chore(mcpbox): use ubuntu:24.04 as base

Signed-off-by: mudler <mudler@localai.io>

.github/workflows/tests.yml

@@ -15,7 +15,7 @@ jobs:
     runs-on: ubuntu-latest
     steps:
     - name: Checkout code
-      uses: actions/checkout@v2
+      uses: actions/checkout@v4
     - run: |
         # Add Docker's official GPG key:
         sudo apt-get update

Dockerfile.mcpbox

@@ -20,10 +20,10 @@ COPY . .
 RUN CGO_ENABLED=0 GOOS=linux go build -o mcpbox ./cmd/mcpbox
 
 # Final stage
-FROM alpine:3.19
+FROM ubuntu:22.04
 
 # Install runtime dependencies
-RUN apk add --no-cache ca-certificates tzdata docker
+RUN apt-get update && apt-get install -y ca-certificates tzdata docker.io bash
 
 # Create non-root user
 #RUN adduser -D -g '' appuser

docker-compose.intel.yaml

@@ -12,6 +12,11 @@ services:
       - /dev/dri/card1
       - /dev/dri/renderD129
 
+  mcpbox:
+    extends:
+      file: docker-compose.yaml
+      service: mcpbox
+
   localrecall:
     extends:
       file: docker-compose.yaml

docker-compose.nvidia.yaml

@@ -17,6 +17,11 @@ services:
               count: 1
               capabilities: [gpu]
 
+  mcpbox:
+    extends:
+      file: docker-compose.yaml
+      service: mcpbox
+
   localrecall:
     extends:
       file: docker-compose.yaml
@@ -30,4 +35,4 @@ services:
   localagi:
     extends:
       file: docker-compose.yaml
-      service: localagi
+      service: localagi

pkg/localoperator/client.go

@@ -1,4 +1,4 @@
-package api
+package localoperator
 
 import (
 	"bytes"

services/actions/browseragentrunner.go

@@ -10,6 +10,10 @@ import (
 	"github.com/sashabaranov/go-openai/jsonschema"
 )
 
+const (
+	MetadataBrowserAgentHistory = "browser_agent_history"
+)
+
 type BrowserAgentRunner struct {
 	baseURL, customActionName string
 	client                    *api.Client
@@ -62,7 +66,7 @@ func (b *BrowserAgentRunner) Run(ctx context.Context, params types.ActionParams)
 
 	return types.ActionResult{
 		Result:   fmt.Sprintf("Browser agent completed successfully. History:\n%s", historyStr),
-		Metadata: map[string]interface{}{"browser_agent_history": stateHistory},
+		Metadata: map[string]interface{}{MetadataBrowserAgentHistory: stateHistory},
 	}, nil
 }
 

services/connectors/discord.go

@@ -30,9 +30,15 @@ func NewDiscord(config map[string]string) *Discord {
 		duration = 5 * time.Minute
 	}
 
+	token := config["token"]
+
+	if !strings.HasPrefix(token, "Bot ") {
+		token = "Bot " + token
+	}
+
 	return &Discord{
 		conversationTracker: NewConversationTracker[string](duration),
-		token:               config["token"],
+		token:               token,
 		defaultChannel:      config["defaultChannel"],
 	}
 }

services/connectors/slack.go

@@ -11,6 +11,7 @@ import (
 	"time"
 
 	"github.com/mudler/LocalAGI/pkg/config"
+	"github.com/mudler/LocalAGI/pkg/localoperator"
 	"github.com/mudler/LocalAGI/pkg/xlog"
 	"github.com/mudler/LocalAGI/pkg/xstrings"
 	"github.com/mudler/LocalAGI/services/actions"
@@ -167,8 +168,38 @@ func replaceUserIDsWithNamesInMessage(api *slack.Client, message string) string
 	return message
 }
 
