From a45d9502dbc1c37176607347409502144c75ae16 Mon Sep 17 00:00:00 2001
From: Ettore Di Giacinto <mudler@localai.io>
Date: Sat, 29 Nov 2025 22:35:33 +0100
Subject: [PATCH 1/2] feat(agent-jobs): add multimedia support

Signed-off-by: Ettore Di Giacinto <mudler@localai.io>
---
 core/http/endpoints/localai/agent_jobs.go |  25 +-
 core/http/views/agent-jobs.html           | 145 +++++++++-
 core/http/views/agent-task-details.html   | 257 +++++++++++++++++-
 core/http/views/backends.html             |   5 +-
 core/http/views/chat.html                 |   5 +-
 core/http/views/models.html               |   5 +-
 core/http/views/text2image.html           |   5 +-
 core/schema/agent_jobs.go                 |  25 ++
 core/services/agent_jobs.go               | 206 +++++++++++++-
 core/services/agent_jobs_test.go          | 316 +++++++++++++++++++++-
 10 files changed, 949 insertions(+), 45 deletions(-)

diff --git a/core/http/endpoints/localai/agent_jobs.go b/core/http/endpoints/localai/agent_jobs.go
index 0e65a241df4b..267547434950 100644
--- a/core/http/endpoints/localai/agent_jobs.go
+++ b/core/http/endpoints/localai/agent_jobs.go
@@ -147,7 +147,28 @@ func ExecuteJobEndpoint(app *application.Application) echo.HandlerFunc {
 			req.Parameters = make(map[string]string)
 		}
 
-		jobID, err := app.AgentJobService().ExecuteJob(req.TaskID, req.Parameters, "api")
+		// Build multimedia struct from request
+		var multimedia *struct {
+			Images []string
+			Videos []string
+			Audios []string
+			Files  []string
+		}
+		if len(req.Images) > 0 || len(req.Videos) > 0 || len(req.Audios) > 0 || len(req.Files) > 0 {
+			multimedia = &struct {
+				Images []string
+				Videos []string
+				Audios []string
+				Files  []string
+			}{
+				Images: req.Images,
+				Videos: req.Videos,
+				Audios: req.Audios,
+				Files:  req.Files,
+			}
+		}
+
+		jobID, err := app.AgentJobService().ExecuteJob(req.TaskID, req.Parameters, "api", multimedia)
 		if err != nil {
 			return c.JSON(http.StatusBadRequest, map[string]string{"error": err.Error()})
 		}
@@ -323,7 +344,7 @@ func ExecuteTaskByNameEndpoint(app *application.Application) echo.HandlerFunc {
 			return c.JSON(http.StatusNotFound, map[string]string{"error": "Task not found: " + name})
 		}
 
-		jobID, err := app.AgentJobService().ExecuteJob(task.ID, params, "api")
+		jobID, err := app.AgentJobService().ExecuteJob(task.ID, params, "api", nil)
 		if err != nil {
 			return c.JSON(http.StatusBadRequest, map[string]string{"error": err.Error()})
 		}
diff --git a/core/http/views/agent-jobs.html b/core/http/views/agent-jobs.html
index b650e1a4ccf7..a439f6bd2333 100644
--- a/core/http/views/agent-jobs.html
+++ b/core/http/views/agent-jobs.html
@@ -365,21 +365,40 @@ <h2 class="text-2xl font-semibold text-[var(--color-text-primary)]">Job History<
          x-cloak
          @click.away="showExecuteTaskModal = false; selectedTaskForExecution = null; executionParameters = {}; executionParametersText = ''"
          class="fixed inset-0 bg-black/50 flex items-center justify-center z-50">
-        <div class="bg-[var(--color-bg-secondary)] border border-[var(--color-primary-border)]/20 rounded-xl p-8 max-w-2xl w-full mx-4">
-            <div class="flex justify-between items-center mb-6">
+        <div class="bg-[var(--color-bg-secondary)] border border-[var(--color-primary-border)]/20 rounded-xl max-w-2xl w-full mx-4 max-h-[90vh] flex flex-col">
+            <div class="flex justify-between items-center p-8 pb-6 border-b border-[var(--color-primary-border)]/20">
                 <h3 class="text-2xl font-semibold text-[var(--color-text-primary)]">Execute Task</h3>
-                <button @click="showExecuteTaskModal = false; selectedTaskForExecution = null; executionParameters = {}; executionParametersText = ''" 
+                        <button @click="showExecuteTaskModal = false; selectedTaskForExecution = null; executionParameters = {}; executionParametersText = ''; executionMultimedia = {images: '', videos: '', audios: '', files: ''}; executeModalTab = 'parameters'"
                         class="text-[var(--color-text-secondary)] hover:text-[var(--color-text-primary)]">
                     <i class="fas fa-times text-xl"></i>
                 </button>
             </div>
             <template x-if="selectedTaskForExecution">
-                <div class="space-y-4">
+                <div class="flex flex-col flex-1 min-h-0">
+                    <div class="flex-1 overflow-y-auto px-8 py-6 space-y-4">
                     <div>
                         <label class="block text-sm font-medium text-[var(--color-text-primary)] mb-2">Task</label>
                         <div class="text-[var(--color-text-secondary)]" x-text="selectedTaskForExecution.name"></div>
                     </div>
-                    <div>
+                    
+                    <!-- Tabs for Parameters and Multimedia -->
+                    <div class="border-b border-[var(--color-primary-border)]/20">
+                        <div class="flex space-x-4">
+                            <button @click="executeModalTab = 'parameters'" 
+                                    :class="executeModalTab === 'parameters' ? 'border-b-2 border-[var(--color-primary)] text-[var(--color-primary)]' : 'text-[var(--color-text-secondary)] hover:text-[var(--color-text-primary)]'"
+                                    class="px-4 py-2 font-medium transition-colors">
+                                Parameters
+                            </button>
+                            <button @click="executeModalTab = 'multimedia'" 
+                                    :class="executeModalTab === 'multimedia' ? 'border-b-2 border-[var(--color-primary)] text-[var(--color-primary)]' : 'text-[var(--color-text-secondary)] hover:text-[var(--color-text-primary)]'"
+                                    class="px-4 py-2 font-medium transition-colors">
+                                Multimedia
+                            </button>
+                        </div>
+                    </div>
+                    
+                    <!-- Parameters Tab -->
+                    <div x-show="executeModalTab === 'parameters'">
                         <label class="block text-sm font-medium text-[var(--color-text-primary)] mb-2">Parameters</label>
                         <p class="text-xs text-[var(--color-text-secondary)] mb-3">
                             Enter parameters as key-value pairs (one per line, format: key=value). 
@@ -393,8 +412,61 @@ <h3 class="text-2xl font-semibold text-[var(--color-text-primary)]">Execute Task
                             Example: <code class="bg-[var(--color-bg-primary)] px-1 py-0.5 rounded text-[var(--color-primary)]">user_name=Alice</code>
                         </p>
                     </div>
-                    <div class="flex justify-end space-x-4">
-                        <button @click="showExecuteTaskModal = false; selectedTaskForExecution = null; executionParameters = {}; executionParametersText = ''" 
+                    
+                    <!-- Multimedia Tab -->
+                    <div x-show="executeModalTab === 'multimedia'" class="space-y-4">
+                        <p class="text-xs text-[var(--color-text-secondary)] mb-3">
+                            Provide multimedia content as URLs or base64-encoded data URIs. You can also upload files which will be converted to base64.
+                        </p>
+                        
+                        <!-- Images -->
+                        <div>
+                            <label class="block text-sm font-medium text-[var(--color-text-primary)] mb-2">Images</label>
+                            <textarea x-model="executionMultimedia.images" 
+                                      rows="3"
+                                      placeholder="https://example.com/image.png&#10;data:image/png;base64,iVBORw0KG..."
+                                      class="w-full bg-[var(--color-bg-primary)] border border-[var(--color-primary-border)]/20 rounded px-4 py-2 text-[var(--color-text-primary)] font-mono text-sm focus:border-[var(--color-primary-border)] focus:ring-2 focus:ring-[#38BDF8]/50"></textarea>
+                            <input type="file" @change="handleFileUpload($event, 'image')" accept="image/*" multiple
+                                   class="mt-2 text-sm text-[var(--color-text-secondary)] file:mr-4 file:py-2 file:px-4 file:rounded file:border-0 file:text-sm file:font-semibold file:bg-[var(--color-primary)] file:text-white hover:file:bg-[var(--color-primary-hover)]">
+                        </div>
+                        
+                        <!-- Videos -->
+                        <div>
+                            <label class="block text-sm font-medium text-[var(--color-text-primary)] mb-2">Videos</label>
+                            <textarea x-model="executionMultimedia.videos" 
+                                      rows="3"
+                                      placeholder="https://example.com/video.mp4&#10;data:video/mp4;base64,..."
+                                      class="w-full bg-[var(--color-bg-primary)] border border-[var(--color-primary-border)]/20 rounded px-4 py-2 text-[var(--color-text-primary)] font-mono text-sm focus:border-[var(--color-primary-border)] focus:ring-2 focus:ring-[#38BDF8]/50"></textarea>
+                            <input type="file" @change="handleFileUpload($event, 'video')" accept="video/*" multiple
+                                   class="mt-2 text-sm text-[var(--color-text-secondary)] file:mr-4 file:py-2 file:px-4 file:rounded file:border-0 file:text-sm file:font-semibold file:bg-[var(--color-primary)] file:text-white hover:file:bg-[var(--color-primary-hover)]">
+                        </div>
+                        
+                        <!-- Audios -->
+                        <div>
+                            <label class="block text-sm font-medium text-[var(--color-text-primary)] mb-2">Audios</label>
+                            <textarea x-model="executionMultimedia.audios" 
+                                      rows="3"
+                                      placeholder="https://example.com/audio.mp3&#10;data:audio/mpeg;base64,..."
+                                      class="w-full bg-[var(--color-bg-primary)] border border-[var(--color-primary-border)]/20 rounded px-4 py-2 text-[var(--color-text-primary)] font-mono text-sm focus:border-[var(--color-primary-border)] focus:ring-2 focus:ring-[#38BDF8]/50"></textarea>
+                            <input type="file" @change="handleFileUpload($event, 'audio')" accept="audio/*" multiple
+                                   class="mt-2 text-sm text-[var(--color-text-secondary)] file:mr-4 file:py-2 file:px-4 file:rounded file:border-0 file:text-sm file:font-semibold file:bg-[var(--color-primary)] file:text-white hover:file:bg-[var(--color-primary-hover)]">
+                        </div>
+                        
+                        <!-- Files -->
+                        <div>
+                            <label class="block text-sm font-medium text-[var(--color-text-primary)] mb-2">Files</label>
+                            <textarea x-model="executionMultimedia.files" 
+                                      rows="3"
+                                      placeholder="https://example.com/file.pdf&#10;data:application/pdf;base64,..."
+                                      class="w-full bg-[var(--color-bg-primary)] border border-[var(--color-primary-border)]/20 rounded px-4 py-2 text-[var(--color-text-primary)] font-mono text-sm focus:border-[var(--color-primary-border)] focus:ring-2 focus:ring-[#38BDF8]/50"></textarea>
+                            <input type="file" @change="handleFileUpload($event, 'file')" multiple
+                                   class="mt-2 text-sm text-[var(--color-text-secondary)] file:mr-4 file:py-2 file:px-4 file:rounded file:border-0 file:text-sm file:font-semibold file:bg-[var(--color-primary)] file:text-white hover:file:bg-[var(--color-primary-hover)]">
+                        </div>
+                    </div>
+                    </div>
+                    
+                    <div class="flex justify-end space-x-4 p-8 pt-6 border-t border-[var(--color-primary-border)]/20 bg-[var(--color-bg-secondary)]">
+                        <button @click="showExecuteTaskModal = false; selectedTaskForExecution = null; executionParameters = {}; executionParametersText = ''; executionMultimedia = {images: '', videos: '', audios: '', files: ''}; executeModalTab = 'parameters'" 
                                 class="px-4 py-2 bg-[var(--color-bg-primary)] hover:bg-[#0A0E1A] text-[var(--color-text-primary)] rounded-lg transition-colors">
                             Cancel
                         </button>
@@ -433,6 +505,13 @@ <h3 class="text-2xl font-semibold text-[var(--color-text-primary)]">Execute Task
                 selectedTaskForExecution: null,
                 executionParameters: {},
                 executionParametersText: '',
+                executionMultimedia: {
+                    images: '',
+                    videos: '',
+                    audios: '',
+                    files: ''
+                },
+                executeModalTab: 'parameters',
                 modelsConfig: [],
                 hasModels: false,
                 hasMCPModels: false,
@@ -528,20 +607,36 @@ <h3 class="text-2xl font-semibold text-[var(--color-text-primary)]">Execute Task
                     // Parse parameters from text
                     this.executionParameters = this.parseParameters(this.executionParametersText);
                     
