feat(browseragent): add browser agent runner action (#55)
Signed-off-by: Ettore Di Giacinto <mudler@localai.io>
This commit is contained in:
Ettore Di Giacinto
committed by
GitHub
GitHub
parent
b5a12a1da6
commit
50e56fe22f
@@ -18,6 +18,7 @@ const (
|
||||
// Actions
|
||||
ActionSearch = "search"
|
||||
ActionCustom = "custom"
|
||||
ActionBrowserAgentRunner = "browser-agent-runner"
|
||||
ActionGithubIssueLabeler = "github-issue-labeler"
|
||||
ActionGithubIssueOpener = "github-issue-opener"
|
||||
ActionGithubIssueCloser = "github-issue-closer"
|
||||
@@ -52,6 +53,7 @@ var AvailableActions = []string{
|
||||
ActionGithubIssueSearcher,
|
||||
ActionGithubRepositoryGet,
|
||||
ActionGithubGetAllContent,
|
||||
ActionBrowserAgentRunner,
|
||||
ActionGithubRepositoryCreateOrUpdate,
|
||||
ActionGithubIssueReader,
|
||||
ActionGithubIssueCommenter,
|
||||
@@ -71,31 +73,34 @@ var AvailableActions = []string{
|
||||
ActionShellcommand,
|
||||
}
|
||||
|
||||
func Actions(a *state.AgentConfig) func(ctx context.Context, pool *state.AgentPool) []types.Action {
|
||||
return func(ctx context.Context, pool *state.AgentPool) []types.Action {
|
||||
allActions := []types.Action{}
|
||||
func Actions(actionsConfigs map[string]string) func(a *state.AgentConfig) func(ctx context.Context, pool *state.AgentPool) []types.Action {
|
||||
return func(a *state.AgentConfig) func(ctx context.Context, pool *state.AgentPool) []types.Action {
|
||||
return func(ctx context.Context, pool *state.AgentPool) []types.Action {
|
||||
allActions := []types.Action{}
|
||||
|
||||
agentName := a.Name
|
||||
agentName := a.Name
|
||||
|
||||
for _, a := range a.Actions {
|
||||
var config map[string]string
|
||||
if err := json.Unmarshal([]byte(a.Config), &config); err != nil {
|
||||
xlog.Error("Error unmarshalling action config", "error", err)
|
||||
continue
|
||||
for _, a := range a.Actions {
|
||||
var config map[string]string
|
||||
if err := json.Unmarshal([]byte(a.Config), &config); err != nil {
|
||||
xlog.Error("Error unmarshalling action config", "error", err)
|
||||
continue
|
||||
}
|
||||
|
||||
a, err := Action(a.Name, agentName, config, pool, actionsConfigs)
|
||||
if err != nil {
|
||||
continue
|
||||
}
|
||||
allActions = append(allActions, a)
|
||||
}
|
||||
|
||||
a, err := Action(a.Name, agentName, config, pool)
|
||||
if err != nil {
|
||||
continue
|
||||
}
|
||||
allActions = append(allActions, a)
|
||||
return allActions
|
||||
}
|
||||
|
||||
return allActions
|
||||
}
|
||||
|
||||
}
|
||||
|
||||
func Action(name, agentName string, config map[string]string, pool *state.AgentPool) (types.Action, error) {
|
||||
func Action(name, agentName string, config map[string]string, pool *state.AgentPool, actionsConfigs map[string]string) (types.Action, error) {
|
||||
var a types.Action
|
||||
var err error
|
||||
|
||||
@@ -114,6 +119,8 @@ func Action(name, agentName string, config map[string]string, pool *state.AgentP
|
||||
a = actions.NewGithubIssueCloser(config)
|
||||
case ActionGithubIssueSearcher:
|
||||
a = actions.NewGithubIssueSearch(config)
|
||||
case ActionBrowserAgentRunner:
|
||||
a = actions.NewBrowserAgentRunner(config, actionsConfigs["browser-agent-runner-base-url"])
|
||||
case ActionGithubIssueReader:
|
||||
a = actions.NewGithubIssueReader(config)
|
||||
case ActionGithubPRReader:
|
||||
@@ -169,6 +176,11 @@ func ActionsConfigMeta() []config.FieldGroup {
|
||||
Label: "Search",
|
||||
Fields: actions.SearchConfigMeta(),
|
||||
},
|
||||
{
|
||||
Name: "browser-agent-runner",
|
||||
Label: "Browser Agent Runner",
|
||||
Fields: actions.BrowserAgentRunnerConfigMeta(),
|
||||
},
|
||||
{
|
||||
Name: "generate_image",
|
||||
Label: "Generate Image",
|
||||
|
||||
116
services/actions/browseragentrunner.go
Normal file
116
services/actions/browseragentrunner.go
Normal file
@@ -0,0 +1,116 @@
|
||||
package actions
|
||||
|
||||
import (
|
||||
"context"
|
||||
"fmt"
|
||||
|
||||
"github.com/mudler/LocalAGI/core/types"
|
||||
"github.com/mudler/LocalAGI/pkg/config"
|
||||