-func generateAttachmentsFromJobResponse(j *types.JobResult) (attachments []slack.Attachment) {
+func generateAttachmentsFromJobResponse(j *types.JobResult, api *slack.Client, channelID, ts string) (attachments []slack.Attachment) {
 	for _, state := range j.State {
+		// coming from the browser agent
+		if history, exists := state.Metadata[actions.MetadataBrowserAgentHistory]; exists {
+			if historyStruct, ok := history.(*localoperator.StateHistory); ok {
+				state := historyStruct.States[len(historyStruct.States)-1]
+				// Decode base64 screenshot and upload to Slack
+				if state.Screenshot != "" {
+					screenshotData, err := base64.StdEncoding.DecodeString(state.Screenshot)
+					if err != nil {
+						xlog.Error(fmt.Sprintf("Error decoding screenshot: %v", err))
+						continue
+					}
+
+					data := string(screenshotData)
+					// Upload the file to Slack
+					_, err = api.UploadFileV2(slack.UploadFileV2Parameters{
+						Reader:          bytes.NewReader(screenshotData),
+						FileSize:        len(data),
+						ThreadTimestamp: ts,
+						Channel:         channelID,
+						Filename:        "screenshot.png",
+						InitialComment:  "Browser Agent Screenshot",
+					})
+					if err != nil {
+						xlog.Error(fmt.Sprintf("Error uploading screenshot: %v", err))
+						continue
+					}
+				}
+			}
+		}
+
 		// coming from the search action
 		if urls, exists := state.Metadata[actions.MetadataUrls]; exists {
 			for _, url := range xstrings.UniqueSlice(urls.([]string)) {
@@ -375,7 +406,7 @@ func replyWithPostMessage(finalResponse string, api *slack.Client, ev *slackeven
 				slack.MsgOptionEnableLinkUnfurl(),
 				slack.MsgOptionText(message, true),
 				slack.MsgOptionPostMessageParameters(postMessageParams),
-				slack.MsgOptionAttachments(generateAttachmentsFromJobResponse(res)...),
+				slack.MsgOptionAttachments(generateAttachmentsFromJobResponse(res, api, ev.Channel, "")...),
 			)
 			if err != nil {
 				xlog.Error(fmt.Sprintf("Error posting message: %v", err))
@@ -387,7 +418,7 @@ func replyWithPostMessage(finalResponse string, api *slack.Client, ev *slackeven
 			slack.MsgOptionEnableLinkUnfurl(),
 			slack.MsgOptionText(res.Response, true),
 			slack.MsgOptionPostMessageParameters(postMessageParams),
-			slack.MsgOptionAttachments(generateAttachmentsFromJobResponse(res)...),
+			slack.MsgOptionAttachments(generateAttachmentsFromJobResponse(res, api, ev.Channel, "")...),
 		//	slack.MsgOptionTS(ts),
 		)
 		if err != nil {
@@ -408,7 +439,7 @@ func replyToUpdateMessage(finalResponse string, api *slack.Client, ev *slackeven
 			slack.MsgOptionLinkNames(true),
 			slack.MsgOptionEnableLinkUnfurl(),
 			slack.MsgOptionText(messages[0], true),
-			slack.MsgOptionAttachments(generateAttachmentsFromJobResponse(res)...),
+			slack.MsgOptionAttachments(generateAttachmentsFromJobResponse(res, api, ev.Channel, msgTs)...),
 		)
 		if err != nil {
 			xlog.Error(fmt.Sprintf("Error updating final message: %v", err))
@@ -435,7 +466,7 @@ func replyToUpdateMessage(finalResponse string, api *slack.Client, ev *slackeven
 			slack.MsgOptionLinkNames(true),
 			slack.MsgOptionEnableLinkUnfurl(),
 			slack.MsgOptionText(finalResponse, true),
-			slack.MsgOptionAttachments(generateAttachmentsFromJobResponse(res)...),
+			slack.MsgOptionAttachments(generateAttachmentsFromJobResponse(res, api, ev.Channel, msgTs)...),
 		)
 		if err != nil {
 			xlog.Error(fmt.Sprintf("Error updating final message: %v", err))

slack.yaml

@@ -23,6 +23,7 @@ oauth_config:
       - commands
       - groups:history
       - files:read
+      - files:write
       - im:history
       - im:read
       - im:write