+                    // Parse multimedia from text (split by newlines, filter empty)
+                    const parseMultimedia = (text) => {
+                        if (!text || !text.trim()) return [];
+                        return text.split('\n')
+                            .map(line => line.trim())
+                            .filter(line => line.length > 0);
+                    };
+                    
+                    const requestBody = {
+                        task_id: this.selectedTaskForExecution.id,
+                        parameters: this.executionParameters,
+                        images: parseMultimedia(this.executionMultimedia.images),
+                        videos: parseMultimedia(this.executionMultimedia.videos),
+                        audios: parseMultimedia(this.executionMultimedia.audios),
+                        files: parseMultimedia(this.executionMultimedia.files)
+                    };
+                    
                     try {
                         const response = await fetch('/api/agent/jobs/execute', {
                             method: 'POST',
                             headers: { 'Content-Type': 'application/json' },
-                            body: JSON.stringify({ 
-                                task_id: this.selectedTaskForExecution.id, 
-                                parameters: this.executionParameters 
-                            })
+                            body: JSON.stringify(requestBody)
                         });
                         if (response.ok) {
                             this.showExecuteTaskModal = false;
                             this.selectedTaskForExecution = null;
                             this.executionParameters = {};
                             this.executionParametersText = '';
+                            this.executionMultimedia = {images: '', videos: '', audios: '', files: ''};
+                            this.executeModalTab = 'parameters';
                             this.fetchJobs();
                         } else {
                             const error = await response.json();
@@ -552,6 +647,34 @@ <h3 class="text-2xl font-semibold text-[var(--color-text-primary)]">Execute Task
                         alert('Failed to execute task: ' + error.message);
                     }
                 },
+                
+                handleFileUpload(event, type) {
+                    const files = event.target.files;
+                    if (!files || files.length === 0) return;
+                    
+                    const dataURIs = [];
+                    let processed = 0;
+                    
+                    for (let i = 0; i < files.length; i++) {
+                        const file = files[i];
+                        const reader = new FileReader();
+                        
+                        reader.onload = (e) => {
+                            const dataURI = e.target.result;
+                            dataURIs.push(dataURI);
+                            processed++;
+                            
+                            if (processed === files.length) {
+                                // Append to existing content
+                                const current = this.executionMultimedia[type + 's'] || '';
+                                const newContent = current ? current + '\n' + dataURIs.join('\n') : dataURIs.join('\n');
+                                this.executionMultimedia[type + 's'] = newContent;
+                            }
+                        };
+                        
+                        reader.readAsDataURL(file);
+                    }
+                },
 
                 async deleteTask(taskId) {
                     if (!confirm('Are you sure you want to delete this task?')) return;
diff --git a/core/http/views/agent-task-details.html b/core/http/views/agent-task-details.html
index 07ede2085597..e3c257ef7ced 100644
--- a/core/http/views/agent-task-details.html
+++ b/core/http/views/agent-task-details.html
@@ -165,6 +165,59 @@ <h2 class="text-2xl font-semibold text-[#E5E7EB] mb-6">Cron Schedule (Optional)<
                     </div>
                 </div>
 
+                <!-- Multimedia Sources Configuration -->
+                <div class="bg-[#1E293B] border border-[#38BDF8]/20 rounded-xl p-8">
+                    <h2 class="text-2xl font-semibold text-[#E5E7EB] mb-6">Multimedia Sources (Optional)</h2>
+                    <p class="text-sm text-[#94A3B8] mb-4">
+                        Configure multimedia sources (images, videos, audios, files) to fetch when cron jobs execute. 
+                        Each source can have custom headers for authentication/authorization. These will be fetched and included in the job execution.
+                    </p>
+                    <div class="space-y-4">
+                        <template x-for="(source, index) in taskForm.multimedia_sources" :key="index">
+                            <div class="bg-[#101827] p-4 rounded border border-[#38BDF8]/10">
+                                <div class="flex justify-between items-center mb-4">
+                                    <h3 class="text-lg font-semibold text-[#E5E7EB]">Multimedia Source <span x-text="index + 1"></span></h3>
+                                    <button type="button" @click="taskForm.multimedia_sources.splice(index, 1)" 
+                                            class="text-red-400 hover:text-red-300">
+                                        <i class="fas fa-trash"></i>
+                                    </button>
+                                </div>
+                                <div class="space-y-4">
+                                    <div>
+                                        <label class="block text-[#E5E7EB] mb-2">Type *</label>
+                                        <select x-model="source.type" required
+                                                class="w-full bg-[#0A0E1A] border border-[#38BDF8]/20 rounded px-4 py-2 text-[#E5E7EB] focus:border-[#38BDF8] focus:ring-2 focus:ring-[#38BDF8]/50">
+                                            <option value="">Select type...</option>
+                                            <option value="image">Image</option>
+                                            <option value="video">Video</option>
+                                            <option value="audio">Audio</option>
+                                            <option value="file">File</option>
+                                        </select>
+                                    </div>
+                                    <div>
+                                        <label class="block text-[#E5E7EB] mb-2">URL *</label>
+                                        <input type="url" x-model="source.url" required
+                                               placeholder="https://example.com/image.png"
+                                               class="w-full bg-[#0A0E1A] border border-[#38BDF8]/20 rounded px-4 py-2 text-[#E5E7EB] focus:border-[#38BDF8] focus:ring-2 focus:ring-[#38BDF8]/50">
+                                        <p class="text-xs text-[#94A3B8] mt-1">URL where multimedia content will be fetched from</p>
+                                    </div>
+                                    <div>
+                                        <label class="block text-[#E5E7EB] mb-2">Headers (JSON)</label>
+                                        <textarea x-model="source.headers_json" rows="3"
+                                                  placeholder='{"Authorization": "Bearer token"}'
+                                                  class="w-full bg-[#0A0E1A] border border-[#38BDF8]/20 rounded px-4 py-2 text-[#E5E7EB] font-mono text-sm focus:border-[#38BDF8] focus:ring-2 focus:ring-[#38BDF8]/50"></textarea>
+                                        <p class="text-xs text-[#94A3B8] mt-1">Custom headers for the HTTP request (e.g., Authorization)</p>
+                                    </div>
+                                </div>
+                            </div>
+                        </template>
+                        <button type="button" @click="addMultimediaSource()" 
+                                class="w-full bg-[#101827] hover:bg-[#0A0E1A] border border-[#38BDF8]/20 border-dashed rounded-lg p-4 text-[#94A3B8] hover:text-[#E5E7EB] transition-colors">
+                            <i class="fas fa-plus mr-2"></i>Add Multimedia Source
+                        </button>
+                    </div>
+                </div>
+
                 <!-- Webhook Configuration -->
                 <div class="bg-[#1E293B] border border-[#38BDF8]/20 rounded-xl p-8">
                     <h2 class="text-2xl font-semibold text-[#E5E7EB] mb-6">Webhooks (Optional)</h2>
@@ -350,6 +403,33 @@ <h3 class="text-lg font-semibold text-[#E5E7EB] mb-3 flex items-center">
                         </p>
                     </div>
                     
