prompt until an executable code is returned or number of attempts reached.
+#' "correctThenSample" first try self-correct "attempt" number of times.
+#' If no executable code is returned. It will sample new responses "attempt" number of times or until executable code
+#' @param attempts Numeric value denoting how many times the code should be sent back for fixing.
+#' @param output.file Optional output file created holding parsed code
+#' @param ... arguments to sendPrompt()
+#' @return A list containing the following elements:
+#' \item{init.response}{A character vector representing the initial prompt response.}
+#' \item{init.blocks}{A list of initial blocks.}
+#' \item{final.blocks}{A list of final blocks.}
+#' \item{code.works}{A boolean value indicating whether the code works.}
+#' \item{exec.result}{A character string representing the execution results.}
+#' \item{tried.attempts}{An integer representing the number of attempts.}
+#' @seealso \code{\link{selfcorrect}},\code{\link{sampleResponse}}
+#' @examples
+#' \dontrun{
+#'
+#' resp.list <- runCodeInResponse(agent,prompt,context=rbionfoExp,correction="sampleMore",attempt=2)
+#' }
+#' @export
+runCodeInResponse<-function(
+  response,
+  prompt=NULL,
+  agent=NULL,
+  context = NULL,
+  correction=c("none","selfcorrect","sampleMore","sampleThenCorrect","correctThenSample"),
+  attempts = 3,
+  output.file = NULL,
+  ...
+){
+
+  # check optional arguments
+  if(any( !is.null(prompt),!is.null(agent),!is.null(context))){
+
+    if( (is.null(prompt)+is.null(agent)+is.null(context))>1 ){
+      stop("\nif any of the following arguments is provided, all of them must be provided:\n",
+           "'prompt','agent' and 'context'\n",
+           "no NULL values are allowed in that case for those arguments"
+      )
+    }
+  }
+
+
+
+  # Clean the code backtick structure and install.packages calls
+  response<-clean_code_blocks(response)
+
+
+
+  # Parse the code
+  blocks <- extractCode(text=response,delimiter="```")
+
+  #execute response
+  if(is.null(output.file)){
+      res<-executeCode(blocks$code, output = "eval",output.file = output.file )
+
+  }else{
+      res<-executeCode(blocks$code, output = "html",output.file = output.file )
+
+  }
+
+  first.res<-list(init.response=response,
+                  init.blocks=blocks,
+                  final.response=response,
+                  final.blocks=blocks,
+                  code.works=!(is.list(res) & ("error" %in% names(res) )),
+                  exec.result=res,
+                  tried.attempts=1)
+
+  # if code works or no correction asked
+  if(correction=="none" | first.res$code.works){
+    return(first.res)
+
+  }else if(correction=="selfcorrect"){
+
+    res.list<-selfcorrect(agent,prompt,
+                       context=context,attempts,output.file,
+                       responseWithError=list(response,res),
+                       ...)
+
+    return(res.list)
+  }else if(correction=="sampleMore"){
+    res.list<-sampleResponse(agent,prompt,
+                        context=context,attempts,output.file,
+                        responseWithError=list(response,res),
+                        ...)
+    return(res.list)
+  }else if(correction=="sampleThenCorrect"){
+
+    res.list<-sampleResponse(agent,prompt,
+                        context=context,attempts,output.file,
+                        responseWithError=list(response,res),
+                        ...)
+    if(!res.list$code.works){
+      res.list<-selfcorrect(agent,prompt,
+                               context=context,attempts,output.file,
+                               responseWithError=list(response,res),
+                               ...)
+    }
+
+    return(res.list)
+  }else if(correction=="correctThenSample"){
+
+    res.list<-selfcorrect(agent,prompt,
+                          context=context,attempts,output.file,
+                          responseWithError=list(response,res),
+                          ...)
+
+    if(!res.list$code.works){
+      res.list<-sampleResponse(agent,prompt,
+                            context=context,attempts,output.file,
+                            responseWithError=list(response,res),
+                            ...)
+    }
+
+    return(res.list)
+  }
+
+
+}