api "github.com/mudler/LocalAGI/pkg/localoperator"
|
||||
"github.com/sashabaranov/go-openai/jsonschema"
|
||||
)
|
||||
|
||||
type BrowserAgentRunner struct {
|
||||
baseURL, customActionName string
|
||||
client *api.Client
|
||||
}
|
||||
|
||||
func NewBrowserAgentRunner(config map[string]string, defaultURL string) *BrowserAgentRunner {
|
||||
if config["baseURL"] == "" {
|
||||
config["baseURL"] = defaultURL
|
||||
}
|
||||
|
||||
client := api.NewClient(config["baseURL"])
|
||||
|
||||
return &BrowserAgentRunner{
|
||||
client: client,
|
||||
baseURL: config["baseURL"],
|
||||
customActionName: config["customActionName"],
|
||||
}
|
||||
}
|
||||
|
||||
func (b *BrowserAgentRunner) Run(ctx context.Context, params types.ActionParams) (types.ActionResult, error) {
|
||||
result := api.AgentRequest{}
|
||||
err := params.Unmarshal(&result)
|
||||
if err != nil {
|
||||
return types.ActionResult{}, fmt.Errorf("failed to unmarshal params: %w", err)
|
||||
}
|
||||
|
||||
req := api.AgentRequest{
|
||||
Goal: result.Goal,
|
||||
MaxAttempts: result.MaxAttempts,
|
||||
MaxNoActionAttempts: result.MaxNoActionAttempts,
|
||||
}
|
||||
|
||||
stateHistory, err := b.client.RunBrowserAgent(req)
|
||||
if err != nil {
|
||||
return types.ActionResult{}, fmt.Errorf("failed to run browser agent: %w", err)
|
||||
}
|
||||
|
||||
// Format the state history into a readable string
|
||||
var historyStr string
|
||||
// for i, state := range stateHistory.States {
|
||||
// historyStr += fmt.Sprintf("State %d:\n", i+1)
|
||||
// historyStr += fmt.Sprintf(" URL: %s\n", state.CurrentURL)
|
||||
// historyStr += fmt.Sprintf(" Title: %s\n", state.PageTitle)
|
||||
// historyStr += fmt.Sprintf(" Description: %s\n\n", state.PageContentDescription)
|
||||
// }
|
||||
|
||||
historyStr += fmt.Sprintf(" URL: %s\n", stateHistory.States[len(stateHistory.States)-1].CurrentURL)
|
||||
historyStr += fmt.Sprintf(" Title: %s\n", stateHistory.States[len(stateHistory.States)-1].PageTitle)
|
||||
historyStr += fmt.Sprintf(" Description: %s\n\n", stateHistory.States[len(stateHistory.States)-1].PageContentDescription)
|
||||
|
||||
return types.ActionResult{
|
||||
Result: fmt.Sprintf("Browser agent completed successfully. History:\n%s", historyStr),
|
||||
}, nil
|
||||
}
|
||||
|
||||
func (b *BrowserAgentRunner) Definition() types.ActionDefinition {
|
||||
actionName := "run_browser_agent"
|
||||
if b.customActionName != "" {
|
||||
actionName = b.customActionName
|
||||
}
|
||||
description := "Run a browser agent to achieve a specific goal, for example: 'Go to https://www.google.com and search for 'LocalAI', and tell me what's on the first page'"
|
||||
return types.ActionDefinition{
|
||||
Name: types.ActionDefinitionName(actionName),
|
||||
Description: description,
|
||||
Properties: map[string]jsonschema.Definition{
|
||||
"goal": {
|
||||
Type: jsonschema.String,
|
||||
Description: "The goal for the browser agent to achieve",
|
||||
},
|
||||
"max_attempts": {
|
||||
Type: jsonschema.Number,
|
||||
Description: "Maximum number of attempts the agent can make (optional)",
|
||||
},
|
||||
"max_no_action_attempts": {
|
||||
Type: jsonschema.Number,
|
||||
Description: "Maximum number of attempts without taking an action (optional)",
|
||||
},
|
||||
},
|
||||
Required: []string{"goal"},
|
||||
}
|
||||
}
|
||||
|
||||
func (a *BrowserAgentRunner) Plannable() bool {
|
||||
return true
|
||||
}
|
||||
|
||||
// BrowserAgentRunnerConfigMeta returns the metadata for Browser Agent Runner action configuration fields
|
||||
func BrowserAgentRunnerConfigMeta() []config.Field {
|
||||
return []config.Field{
|
||||
{
|
||||
Name: "baseURL",
|
||||
Label: "Base URL",
|
||||
Type: config.FieldTypeText,
|
||||
Required: false,
|
||||
HelpText: "Base URL of the LocalOperator API",
|
||||
},
|
||||
{
|
||||
Name: "customActionName",
|
||||
Label: "Custom Action Name",
|
||||
Type: config.FieldTypeText,
|
||||
HelpText: "Custom name for this action",
|
||||
},
|
||||
}
|
||||
}
|
||||
Reference in New Issue
Block a user