+                    <!-- Execute Task with Multimedia -->
+                    <div>
+                        <h3 class="text-lg font-semibold text-[#E5E7EB] mb-3 flex items-center">
+                            <i class="fas fa-images text-[#38BDF8] mr-2"></i>
+                            Execute Task with Multimedia (Images)
+                        </h3>
+                        <div class="bg-[#101827] border border-[#38BDF8]/10 rounded-lg p-4">
+                            <pre class="text-xs text-[#E5E7EB] font-mono overflow-x-auto"><code>curl -X POST {{ .BaseURL }}api/agent/jobs/execute \
+  -H "Content-Type: application/json" \
+  -H "Authorization: Bearer YOUR_API_KEY" \
+  -d '{
+    "task_id": "<span x-text="task ? task.id : 'task-uuid'"></span>",
+    "parameters": {
+      "user_name": "Alice",
+      "task_description": "Analyze this image"
+    },
+    "images": [
+      "https://example.com/image.png",
+      "data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAAAEAAAABCAYAAAAfFcSJAAAADUlEQVR42mNk+M9QDwADhgGAWjR9awAAAABJRU5ErkJggg=="
+    ]
+  }'</code></pre>
+                        </div>
+                        <p class="text-xs text-[#94A3B8] mt-2">
+                            You can provide multimedia content as URLs or base64-encoded data URIs. Supported types: <code class="bg-[#101827] px-1 py-0.5 rounded text-[#38BDF8]">images</code>, <code class="bg-[#101827] px-1 py-0.5 rounded text-[#38BDF8]">videos</code>, <code class="bg-[#101827] px-1 py-0.5 rounded text-[#38BDF8]">audios</code>, and <code class="bg-[#101827] px-1 py-0.5 rounded text-[#38BDF8]">files</code>.
+                        </p>
+                    </div>
+                    
                     <!-- Check Job Status -->
                     <div>
                         <h3 class="text-lg font-semibold text-[#E5E7EB] mb-3 flex items-center">
@@ -478,23 +558,42 @@ <h2 class="text-2xl font-semibold text-[#E5E7EB]">Job History</h2>
     <!-- Execute Task Modal -->
     <div x-show="showExecuteTaskModal" 
          x-cloak
-         @click.away="showExecuteTaskModal = false; executionParameters = {}; executionParametersText = ''"
+         @click.away="showExecuteTaskModal = false; executionParameters = {}; executionParametersText = ''; executionMultimedia = {images: '', videos: '', audios: '', files: ''}; executeModalTab = 'parameters'"
          class="fixed inset-0 bg-black/50 flex items-center justify-center z-50">
-        <div class="bg-[#1E293B] border border-[#38BDF8]/20 rounded-xl p-8 max-w-2xl w-full mx-4">
-            <div class="flex justify-between items-center mb-6">
+        <div class="bg-[#1E293B] border border-[#38BDF8]/20 rounded-xl max-w-2xl w-full mx-4 max-h-[90vh] flex flex-col" @click.stop>
+            <div class="flex justify-between items-center p-8 pb-6 border-b border-[#38BDF8]/20">
                 <h3 class="text-2xl font-semibold text-[#E5E7EB]">Execute Task</h3>
-                <button @click="showExecuteTaskModal = false; executionParameters = {}; executionParametersText = ''" 
+                <button @click="showExecuteTaskModal = false; executionParameters = {}; executionParametersText = ''; executionMultimedia = {images: '', videos: '', audios: '', files: ''}; executeModalTab = 'parameters'" 
                         class="text-[#94A3B8] hover:text-[#E5E7EB]">
                     <i class="fas fa-times text-xl"></i>
                 </button>
             </div>
             <template x-if="task">
-                <div class="space-y-4">
+                <div class="flex flex-col flex-1 min-h-0">
+                    <div class="flex-1 overflow-y-auto px-8 py-6 space-y-4">
                     <div>
                         <label class="block text-sm font-medium text-[#E5E7EB] mb-2">Task</label>
                         <div class="text-[#94A3B8]" x-text="task.name"></div>
                     </div>
-                    <div>
+                    
+                    <!-- Tabs for Parameters and Multimedia -->
+                    <div class="border-b border-[#38BDF8]/20">
+                        <div class="flex space-x-4">
+                            <button @click="executeModalTab = 'parameters'" 
+                                    :class="executeModalTab === 'parameters' ? 'border-b-2 border-[#38BDF8] text-[#38BDF8]' : 'text-[#94A3B8] hover:text-[#E5E7EB]'"
+                                    class="px-4 py-2 font-medium transition-colors">
+                                Parameters
+                            </button>
+                            <button @click="executeModalTab = 'multimedia'" 
+                                    :class="executeModalTab === 'multimedia' ? 'border-b-2 border-[#38BDF8] text-[#38BDF8]' : 'text-[#94A3B8] hover:text-[#E5E7EB]'"
+                                    class="px-4 py-2 font-medium transition-colors">
+                                Multimedia
+                            </button>
+                        </div>
+                    </div>
+                    
+                    <!-- Parameters Tab -->
+                    <div x-show="executeModalTab === 'parameters'">
                         <label class="block text-sm font-medium text-[#E5E7EB] mb-2">Parameters</label>
                         <p class="text-xs text-[#94A3B8] mb-3">
                             Enter parameters as key-value pairs (one per line, format: key=value). 
@@ -508,8 +607,61 @@ <h3 class="text-2xl font-semibold text-[#E5E7EB]">Execute Task</h3>
                             Example: <code class="bg-[#101827] px-1 py-0.5 rounded text-[#38BDF8]">user_name=Alice</code>
                         </p>
                     </div>
-                    <div class="flex justify-end space-x-4">
-                        <button @click="showExecuteTaskModal = false; executionParameters = {}; executionParametersText = ''" 
+                    
+                    <!-- Multimedia Tab -->
+                    <div x-show="executeModalTab === 'multimedia'" class="space-y-4">
+                        <p class="text-xs text-[#94A3B8] mb-3">
+                            Provide multimedia content as URLs or base64-encoded data URIs. You can also upload files which will be converted to base64.
+                        </p>
+                        
+                        <!-- Images -->
+                        <div>
+                            <label class="block text-sm font-medium text-[#E5E7EB] mb-2">Images</label>
+                            <textarea x-model="executionMultimedia.images" 
+                                      rows="3"
+                                      placeholder="https://example.com/image.png&#10;data:image/png;base64,iVBORw0KG..."
+                                      class="w-full bg-[#101827] border border-[#38BDF8]/20 rounded px-4 py-2 text-[#E5E7EB] font-mono text-sm focus:border-[#38BDF8] focus:ring-2 focus:ring-[#38BDF8]/50"></textarea>
+                            <input type="file" @change="handleFileUpload($event, 'image')" accept="image/*" multiple
+                                   class="mt-2 text-sm text-[#94A3B8] file:mr-4 file:py-2 file:px-4 file:rounded file:border-0 file:text-sm file:font-semibold file:bg-[#38BDF8] file:text-white hover:file:bg-[#38BDF8]/80">
+                        </div>
+                        
+                        <!-- Videos -->
+                        <div>
+                            <label class="block text-sm font-medium text-[#E5E7EB] mb-2">Videos</label>
+                            <textarea x-model="executionMultimedia.videos" 
+                                      rows="3"
+                                      placeholder="https://example.com/video.mp4&#10;data:video/mp4;base64,..."
+                                      class="w-full bg-[#101827] border border-[#38BDF8]/20 rounded px-4 py-2 text-[#E5E7EB] font-mono text-sm focus:border-[#38BDF8] focus:ring-2 focus:ring-[#38BDF8]/50"></textarea>
+                            <input type="file" @change="handleFileUpload($event, 'video')" accept="video/*" multiple
+                                   class="mt-2 text-sm text-[#94A3B8] file:mr-4 file:py-2 file:px-4 file:rounded file:border-0 file:text-sm file:font-semibold file:bg-[#38BDF8] file:text-white hover:file:bg-[#38BDF8]/80">
+                        </div>
+                        
+                        <!-- Audios -->
+                        <div>
+                            <label class="block text-sm font-medium text-[#E5E7EB] mb-2">Audios</label>
+                            <textarea x-model="executionMultimedia.audios" 
+                                      rows="3"
+                                      placeholder="https://example.com/audio.mp3&#10;data:audio/mpeg;base64,..."
+                                      class="w-full bg-[#101827] border border-[#38BDF8]/20 rounded px-4 py-2 text-[#E5E7EB] font-mono text-sm focus:border-[#38BDF8] focus:ring-2 focus:ring-[#38BDF8]/50"></textarea>
+                            <input type="file" @change="handleFileUpload($event, 'audio')" accept="audio/*" multiple
+                                   class="mt-2 text-sm text-[#94A3B8] file:mr-4 file:py-2 file:px-4 file:rounded file:border-0 file:text-sm file:font-semibold file:bg-[#38BDF8] file:text-white hover:file:bg-[#38BDF8]/80">
+                        </div>
+                        
+                        <!-- Files -->
+                        <div>
+                            <label class="block text-sm font-medium text-[#E5E7EB] mb-2">Files</label>
+                            <textarea x-model="executionMultimedia.files" 
+                                      rows="3"
+                                      placeholder="https://example.com/file.pdf&#10;data:application/pdf;base64,..."
+                                      class="w-full bg-[#101827] border border-[#38BDF8]/20 rounded px-4 py-2 text-[#E5E7EB] font-mono text-sm focus:border-[#38BDF8] focus:ring-2 focus:ring-[#38BDF8]/50"></textarea>
+                            <input type="file" @change="handleFileUpload($event, 'file')" multiple
+                                   class="mt-2 text-sm text-[#94A3B8] file:mr-4 file:py-2 file:px-4 file:rounded file:border-0 file:text-sm file:font-semibold file:bg-[#38BDF8] file:text-white hover:file:bg-[#38BDF8]/80">
+                        </div>
+                    </div>
+                    </div>
+                    
+                    <div class="flex justify-end space-x-4 p-8 pt-6 border-t border-[#38BDF8]/20 bg-[#1E293B]">
+                        <button @click="showExecuteTaskModal = false; executionParameters = {}; executionParametersText = ''; executionMultimedia = {images: '', videos: '', audios: '', files: ''}; executeModalTab = 'parameters'" 
                                 class="px-4 py-2 bg-[#101827] hover:bg-[#0A0E1A] text-[#E5E7EB] rounded-lg transition-colors">
                             Cancel
                         </button>
@@ -533,6 +685,13 @@ <h3 class="text-2xl font-semibold text-[#E5E7EB]">Execute Task</h3>
                 showExecuteTaskModal: false,
                 executionParameters: {},
                 executionParametersText: '',
+                executionMultimedia: {
+                    images: '',
+                    videos: '',
+                    audios: '',
+                    files: ''
+                },
+                executeModalTab: 'parameters',
                 isNewTask: false,
                 isEditMode: false,
                 taskForm: {
@@ -543,7 +702,8 @@ <h3 class="text-2xl font-semibold text-[#E5E7EB]">Execute Task</h3>
                     enabled: true,
                     cron: '',
                     cron_parameters: {},
-                    webhooks: []
+                    webhooks: [],
+                    multimedia_sources: []
                 },
                 cronError: '',
                 cronParametersText: '',
@@ -600,6 +760,15 @@ <h3 class="text-2xl font-semibold text-[#E5E7EB]">Execute Task</h3>
                             // Note: Legacy fields (webhook_url, webhook_auth, webhook_template) should be migrated
                             // by the backend, so we don't need to handle them here
                             
