diff --git a/client/agent.go b/client/agent.go
index f669f4f5..f3773ac2 100644
--- a/client/agent.go
+++ b/client/agent.go
@@ -22,16 +22,17 @@ type MentionedItem struct {
 
 // AgentQARequest agent Q&A request payload.
 type AgentQARequest struct {
-	Query            string          `json:"query"`                        // Required query text
-	KnowledgeBaseIDs []string        `json:"knowledge_base_ids,omitempty"` // Optional KBs for this query
-	KnowledgeIDs     []string        `json:"knowledge_ids,omitempty"`      // Optional specific knowledge IDs for this query
-	AgentEnabled     bool            `json:"agent_enabled"`                // Whether to run in agent mode
-	AgentID          string          `json:"agent_id,omitempty"`           // Optional custom agent ID
-	WebSearchEnabled bool            `json:"web_search_enabled"`           // Whether to enable web search
-	SummaryModelID   string          `json:"summary_model_id,omitempty"`   // Optional summary model override
-	MentionedItems   []MentionedItem `json:"mentioned_items,omitempty"`    // @mentioned knowledge bases and files
-	DisableTitle     bool            `json:"disable_title,omitempty"`      // Whether to disable auto title generation
-	MCPServiceIDs    []string        `json:"mcp_service_ids,omitempty"`    // Optional MCP service allow list (deprecated)
+	Query            string            `json:"query"`                        // Required query text
+	KnowledgeBaseIDs []string          `json:"knowledge_base_ids,omitempty"` // Optional KBs for this query
+	KnowledgeIDs     []string          `json:"knowledge_ids,omitempty"`      // Optional specific knowledge IDs for this query
+	AgentEnabled     bool              `json:"agent_enabled"`                // Whether to run in agent mode
+	AgentID          string            `json:"agent_id,omitempty"`           // Optional custom agent ID
+	WebSearchEnabled bool              `json:"web_search_enabled"`           // Whether to enable web search
+	SummaryModelID   string            `json:"summary_model_id,omitempty"`   // Optional summary model override
+	MentionedItems   []MentionedItem   `json:"mentioned_items,omitempty"`    // @mentioned knowledge bases and files
+	DisableTitle     bool              `json:"disable_title,omitempty"`      // Whether to disable auto title generation
+	MCPServiceIDs    []string          `json:"mcp_service_ids,omitempty"`    // Optional MCP service allow list (deprecated)
+	Images           []ImageAttachment `json:"images,omitempty"`             // Attached images for multimodal chat
 }
 
 // AgentResponseType defines the type of agent response
diff --git a/client/organization.go b/client/organization.go
index ee6c4023..a382c537 100644
--- a/client/organization.go
+++ b/client/organization.go
@@ -126,6 +126,19 @@ type JoinRequestResponse struct {
 	ReviewedAt    *time.Time `json:"reviewed_at,omitempty"`
 }
 
+// UserInfo represents user information for API responses
+type UserInfo struct {
+	ID                  string    `json:"id"`
+	Username            string    `json:"username"`
+	Email               string    `json:"email"`
+	Avatar              string    `json:"avatar"`
+	TenantID            uint64    `json:"tenant_id"`
+	IsActive            bool      `json:"is_active"`
+	CanAccessAllTenants bool      `json:"can_access_all_tenants"`
+	CreatedAt           time.Time `json:"created_at"`
+	UpdatedAt           time.Time `json:"updated_at"`
+}
+
 // SharedKnowledgeBaseInfo represents a shared knowledge base
 type SharedKnowledgeBaseInfo struct {
 	ShareID        string    `json:"share_id"`
diff --git a/client/session.go b/client/session.go
index 9ea3621b..831fc4c6 100644
--- a/client/session.go
+++ b/client/session.go
@@ -198,16 +198,26 @@ func (c *Client) GenerateTitle(ctx context.Context, sessionID string, request *G
 	return response.Data, nil
 }
 
+// ImageAttachment represents an image in a chat request.
+// Frontend sends base64 data in the Data field; the backend saves, runs VLM analysis,
+// and populates URL/Caption before proceeding with the chat pipeline.
+type ImageAttachment struct {
+	Data    string `json:"data,omitempty"`    // base64 data URI (data:image/png;base64,...)
+	URL     string `json:"url,omitempty"`     // serving URL after saving to storage
+	Caption string `json:"caption,omitempty"` // VLM analysis result
+}
+
 // KnowledgeQARequest knowledge Q&A request
 type KnowledgeQARequest struct {
-	Query            string   `json:"query"`              // Query text for knowledge base search
-	KnowledgeBaseIDs []string `json:"knowledge_base_ids"` // Selected knowledge base IDs for this request
-	KnowledgeIDs     []string `json:"knowledge_ids"`      // Selected knowledge IDs for this request
-	AgentEnabled     bool     `json:"agent_enabled"`      // Whether agent mode is enabled for this request
-	AgentID          string   `json:"agent_id"`           // Selected custom agent ID for this request
-	WebSearchEnabled bool     `json:"web_search_enabled"` // Whether web search is enabled for this request
-	SummaryModelID   string   `json:"summary_model_id"`   // Optional summary model ID (overrides session default)
-	DisableTitle     bool     `json:"disable_title"`      // Whether to disable auto title generation
+	Query            string            `json:"query"`              // Query text for knowledge base search
+	KnowledgeBaseIDs []string          `json:"knowledge_base_ids"` // Selected knowledge base IDs for this request
+	KnowledgeIDs     []string          `json:"knowledge_ids"`      // Selected knowledge IDs for this request
+	AgentEnabled     bool              `json:"agent_enabled"`      // Whether agent mode is enabled for this request
+	AgentID          string            `json:"agent_id"`           // Selected custom agent ID for this request
+	WebSearchEnabled bool              `json:"web_search_enabled"` // Whether web search is enabled for this request
+	SummaryModelID   string            `json:"summary_model_id"`   // Optional summary model ID (overrides session default)
+	DisableTitle     bool              `json:"disable_title"`      // Whether to disable auto title generation
+	Images           []ImageAttachment `json:"images,omitempty"`   // Attached images for multimodal chat
 }
 
 // LLMToolCall represents a function/tool call from the LLM