Fix parsing in filter afforder

afforder/afforderstrategy/filterafforder/filter_afforder.go

@@ -43,24 +43,6 @@ func (fa *FilterAfforder) filterBrowserDisplay(ctx context.Context, br *browser.
  rawBrowserDisplay := br.GetDisplay().MD
  numberedBrowserDisplay := displayBrowserContentWithLineno(rawBrowserDisplay)
  trajDisplay := traj.GetAbbreviatedText()
- messages := []*llm.Message{
- {
- Role: llm.MessageRoleSystem,
- Content: systemPromptToFilterAffordances,
- },
- {
- Role: llm.MessageRoleUser,
- Content: fmt.Sprintf(`----- START BROWSER -----
-%s
------ END BROWSER -----
-
------ START TRAJECTORY -----
-%s
------ END TRAJECTORY -----
-
-First, list a description of the next action that should be taken. Then, list a sequence of the irrelevant lines. These lines will be deleted and the remaining lines will be displayed as the web browser for your next action.`, numberedBrowserDisplay, trajDisplay),
- },
- }
  functionDef := &llm.FunctionDef{
  Name: "filter_irrelevant_lines",
  Description: "Filter out the irrelevant lines from the browser display. The remaining lines will be displayed as the web browser for your next action.",
@@ -82,17 +64,35 @@ First, list a description of the next action that should be taken. Then, list a
  Required: []string{"next_action_description", "irrelevant_lines"},
  },
  }
+ messages := []*llm.Message{
+ {
+ Role: llm.MessageRoleSystem,
+ Content: systemPromptToFilterAffordances,
+ },
+ {
+ Role: llm.MessageRoleUser,
+ Content: fmt.Sprintf(`----- START BROWSER -----
+%s
+----- END BROWSER -----
+
+----- START TRAJECTORY -----
+%s
+----- END TRAJECTORY -----
+
+First, list a description of the next action that should be taken. Then, list a sequence of the irrelevant lines. These lines will be deleted and the remaining lines will be displayed as the web browser for your next action. Use the %s function call.`, numberedBrowserDisplay, trajDisplay, functionDef.Name),
+ },
+ }
  var args map[string]interface{}
  if res, err := fa.models.ChatModels[llm.ChatModelGPT4].Message(ctx, messages, &llm.MessageOptions{
  Temperature: 0.0,
  Functions: []*llm.FunctionDef{functionDef},
- FunctionCall: "filter_affordances",
+ FunctionCall: functionDef.Name,
  }); err != nil {
  return "", fmt.Errorf("failed to get response from chat model: %w", err)
  } else if res.FunctionCall == nil {
  return "", fmt.Errorf("response from chat model did not include a function call")
- } else if res.FunctionCall.Name != "filter_affordances" {
- return "", fmt.Errorf("response from chat model did not include a function call to filter_affordances")
+ } else if res.FunctionCall.Name != functionDef.Name {
+ return "", fmt.Errorf("response from chat model did not include a function call to %s", functionDef.Name)
  } else if err := json.Unmarshal([]byte(res.FunctionCall.Arguments), &args); err != nil {
  return "", fmt.Errorf("failed to unmarshal arguments from chat model response: %w", err)
  } else if irrelevantLinesRes, ok := args["irrelevant_lines"].([]interface{}); !ok {
@@ -124,7 +124,7 @@ func displayBrowserContentWithLineno(s string) string {
 
 func parseIrrelevantLinesResult(res []string) (map[int]struct{}, error) {
  irrelevantLines := make(map[int]struct{})
- re := regexp.MustCompile(`^(\d+(:|-)\d+|:\d+|\d+:|<\d+>) description="(.+)"$`)
+ re := regexp.MustCompile(`^(\d+(:|-)\d+|:\d+|\d+:|<\d+>|\d+) description=['"]?(.+)['"]?$`)
  for _, line := range res {
  matches := re.FindStringSubmatch(line)
  if matches == nil {

afforder/strategy.go

@@ -11,8 +11,8 @@ import (
 type AfforderStrategyID string
 
 const (
- AfforderStrategyIDFunctionAfforder AfforderStrategyID = "function_afforder"
- AfforderStrategyIDFilterAfforder AfforderStrategyID = "filter_afforder"
+ AfforderStrategyIDFunctionAfforder AfforderStrategyID = "function"
+ AfforderStrategyIDFilterAfforder AfforderStrategyID = "filter"
 )
 
 const DefaultAfforderStrategyID = AfforderStrategyIDFunctionAfforder

browser/browser.go

@@ -57,12 +57,12 @@ func (b *Browser) updateDisplay() error {
  return fmt.Errorf("error getting location: %w", err)
  } else if html, err := b.getHTML(); err != nil {
  return fmt.Errorf("error getting html for location %s: %w", location, err)
+ } else if md, err := b.translators[language.LanguageMD].Translate(html); err != nil {
+ return fmt.Errorf("error translating html to %s for location %s: %w", language.LanguageMD, location, err)
  } else {
  b.display.HTML = html
  b.display.Location = location
- if b.display.MD == "" {
- b.display.MD = "No MD display available yet."
- }
+ b.display.MD = md
  return nil
  }
 }

cmd/shell/shell.go

@@ -21,8 +21,8 @@ func main() {
  runHeadful := flag.Bool("headful", false, "run the browser in non-headless mode")
  logPath := flag.String("log-path", "out", "the path to write the trajectory and browser display to")
  initialURL := flag.String("url", "https://www.google.com", "the initial url to visit")
- actorStrategy := flag.String("actor-strategy", "base_llm", "the actor strategy to use; one of [base_llm, reflexion]")
- afforderStrategy := flag.String("afforder-strategy", "function_afforder", "the afforder strategy to use; one of [function_afforder, filter_afforder]")
+ actorStrategy := flag.String("actor-strategy", "base", "the actor strategy to use; one of [\"base\", \"reflexion\"]")
+ afforderStrategy := flag.String("afforder-strategy", "function", "the afforder strategy to use; one of [\"function\", filter\"]")
  verbose := flag.Bool("verbose", false, "whether to print verbose debug logs")
  flag.Parse()