+                            // Handle multimedia sources
+                            let multimediaSources = [];
+                            if (this.task.multimedia_sources && Array.isArray(this.task.multimedia_sources) && this.task.multimedia_sources.length > 0) {
+                                multimediaSources = this.task.multimedia_sources.map(ms => ({
+                                    ...ms,
+                                    headers_json: JSON.stringify(ms.headers || {}, null, 2)
+                                }));
+                            }
+                            
                             // Convert cron_parameters to text format
                             let cronParamsText = '';
                             if (this.task.cron_parameters && Object.keys(this.task.cron_parameters).length > 0) {
@@ -616,7 +785,8 @@ <h3 class="text-2xl font-semibold text-[#E5E7EB]">Execute Task</h3>
                                 enabled: this.task.enabled !== undefined ? this.task.enabled : true,
                                 cron: this.task.cron || '',
                                 cron_parameters: this.task.cron_parameters || {},
-                                webhooks: webhooks
+                                webhooks: webhooks,
+                                multimedia_sources: multimediaSources
                             };
                             this.cronParametersText = cronParamsText;
                         } else {
@@ -667,6 +837,19 @@ <h3 class="text-2xl font-semibold text-[#E5E7EB]">Execute Task</h3>
                     this.taskForm.webhooks.push(webhook);
                 },
 
+                addMultimediaSource() {
+                    const source = {
+                        type: '',
+                        url: '',
+                        headers: {},
+                        headers_json: '{}'
+                    };
+                    if (!this.taskForm.multimedia_sources) {
+                        this.taskForm.multimedia_sources = [];
+                    }
+                    this.taskForm.multimedia_sources.push(source);
+                },
+
                 updateCronParameters() {
                     // Parse text input into parameters object
                     const params = {};
@@ -791,6 +974,8 @@ <h3 class="text-2xl font-semibold text-[#E5E7EB]">Execute Task</h3>
                 showExecuteModal() {
                     this.executionParameters = {};
                     this.executionParametersText = '';
+                    this.executionMultimedia = {images: '', videos: '', audios: '', files: ''};
+                    this.executeModalTab = 'parameters';
                     this.showExecuteTaskModal = true;
                 },
 
@@ -821,19 +1006,35 @@ <h3 class="text-2xl font-semibold text-[#E5E7EB]">Execute Task</h3>
                     // Parse parameters from text
                     this.executionParameters = this.parseParameters(this.executionParametersText);
                     
+                    // Parse multimedia from text (split by newlines, filter empty)
+                    const parseMultimedia = (text) => {
+                        if (!text || !text.trim()) return [];
+                        return text.split('\n')
+                            .map(line => line.trim())
+                            .filter(line => line.length > 0);
+                    };
+                    
+                    const requestBody = {
+                        task_id: this.task.id,
+                        parameters: this.executionParameters,
+                        images: parseMultimedia(this.executionMultimedia.images),
+                        videos: parseMultimedia(this.executionMultimedia.videos),
+                        audios: parseMultimedia(this.executionMultimedia.audios),
+                        files: parseMultimedia(this.executionMultimedia.files)
+                    };
+                    
                     try {
                         const response = await fetch('/api/agent/jobs/execute', {
                             method: 'POST',
                             headers: { 'Content-Type': 'application/json' },
-                            body: JSON.stringify({ 
-                                task_id: this.task.id, 
-                                parameters: this.executionParameters 
-                            })
+                            body: JSON.stringify(requestBody)
                         });
                         if (response.ok) {
                             this.showExecuteTaskModal = false;
                             this.executionParameters = {};
                             this.executionParametersText = '';
+                            this.executionMultimedia = {images: '', videos: '', audios: '', files: ''};
+                            this.executeModalTab = 'parameters';
                             this.fetchJobs();
                         } else {
                             const error = await response.json();
@@ -844,6 +1045,34 @@ <h3 class="text-2xl font-semibold text-[#E5E7EB]">Execute Task</h3>
                         alert('Failed to execute task: ' + error.message);
                     }
                 },
+                
+                handleFileUpload(event, type) {
+                    const files = event.target.files;
+                    if (!files || files.length === 0) return;
+                    
+                    const dataURIs = [];
+                    let processed = 0;
+                    
+                    for (let i = 0; i < files.length; i++) {
+                        const file = files[i];
+                        const reader = new FileReader();
+                        
+                        reader.onload = (e) => {
+                            const dataURI = e.target.result;
+                            dataURIs.push(dataURI);
+                            processed++;
+                            
+                            if (processed === files.length) {
+                                // Append to existing content
+                                const current = this.executionMultimedia[type + 's'] || '';
+                                const newContent = current ? current + '\n' + dataURIs.join('\n') : dataURIs.join('\n');
+                                this.executionMultimedia[type + 's'] = newContent;
+                            }
+                        };
+                        
+                        reader.readAsDataURL(file);
+                    }
+                },
 
