feat: Implement vision capabilities and image generation tools, adding media handling, dedicated configurations, and trace optimization for image data.

cmd/gateway.go

@@ -182,6 +182,10 @@ func runGateway() {
 	toolsReg.Register(webFetchTool)
 	slog.Info("web_fetch tool enabled")
 
+	// Vision fallback tool (for non-vision providers like MiniMax)
+	toolsReg.Register(tools.NewReadImageTool(providerRegistry))
+	toolsReg.Register(tools.NewCreateImageTool(providerRegistry))
+
 	// TTS (text-to-speech) system
 	ttsMgr := setupTTS(cfg)
 	if ttsMgr != nil {

cmd/gateway_consumer.go

@@ -150,6 +150,7 @@ func consumeInboundMessages(ctx context.Context, msgBus *bus.MessageBus, agents
 		outCh := sched.ScheduleWithOpts(ctx, "main", agent.RunRequest{
 			SessionKey:        sessionKey,
 			Message:           msg.Content,
+			Media:             msg.Media,
 			Channel:           msg.Channel,
 			ChatID:            msg.ChatID,
 			PeerKind:          peerKind,
@@ -225,12 +226,28 @@ func consumeInboundMessages(ctx context.Context, msgBus *bus.MessageBus, agents
 			}
 
 			// Publish response back to the channel
-			msgBus.PublishOutbound(bus.OutboundMessage{
+			outMsg := bus.OutboundMessage{
 				Channel:  channel,
 				ChatID:   chatID,
 				Content:  outcome.Result.Content,
 				Metadata: meta,
-			})
+			}
+
+			// Convert media results from agent run to outbound media attachments
+			for _, mr := range outcome.Result.Media {
+				outMsg.Media = append(outMsg.Media, bus.MediaAttachment{
+					URL:         mr.Path,
+					ContentType: mr.ContentType,
+				})
+				if mr.AsVoice {
+					if outMsg.Metadata == nil {
+						outMsg.Metadata = make(map[string]string)
+					}
+					outMsg.Metadata["audio_as_voice"] = "true"
+				}
+			}
+
+			msgBus.PublishOutbound(outMsg)
 		}(msg.Channel, msg.ChatID, sessionKey, runID, outMeta)
 	}
 

cmd/gateway_providers.go

@@ -3,6 +3,7 @@ package cmd
 import (
 	"context"
 	"log/slog"
+	"strings"
 
 	"github.com/nextlevelbuilder/goclaw/internal/config"
 	"github.com/nextlevelbuilder/goclaw/internal/providers"
@@ -51,7 +52,8 @@ func registerProviders(registry *providers.Registry, cfg *config.Config) {
 	}
 
 	if cfg.Providers.MiniMax.APIKey != "" {
-		registry.Register(providers.NewOpenAIProvider("minimax", cfg.Providers.MiniMax.APIKey, "https://api.minimax.io/v1", "MiniMax-M2.5"))
+		registry.Register(providers.NewOpenAIProvider("minimax", cfg.Providers.MiniMax.APIKey, "https://api.minimax.io/v1", "MiniMax-M2.5").
+			WithChatPath("/text/chatcompletion_v2"))
 		slog.Info("registered provider", "name", "minimax")
 	}
 
@@ -82,7 +84,12 @@ func registerProvidersFromDB(registry *providers.Registry, provStore store.Provi
 		if p.ProviderType == "anthropic_native" {
 			registry.Register(providers.NewAnthropicProvider(p.APIKey))
 		} else {
-			registry.Register(providers.NewOpenAIProvider(p.Name, p.APIKey, p.APIBase, ""))
+			prov := providers.NewOpenAIProvider(p.Name, p.APIKey, p.APIBase, "")
+			// MiniMax native API uses a different chat path for vision support.
+			if p.Name == "minimax" && strings.Contains(p.APIBase, "minimax.io") {
+				prov.WithChatPath("/text/chatcompletion_v2")
+			}
+			registry.Register(prov)
 		}
 		slog.Info("registered provider from DB", "name", p.Name)
 	}