                 async deleteTask() {
                     if (!confirm('Are you sure you want to delete this task? This will also delete all associated jobs.')) return;
diff --git a/core/http/views/backends.html b/core/http/views/backends.html
index 815823f74a9c..8a3928a470ce 100644
--- a/core/http/views/backends.html
+++ b/core/http/views/backends.html
@@ -75,13 +75,14 @@ <h3 class="text-xl font-semibold text-[#E5E7EB] mb-4 flex items-center">
                         Find Backend Components
                     </h3>
                     <div class="relative">
-                        <div class="absolute inset-y-0 start-0 flex items-center ps-4 pointer-events-none">
+                        <div class="absolute inset-y-0 start-0 flex items-center ps-4 pointer-events-none z-10">
                             <i class="fas fa-search text-[#94A3B8]"></i>
                         </div>
                         <input 
                             x-model="searchTerm"
                             @input.debounce.500ms="fetchBackends()"
-                            class="input w-full pl-12 pr-16 py-4" 
+                            class="input w-full pr-16 py-4" 
+                            style="padding-left: 3.5rem !important;"
                             type="search" 
                             placeholder="Search backends by name, description or type...">
                         <span class="absolute right-4 top-4" x-show="loading">
diff --git a/core/http/views/chat.html b/core/http/views/chat.html
index 232ae1c0e3fe..cbc2539ae505 100644
--- a/core/http/views/chat.html
+++ b/core/http/views/chat.html
@@ -717,9 +717,10 @@ <h2 class="text-xs font-semibold text-[var(--color-text-secondary)] uppercase tr
                 type="text"
                 x-model="searchQuery"
                 placeholder="Search conversations..."
-                class="w-full bg-[var(--color-bg-primary)] text-[var(--color-text-primary)] border border-[var(--color-bg-secondary)] focus:border-[var(--color-primary-border)] focus:ring-1 focus:ring-[var(--color-primary)]/50 rounded px-2 py-1.5 pl-7 text-xs placeholder-[var(--color-text-secondary)]"
+                class="w-full bg-[var(--color-bg-primary)] text-[var(--color-text-primary)] border border-[var(--color-bg-secondary)] focus:border-[var(--color-primary-border)] focus:ring-1 focus:ring-[var(--color-primary)]/50 rounded py-1.5 pr-2 text-xs placeholder-[var(--color-text-secondary)]"
+                style="padding-left: 2rem !important;"
               />
-              <i class="fa-solid fa-search absolute left-2 top-1/2 transform -translate-y-1/2 text-[var(--color-text-secondary)] text-xs"></i>
+              <i class="fa-solid fa-search absolute left-2.5 top-1/2 transform -translate-y-1/2 text-[var(--color-text-secondary)] text-xs pointer-events-none z-10"></i>
               <button
                 x-show="searchQuery.length > 0"
                 @click="searchQuery = ''"
diff --git a/core/http/views/models.html b/core/http/views/models.html
index 6e84472f25cc..d83cee2d853f 100644
--- a/core/http/views/models.html
+++ b/core/http/views/models.html
@@ -80,13 +80,14 @@ <h3 class="text-xl font-semibold text-[var(--color-text-primary)] mb-4 flex item
                         Find Your Perfect Model
                     </h3>
                     <div class="relative">
-                        <div class="absolute inset-y-0 start-0 flex items-center ps-4 pointer-events-none">
+                        <div class="absolute inset-y-0 start-0 flex items-center ps-4 pointer-events-none z-10">
                             <i class="fas fa-search text-[var(--color-text-secondary)]"></i>
                         </div>
                         <input
                             x-model="searchTerm"
                             @input.debounce.500ms="fetchModels()"
-                            class="input w-full pl-12 pr-16 py-4"
+                            class="input w-full pr-16 py-4"
+                            style="padding-left: 3.5rem !important;"
                             type="search"
                             placeholder="Search models by name, tag, or description...">
                         <span class="absolute right-4 top-4" x-show="loading">
diff --git a/core/http/views/text2image.html b/core/http/views/text2image.html
index b8e28c8ba1d6..69f0a925400e 100644
--- a/core/http/views/text2image.html
+++ b/core/http/views/text2image.html
@@ -56,7 +56,7 @@ <h2 class="h3 mb-6">Generate an Image</h2>
                 <input id="image-model" type="hidden" value="{{.Model}}">
                 <form id="genimage" action="text2image/{{.Model}}" method="get" class="mb-8">
                     <div class="relative">
-                        <div class="absolute inset-y-0 left-0 flex items-center pl-4">
+                        <div class="absolute inset-y-0 left-0 flex items-center pl-4 pointer-events-none z-10">
                             <i class="fas fa-magic text-[var(--color-primary)]"></i>
                         </div>
                         <input
@@ -65,7 +65,8 @@ <h2 class="h3 mb-6">Generate an Image</h2>
                             name="input"
                             placeholder="Describe the image you want to generate..."
                             autocomplete="off"
-                            class="input w-full pl-12 pr-12 py-4 text-lg"
+                            class="input w-full pr-12 py-4 text-lg"
+                            style="padding-left: 3.5rem !important;"
                             required
                         />
                         <span id="loader" class="my-2 loader absolute right-4 top-4 hidden">
diff --git a/core/schema/agent_jobs.go b/core/schema/agent_jobs.go
index 4d867416adab..b0453a293be9 100644
--- a/core/schema/agent_jobs.go
+++ b/core/schema/agent_jobs.go
@@ -27,6 +27,10 @@ type Task struct {
 	// - {{.Status}} - Job status string
 	Webhooks []WebhookConfig `json:"webhooks,omitempty"` // Webhook configs for job completion notifications
 
+	// Multimedia sources (for cron jobs)
+	// URLs to fetch multimedia content from when cron job executes
+	// Each source can have custom headers for authentication/authorization
+	MultimediaSources []MultimediaSourceConfig `json:"multimedia_sources,omitempty"` // Multimedia sources for cron jobs
 }
 
 // WebhookConfig represents configuration for sending webhook notifications
@@ -44,6 +48,14 @@ type WebhookConfig struct {
 	// - {{.Status}} - Job status string
 }
 
+// MultimediaSourceConfig represents configuration for fetching multimedia content
+// Used in cron jobs to periodically fetch multimedia from URLs with custom headers
+type MultimediaSourceConfig struct {
+	Type    string            `json:"type"`              // "image", "video", "audio", "file"
+	URL     string            `json:"url"`              // URL to fetch from
+	Headers map[string]string `json:"headers,omitempty"` // Custom headers for HTTP request (e.g., Authorization)
+}
+
 // JobStatus represents the status of a job
 type JobStatus string
 
@@ -75,6 +87,13 @@ type Job struct {
 
 	// Execution traces (reasoning, tool calls, tool results)
 	Traces []JobTrace `json:"traces,omitempty"`
+
+	// Multimedia content (for manual execution)
+	// Can contain URLs or base64-encoded data URIs
+	Images []string `json:"images,omitempty"` // List of image URLs or base64 strings
+	Videos []string `json:"videos,omitempty"` // List of video URLs or base64 strings
+	Audios []string `json:"audios,omitempty"` // List of audio URLs or base64 strings
+	Files  []string `json:"files,omitempty"`  // List of file URLs or base64 strings
 }
 
 // JobTrace represents a single execution trace entry
@@ -90,6 +109,12 @@ type JobTrace struct {
 type JobExecutionRequest struct {
 	TaskID     string            `json:"task_id"`    // Required
 	Parameters map[string]string `json:"parameters"` // Optional, for templating
+	// Multimedia content (optional, for manual execution)
+	// Can contain URLs or base64-encoded data URIs
+	Images []string `json:"images,omitempty"` // List of image URLs or base64 strings
+	Videos []string `json:"videos,omitempty"` // List of video URLs or base64 strings
+	Audios []string `json:"audios,omitempty"` // List of audio URLs or base64 strings
+	Files  []string `json:"files,omitempty"`  // List of file URLs or base64 strings
 }
 
 // JobExecutionResponse represents the response after creating a job
diff --git a/core/services/agent_jobs.go b/core/services/agent_jobs.go
index 1702e6d6d885..553ef1c61a62 100644
--- a/core/services/agent_jobs.go
+++ b/core/services/agent_jobs.go
@@ -3,6 +3,7 @@ package services
 import (
 	"bytes"
 	"context"
+	"encoding/base64"
 	"encoding/json"
 	"errors"
 	"fmt"
@@ -366,7 +367,13 @@ func (s *AgentJobService) buildPrompt(templateStr string, params map[string]stri
 }
 
 // ExecuteJob creates and queues a job for execution
-func (s *AgentJobService) ExecuteJob(taskID string, params map[string]string, triggeredBy string) (string, error) {
+// multimedia can be nil for backward compatibility
+func (s *AgentJobService) ExecuteJob(taskID string, params map[string]string, triggeredBy string, multimedia *struct {
+	Images []string
+	Videos []string
+	Audios []string
+	Files  []string
+}) (string, error) {
 	task := s.tasks.Get(taskID)
 	if task.ID == "" {
 		return "", fmt.Errorf("task not found: %s", taskID)
@@ -388,6 +395,52 @@ func (s *AgentJobService) ExecuteJob(taskID string, params map[string]string, tr
 		TriggeredBy: triggeredBy,
 	}
 
+	// Handle multimedia: merge task-level (for cron) and job-level (for manual execution)
+	if triggeredBy == "cron" && len(task.MultimediaSources) > 0 {
+		// Fetch multimedia from task sources
+		job.Images = []string{}
+		job.Videos = []string{}
+		job.Audios = []string{}
+		job.Files = []string{}
+
+		for _, source := range task.MultimediaSources {
+			// Fetch content from URL with custom headers
+			dataURI, err := s.fetchMultimediaFromURL(source.URL, source.Headers, source.Type)
+			if err != nil {
+				log.Warn().Err(err).Str("url", source.URL).Str("type", source.Type).Msg("Failed to fetch multimedia from task source")
+				continue
+			}
+
+			// Add to appropriate slice based on type
+			switch source.Type {
+			case "image":
+				job.Images = append(job.Images, dataURI)
+			case "video":
+				job.Videos = append(job.Videos, dataURI)
+			case "audio":
+				job.Audios = append(job.Audios, dataURI)
+			case "file":
+				job.Files = append(job.Files, dataURI)
+			}
+		}
+	}
+
+	// Override with job-level multimedia if provided (manual execution takes precedence)
+	if multimedia != nil {
+		if len(multimedia.Images) > 0 {
+			job.Images = multimedia.Images
+		}
+		if len(multimedia.Videos) > 0 {
+			job.Videos = multimedia.Videos
+		}
+		if len(multimedia.Audios) > 0 {
+			job.Audios = multimedia.Audios
+		}
+		if len(multimedia.Files) > 0 {
+			job.Files = multimedia.Files
+		}
+	}
+
 	// Store job
 	s.jobs.Set(jobID, job)
 
@@ -512,6 +565,108 @@ func (s *AgentJobService) DeleteJob(id string) error {
 	return nil
 }
 
+type multimediaContent struct {
+	url       string
+	mediaType string
+}
+
+func (mu multimediaContent) URL() string {
+	return mu.url
+}
+
+// fetchMultimediaFromURL fetches multimedia content from a URL with custom headers
+// and converts it to a data URI string
+func (s *AgentJobService) fetchMultimediaFromURL(url string, headers map[string]string, mediaType string) (string, error) {
+	// Create HTTP request
+	req, err := http.NewRequest("GET", url, nil)
+	if err != nil {
+		return "", fmt.Errorf("failed to create request: %w", err)
+	}
+
+	// Set custom headers
+	for key, value := range headers {
+		req.Header.Set(key, value)
+	}
+
+	// Execute request
+	client := &http.Client{Timeout: 30 * time.Second}
+	resp, err := client.Do(req)
+	if err != nil {
+		return "", fmt.Errorf("failed to fetch URL: %w", err)
+	}
+	defer resp.Body.Close()
+
+	if resp.StatusCode < 200 || resp.StatusCode >= 300 {
+		return "", fmt.Errorf("HTTP error: %d", resp.StatusCode)
+	}
+
+	// Read content
+	data, err := io.ReadAll(resp.Body)
+	if err != nil {
+		return "", fmt.Errorf("failed to read response: %w", err)
+	}
+
+	// Encode to base64
+	encoded := base64.StdEncoding.EncodeToString(data)
+
+	// Determine MIME type
+	mimeType := s.getMimeTypeForMediaType(mediaType)
+	if contentType := resp.Header.Get("Content-Type"); contentType != "" {
+		mimeType = contentType
+	}
+
+	// Return as data URI
+	return fmt.Sprintf("data:%s;base64,%s", mimeType, encoded), nil
+}
+
+// getMimeTypeForMediaType returns the default MIME type for a media type
+func (s *AgentJobService) getMimeTypeForMediaType(mediaType string) string {
+	switch mediaType {
+	case "image":
+		return "image/png"
+	case "video":
+		return "video/mp4"
+	case "audio":
+		return "audio/mpeg"
+	case "file":
+		return "application/octet-stream"
+	default:
+		return "application/octet-stream"
+	}
+}
+
+// convertToMultimediaContent converts a slice of strings (URLs or base64) to multimediaContent objects
+func (s *AgentJobService) convertToMultimediaContent(items []string, mediaType string) ([]cogito.Multimedia, error) {
+	result := make([]cogito.Multimedia, 0, len(items))
+
+	for _, item := range items {
+		if item == "" {
+			continue
+		}
+
+		// Check if it's already a data URI
+		if strings.HasPrefix(item, "data:") {
+			result = append(result, multimediaContent{url: item, mediaType: mediaType})
+			continue
+		}
+
+		// Check if it's a URL
+		if strings.HasPrefix(item, "http://") || strings.HasPrefix(item, "https://") {
+			// Pass URL directly to cogito (it handles fetching)
+			result = append(result, multimediaContent{url: item, mediaType: mediaType})
+			continue
+		}
+
+		// Assume it's base64 without data URI prefix
+		// Add appropriate prefix based on media type
+		mimeType := s.getMimeTypeForMediaType(mediaType)
+		dataURI := fmt.Sprintf("data:%s;base64,%s", mimeType, item)
+		result = append(result, multimediaContent{url: dataURI, mediaType: mediaType})
+	}
+
+	return result, nil
+}
+
 // executeJobInternal executes a job using cogito
 func (s *AgentJobService) executeJobInternal(job schema.Job, task schema.Task, ctx context.Context) error {
 	// Update job status to running
@@ -585,7 +740,51 @@ func (s *AgentJobService) executeJobInternal(job schema.Job, task schema.Task, c
 
 	// Create cogito fragment
 	fragment := cogito.NewEmptyFragment()
-	fragment = fragment.AddMessage("user", prompt)
+
+	// Collect all multimedia content
+	multimediaItems := []cogito.Multimedia{}
+
+	// Convert images
+	if len(job.Images) > 0 {
+		images, err := s.convertToMultimediaContent(job.Images, "image")
+		if err != nil {
+			log.Warn().Err(err).Str("job_id", job.ID).Msg("Failed to convert images")
+		} else {
+			multimediaItems = append(multimediaItems, images...)
+		}
+	}
+
+	// Convert videos
+	if len(job.Videos) > 0 {
+		videos, err := s.convertToMultimediaContent(job.Videos, "video")
+		if err != nil {
+			log.Warn().Err(err).Str("job_id", job.ID).Msg("Failed to convert videos")
+		} else {
+			multimediaItems = append(multimediaItems, videos...)
+		}
+	}
+
+	// Convert audios
+	if len(job.Audios) > 0 {
+		audios, err := s.convertToMultimediaContent(job.Audios, "audio")
+		if err != nil {
+			log.Warn().Err(err).Str("job_id", job.ID).Msg("Failed to convert audios")
+		} else {
+			multimediaItems = append(multimediaItems, audios...)
+		}
+	}
+
+	// Convert files
+	if len(job.Files) > 0 {
+		files, err := s.convertToMultimediaContent(job.Files, "file")
+		if err != nil {
+			log.Warn().Err(err).Str("job_id", job.ID).Msg("Failed to convert files")
+		} else {
+			multimediaItems = append(multimediaItems, files...)
+		}
+	}
+
+	fragment = fragment.AddMessage("user", prompt, multimediaItems...)
 
 	// Get API address and key
 	_, port, err := net.SplitHostPort(s.appConfig.APIAddress)
@@ -845,7 +1044,8 @@ func (s *AgentJobService) ScheduleCronTask(task schema.Task) error {
 	}
 	entryID, err := s.cronScheduler.AddFunc(cronExpr, func() {
 		// Create job for cron execution with configured parameters
-		_, err := s.ExecuteJob(task.ID, cronParams, "cron")
+		// Multimedia will be fetched from task sources in ExecuteJob
+		_, err := s.ExecuteJob(task.ID, cronParams, "cron", nil)
 		if err != nil {
 			log.Error().Err(err).Str("task_id", task.ID).Msg("Failed to execute cron job")
 		}
diff --git a/core/services/agent_jobs_test.go b/core/services/agent_jobs_test.go
index 9851ba1e9cfc..1b902049d5c8 100644
--- a/core/services/agent_jobs_test.go
+++ b/core/services/agent_jobs_test.go
@@ -153,7 +153,7 @@ var _ = Describe("AgentJobService", func() {
 
 		It("should create and queue a job", func() {
 			params := map[string]string{"name": "World"}
-			jobID, err := service.ExecuteJob(taskID, params, "test")
+			jobID, err := service.ExecuteJob(taskID, params, "test", nil)
 			Expect(err).NotTo(HaveOccurred())
 			Expect(jobID).NotTo(BeEmpty())
 
@@ -166,12 +166,12 @@ var _ = Describe("AgentJobService", func() {
 
 		It("should list jobs with filters", func() {
 			params := map[string]string{}
-			jobID1, err := service.ExecuteJob(taskID, params, "test")
+			jobID1, err := service.ExecuteJob(taskID, params, "test", nil)
 			Expect(err).NotTo(HaveOccurred())
 
 			time.Sleep(10 * time.Millisecond) // Ensure different timestamps
 
-			jobID2, err := service.ExecuteJob(taskID, params, "test")
+			jobID2, err := service.ExecuteJob(taskID, params, "test", nil)
 			Expect(err).NotTo(HaveOccurred())
 
 			allJobs := service.ListJobs(nil, nil, 0)
@@ -195,7 +195,7 @@ var _ = Describe("AgentJobService", func() {
 
 		It("should cancel a pending job", func() {
 			params := map[string]string{}
-			jobID, err := service.ExecuteJob(taskID, params, "test")
+			jobID, err := service.ExecuteJob(taskID, params, "test", nil)
 			Expect(err).NotTo(HaveOccurred())
 
 			err = service.CancelJob(jobID)
@@ -208,7 +208,7 @@ var _ = Describe("AgentJobService", func() {
 
 		It("should delete a job", func() {
 			params := map[string]string{}
-			jobID, err := service.ExecuteJob(taskID, params, "test")
+			jobID, err := service.ExecuteJob(taskID, params, "test", nil)
 			Expect(err).NotTo(HaveOccurred())
 
 			err = service.DeleteJob(jobID)
@@ -258,7 +258,7 @@ var _ = Describe("AgentJobService", func() {
 			Expect(err).NotTo(HaveOccurred())
 
 			params := map[string]string{}
-			jobID, err := service.ExecuteJob(taskID, params, "test")
+			jobID, err := service.ExecuteJob(taskID, params, "test", nil)
 			Expect(err).NotTo(HaveOccurred())
 
 			service.SaveJobsToFile()
@@ -311,7 +311,7 @@ var _ = Describe("AgentJobService", func() {
 			Expect(err).NotTo(HaveOccurred())
 
 			params := map[string]string{}
-			jobID, err := service.ExecuteJob(taskID, params, "test")
+			jobID, err := service.ExecuteJob(taskID, params, "test", nil)
 			Expect(err).NotTo(HaveOccurred())
 
 			// Manually set job creation time to be old
@@ -329,4 +329,306 @@ var _ = Describe("AgentJobService", func() {
 			Expect(err).NotTo(HaveOccurred())
 		})
 	})
+
+	Describe("Multimedia support", func() {
+		Describe("Task multimedia sources", func() {
+			It("should create a task with multimedia sources", func() {
+				task := schema.Task{
+					Name:   "Multimedia Task",
+					Model:  "test-model",
+					Prompt: "Analyze this image",
+					MultimediaSources: []schema.MultimediaSourceConfig{
+						{
+							Type:    "image",
+							URL:     "https://example.com/image.png",
+							Headers: map[string]string{"Authorization": "Bearer token123"},
+						},
+						{
+							Type: "video",
+							URL:  "https://example.com/video.mp4",
+						},
+					},
+				}
+
+				id, err := service.CreateTask(task)
+				Expect(err).NotTo(HaveOccurred())
+				Expect(id).NotTo(BeEmpty())
+
+				retrieved, err := service.GetTask(id)
+				Expect(err).NotTo(HaveOccurred())
+				Expect(retrieved.MultimediaSources).To(HaveLen(2))
+				Expect(retrieved.MultimediaSources[0].Type).To(Equal("image"))
+				Expect(retrieved.MultimediaSources[0].URL).To(Equal("https://example.com/image.png"))
+				Expect(retrieved.MultimediaSources[0].Headers["Authorization"]).To(Equal("Bearer token123"))
+				Expect(retrieved.MultimediaSources[1].Type).To(Equal("video"))
+			})
+
+			It("should save and load tasks with multimedia sources from file", func() {
+				task := schema.Task{
+					Name:   "Persistent Multimedia Task",
+					Model:  "test-model",
+					Prompt: "Test prompt",
+					MultimediaSources: []schema.MultimediaSourceConfig{
+						{
+							Type: "audio",
+							URL:  "https://example.com/audio.mp3",
+							Headers: map[string]string{
+								"X-Custom-Header": "value",
+							},
+						},
+					},
+				}
+
+				id, err := service.CreateTask(task)
+				Expect(err).NotTo(HaveOccurred())
+
+				// Create a new service instance to test loading
+				newService := services.NewAgentJobService(
+					appConfig,
+					modelLoader,
+					configLoader,
+					evaluator,
+				)
+
+				err = newService.LoadTasksFromFile()
+				Expect(err).NotTo(HaveOccurred())
+
+				retrieved, err := newService.GetTask(id)
+				Expect(err).NotTo(HaveOccurred())
+				Expect(retrieved.Name).To(Equal("Persistent Multimedia Task"))
+				Expect(retrieved.MultimediaSources).To(HaveLen(1))
+				Expect(retrieved.MultimediaSources[0].Type).To(Equal("audio"))
+				Expect(retrieved.MultimediaSources[0].URL).To(Equal("https://example.com/audio.mp3"))
+				Expect(retrieved.MultimediaSources[0].Headers["X-Custom-Header"]).To(Equal("value"))
+			})
+		})
+
+		Describe("Job multimedia", func() {
+			var taskID string
+
+			BeforeEach(func() {
+				task := schema.Task{
+					Name:    "Test Task",
+					Model:   "test-model",
+					Prompt:  "Hello {{.name}}",
+					Enabled: true,
+				}
+				var err error
+				taskID, err = service.CreateTask(task)
+				Expect(err).NotTo(HaveOccurred())
+			})
+
+			It("should create a job with multimedia content", func() {
+				params := map[string]string{"name": "World"}
+				multimedia := &struct {
+					Images []string
+					Videos []string
+					Audios []string
+					Files  []string
+				}{
+					Images: []string{"https://example.com/image1.png", "data:image/png;base64,iVBORw0KG"},
+					Videos: []string{"https://example.com/video.mp4"},
+					Audios: []string{"data:audio/mpeg;base64,SUQzBAAAAA"},
+					Files:  []string{"https://example.com/file.pdf"},
+				}
+
+				jobID, err := service.ExecuteJob(taskID, params, "test", multimedia)
+				Expect(err).NotTo(HaveOccurred())
+				Expect(jobID).NotTo(BeEmpty())
+
+				job, err := service.GetJob(jobID)
+				Expect(err).NotTo(HaveOccurred())
+				Expect(job.TaskID).To(Equal(taskID))
+				Expect(job.Images).To(HaveLen(2))
+				Expect(job.Images[0]).To(Equal("https://example.com/image1.png"))
+				Expect(job.Images[1]).To(Equal("data:image/png;base64,iVBORw0KG"))
+				Expect(job.Videos).To(HaveLen(1))
+				Expect(job.Videos[0]).To(Equal("https://example.com/video.mp4"))
+				Expect(job.Audios).To(HaveLen(1))
+				Expect(job.Audios[0]).To(Equal("data:audio/mpeg;base64,SUQzBAAAAA"))
+				Expect(job.Files).To(HaveLen(1))
+				Expect(job.Files[0]).To(Equal("https://example.com/file.pdf"))
+			})
+
+			It("should create a job with partial multimedia (only images)", func() {
+				params := map[string]string{}
+				multimedia := &struct {
+					Images []string
+					Videos []string
+					Audios []string
+					Files  []string
+				}{
+					Images: []string{"https://example.com/image.png"},
+				}
+
+				jobID, err := service.ExecuteJob(taskID, params, "test", multimedia)
+				Expect(err).NotTo(HaveOccurred())
+
+				job, err := service.GetJob(jobID)
+				Expect(err).NotTo(HaveOccurred())
+				Expect(job.Images).To(HaveLen(1))
+				Expect(job.Videos).To(BeEmpty())
+				Expect(job.Audios).To(BeEmpty())
+				Expect(job.Files).To(BeEmpty())
+			})
+
+			It("should create a job without multimedia (nil)", func() {
+				params := map[string]string{"name": "Test"}
+				jobID, err := service.ExecuteJob(taskID, params, "test", nil)
+				Expect(err).NotTo(HaveOccurred())
+
+				job, err := service.GetJob(jobID)
+				Expect(err).NotTo(HaveOccurred())
+				Expect(job.Images).To(BeEmpty())
+				Expect(job.Videos).To(BeEmpty())
+				Expect(job.Audios).To(BeEmpty())
+				Expect(job.Files).To(BeEmpty())
+			})
+
+			It("should save and load jobs with multimedia from file", func() {
+				params := map[string]string{}
+				multimedia := &struct {
+					Images []string
+					Videos []string
+					Audios []string
+					Files  []string
+				}{
+					Images: []string{"https://example.com/image.png"},
+					Videos: []string{"https://example.com/video.mp4"},
+				}
+
+				jobID, err := service.ExecuteJob(taskID, params, "test", multimedia)
+				Expect(err).NotTo(HaveOccurred())
+
+				// Wait a bit for async save to complete
+				time.Sleep(50 * time.Millisecond)
+
+				// Ensure directory exists before saving
+				err = os.MkdirAll(tempDir, 0755)
+				Expect(err).NotTo(HaveOccurred())
+
+				err = service.SaveJobsToFile()
+				Expect(err).NotTo(HaveOccurred())
+
+				// Create a new service instance to test loading
+				newService := services.NewAgentJobService(
+					appConfig,
+					modelLoader,
+					configLoader,
+					evaluator,
+				)
+
+				err = newService.LoadJobsFromFile()
+				Expect(err).NotTo(HaveOccurred())
+
+				retrieved, err := newService.GetJob(jobID)
+				Expect(err).NotTo(HaveOccurred())
+				Expect(retrieved.TaskID).To(Equal(taskID))
+				Expect(retrieved.Images).To(HaveLen(1))
+				Expect(retrieved.Images[0]).To(Equal("https://example.com/image.png"))
+				Expect(retrieved.Videos).To(HaveLen(1))
+				Expect(retrieved.Videos[0]).To(Equal("https://example.com/video.mp4"))
+			})
+		})
+
+		Describe("Multimedia format handling", func() {
+			var taskID string
+
+			BeforeEach(func() {
+				task := schema.Task{
+					Name:    "Test Task",
+					Model:   "test-model",
+					Prompt:  "Test prompt",
+					Enabled: true,
+				}
+				var err error
+				taskID, err = service.CreateTask(task)
+				Expect(err).NotTo(HaveOccurred())
+			})
+
+			It("should handle URLs correctly", func() {
+				multimedia := &struct {
+					Images []string
+					Videos []string
+					Audios []string
+					Files  []string
+				}{
+					Images: []string{"https://example.com/image.png"},
+					Videos: []string{"http://example.com/video.mp4"},
+				}
+
+				jobID, err := service.ExecuteJob(taskID, map[string]string{}, "test", multimedia)
+				Expect(err).NotTo(HaveOccurred())
+
+				job, err := service.GetJob(jobID)
+				Expect(err).NotTo(HaveOccurred())
+				Expect(job.Images[0]).To(Equal("https://example.com/image.png"))
+				Expect(job.Videos[0]).To(Equal("http://example.com/video.mp4"))
+			})
+
+			It("should handle data URIs correctly", func() {
+				multimedia := &struct {
+					Images []string
+					Videos []string
+					Audios []string
+					Files  []string
+				}{
+					Images: []string{"data:image/png;base64,iVBORw0KG"},
+					Videos: []string{"data:video/mp4;base64,AAAAIGZ0eXBpc29t"},
+				}
+
+				jobID, err := service.ExecuteJob(taskID, map[string]string{}, "test", multimedia)
+				Expect(err).NotTo(HaveOccurred())
+
+				job, err := service.GetJob(jobID)
+				Expect(err).NotTo(HaveOccurred())
+				Expect(job.Images[0]).To(Equal("data:image/png;base64,iVBORw0KG"))
+				Expect(job.Videos[0]).To(Equal("data:video/mp4;base64,AAAAIGZ0eXBpc29t"))
+			})
+
+			It("should handle base64 strings (will be converted during execution)", func() {
+				// Base64 strings without data URI prefix should be stored as-is
+				// They will be converted to data URIs during execution
+				multimedia := &struct {
+					Images []string
+					Videos []string
+					Audios []string
+					Files  []string
+				}{
+					Images: []string{"iVBORw0KGgoAAAANSUhEUgAAAAEAAAABCAYAAAAfFcSJAAAADUlEQVR42mNk+M9QDwADhgGAWjR9awAAAABJRU5ErkJggg=="},
+				}
+
+				jobID, err := service.ExecuteJob(taskID, map[string]string{}, "test", multimedia)
+				Expect(err).NotTo(HaveOccurred())
+
+				job, err := service.GetJob(jobID)
+				Expect(err).NotTo(HaveOccurred())
+				// The base64 string is stored as-is in the job
+				Expect(job.Images[0]).To(Equal("iVBORw0KGgoAAAANSUhEUgAAAAEAAAABCAYAAAAfFcSJAAAADUlEQVR42mNk+M9QDwADhgGAWjR9awAAAABJRU5ErkJggg=="))
+			})
+
+			It("should handle empty multimedia arrays", func() {
+				multimedia := &struct {
+					Images []string
+					Videos []string
+					Audios []string
+					Files  []string
+				}{
+					Images: []string{""},
+					Videos: []string{},
+				}
+
+				jobID, err := service.ExecuteJob(taskID, map[string]string{}, "test", multimedia)
+				Expect(err).NotTo(HaveOccurred())
+
+				job, err := service.GetJob(jobID)
+				Expect(err).NotTo(HaveOccurred())
+				// Empty strings are stored in the job but will be filtered during execution
+				// The job stores what was provided, filtering happens in convertToMultimediaContent
+				Expect(job.Images).To(HaveLen(1))
+				Expect(job.Images[0]).To(Equal(""))
+				Expect(job.Videos).To(BeEmpty())
+			})
+		})
+	})
 })

From 313b20cdc07d7fb5b4d10e647a38f3880e0661d4 Mon Sep 17 00:00:00 2001
From: Ettore Di Giacinto <mudler@localai.io>
Date: Sun, 30 Nov 2025 12:13:30 +0100
Subject: [PATCH 2/2] Refactoring

Signed-off-by: Ettore Di Giacinto <mudler@localai.io>
---
 core/http/endpoints/localai/agent_jobs.go | 17 ++------
 core/schema/agent_jobs.go                 |  9 +++-
 core/services/agent_jobs.go               | 38 +++++++++--------
 core/services/agent_jobs_test.go          | 50 ++++-------------------
 4 files changed, 38 insertions(+), 76 deletions(-)

diff --git a/core/http/endpoints/localai/agent_jobs.go b/core/http/endpoints/localai/agent_jobs.go
index 267547434950..c46a0208a10f 100644
--- a/core/http/endpoints/localai/agent_jobs.go
+++ b/core/http/endpoints/localai/agent_jobs.go
@@ -148,19 +148,9 @@ func ExecuteJobEndpoint(app *application.Application) echo.HandlerFunc {
 		}
 
 		// Build multimedia struct from request
-		var multimedia *struct {
-			Images []string
-			Videos []string
-			Audios []string
-			Files  []string
-		}
+		var multimedia *schema.MultimediaAttachment
 		if len(req.Images) > 0 || len(req.Videos) > 0 || len(req.Audios) > 0 || len(req.Files) > 0 {
-			multimedia = &struct {
-				Images []string
-				Videos []string
-				Audios []string
-				Files  []string
-			}{
+			multimedia = &schema.MultimediaAttachment{
 				Images: req.Images,
 				Videos: req.Videos,
 				Audios: req.Audios,
@@ -302,7 +292,7 @@ func ExecuteTaskByNameEndpoint(app *application.Application) echo.HandlerFunc {
 	return func(c echo.Context) error {
 		name := c.Param("name")
 		var params map[string]string
-		
+
 		// Try to bind parameters from request body
 		// If body is empty or invalid, use empty params
 		if c.Request().ContentLength > 0 {
@@ -357,4 +347,3 @@ func ExecuteTaskByNameEndpoint(app *application.Application) echo.HandlerFunc {
 		})
 	}
 }
-
diff --git a/core/schema/agent_jobs.go b/core/schema/agent_jobs.go
index b0453a293be9..ac5d13fc6c9f 100644
--- a/core/schema/agent_jobs.go
+++ b/core/schema/agent_jobs.go
@@ -52,10 +52,17 @@ type WebhookConfig struct {
 // Used in cron jobs to periodically fetch multimedia from URLs with custom headers
 type MultimediaSourceConfig struct {
 	Type    string            `json:"type"`              // "image", "video", "audio", "file"
-	URL     string            `json:"url"`              // URL to fetch from
+	URL     string            `json:"url"`               // URL to fetch from
 	Headers map[string]string `json:"headers,omitempty"` // Custom headers for HTTP request (e.g., Authorization)
 }
 
+type MultimediaAttachment struct {
+	Images []string `json:"images,omitempty"`
+	Videos []string `json:"videos,omitempty"`
+	Audios []string `json:"audios,omitempty"`
+	Files  []string `json:"files,omitempty"`
+}
+
 // JobStatus represents the status of a job
 type JobStatus string
 
diff --git a/core/services/agent_jobs.go b/core/services/agent_jobs.go
index 553ef1c61a62..0693400d5956 100644
--- a/core/services/agent_jobs.go
+++ b/core/services/agent_jobs.go
@@ -73,6 +73,13 @@ type JobExecution struct {
 	Cancel context.CancelFunc
 }
 
+const (
+	JobImageType = "image"
+	JobVideoType = "video"
+	JobAudioType = "audio"
+	JobFileType  = "file"
+)
+
 // NewAgentJobService creates a new AgentJobService instance
 func NewAgentJobService(
 	appConfig *config.ApplicationConfig,
@@ -368,12 +375,7 @@ func (s *AgentJobService) buildPrompt(templateStr string, params map[string]stri
 
 // ExecuteJob creates and queues a job for execution
 // multimedia can be nil for backward compatibility
-func (s *AgentJobService) ExecuteJob(taskID string, params map[string]string, triggeredBy string, multimedia *struct {
-	Images []string
-	Videos []string
-	Audios []string
-	Files  []string
-}) (string, error) {
+func (s *AgentJobService) ExecuteJob(taskID string, params map[string]string, triggeredBy string, multimedia *schema.MultimediaAttachment) (string, error) {
 	task := s.tasks.Get(taskID)
 	if task.ID == "" {
 		return "", fmt.Errorf("task not found: %s", taskID)
@@ -413,13 +415,13 @@ func (s *AgentJobService) ExecuteJob(taskID string, params map[string]string, tr
 
 			// Add to appropriate slice based on type
 			switch source.Type {
-			case "image":
+			case JobImageType:
 				job.Images = append(job.Images, dataURI)
-			case "video":
+			case JobVideoType:
 				job.Videos = append(job.Videos, dataURI)
-			case "audio":
+			case JobAudioType:
 				job.Audios = append(job.Audios, dataURI)
-			case "file":
+			case JobFileType:
 				job.Files = append(job.Files, dataURI)
 			}
 		}
@@ -622,13 +624,13 @@ func (s *AgentJobService) fetchMultimediaFromURL(url string, headers map[string]
 // getMimeTypeForMediaType returns the default MIME type for a media type
 func (s *AgentJobService) getMimeTypeForMediaType(mediaType string) string {
 	switch mediaType {
-	case "image":
+	case JobImageType:
 		return "image/png"
-	case "video":
+	case JobVideoType:
 		return "video/mp4"
-	case "audio":
+	case JobAudioType:
 		return "audio/mpeg"
-	case "file":
+	case JobFileType:
 		return "application/octet-stream"
 	default:
 		return "application/octet-stream"
@@ -746,7 +748,7 @@ func (s *AgentJobService) executeJobInternal(job schema.Job, task schema.Task, c
 
 	// Convert images
 	if len(job.Images) > 0 {
-		images, err := s.convertToMultimediaContent(job.Images, "image")
+		images, err := s.convertToMultimediaContent(job.Images, JobImageType)
 		if err != nil {
 			log.Warn().Err(err).Str("job_id", job.ID).Msg("Failed to convert images")
 		} else {
@@ -756,7 +758,7 @@ func (s *AgentJobService) executeJobInternal(job schema.Job, task schema.Task, c
 
 	// Convert videos
 	if len(job.Videos) > 0 {
-		videos, err := s.convertToMultimediaContent(job.Videos, "video")
+		videos, err := s.convertToMultimediaContent(job.Videos, JobVideoType)
 		if err != nil {
 			log.Warn().Err(err).Str("job_id", job.ID).Msg("Failed to convert videos")
 		} else {
@@ -766,7 +768,7 @@ func (s *AgentJobService) executeJobInternal(job schema.Job, task schema.Task, c
 
 	// Convert audios
 	if len(job.Audios) > 0 {
-		audios, err := s.convertToMultimediaContent(job.Audios, "audio")
+		audios, err := s.convertToMultimediaContent(job.Audios, JobAudioType)
 		if err != nil {
 			log.Warn().Err(err).Str("job_id", job.ID).Msg("Failed to convert audios")
 		} else {
@@ -776,7 +778,7 @@ func (s *AgentJobService) executeJobInternal(job schema.Job, task schema.Task, c
 
 	// Convert files
 	if len(job.Files) > 0 {
-		files, err := s.convertToMultimediaContent(job.Files, "file")
+		files, err := s.convertToMultimediaContent(job.Files, JobFileType)
 		if err != nil {
 			log.Warn().Err(err).Str("job_id", job.ID).Msg("Failed to convert files")
 		} else {
diff --git a/core/services/agent_jobs_test.go b/core/services/agent_jobs_test.go
index 1b902049d5c8..447f3e6b4d56 100644
--- a/core/services/agent_jobs_test.go
+++ b/core/services/agent_jobs_test.go
@@ -420,12 +420,7 @@ var _ = Describe("AgentJobService", func() {
 
 			It("should create a job with multimedia content", func() {
 				params := map[string]string{"name": "World"}
-				multimedia := &struct {
-					Images []string
-					Videos []string
-					Audios []string
-					Files  []string
-				}{
+				multimedia := &schema.MultimediaAttachment{
 					Images: []string{"https://example.com/image1.png", "data:image/png;base64,iVBORw0KG"},
 					Videos: []string{"https://example.com/video.mp4"},
 					Audios: []string{"data:audio/mpeg;base64,SUQzBAAAAA"},
@@ -452,12 +447,7 @@ var _ = Describe("AgentJobService", func() {
 
 			It("should create a job with partial multimedia (only images)", func() {
 				params := map[string]string{}
-				multimedia := &struct {
-					Images []string
-					Videos []string
-					Audios []string
-					Files  []string
-				}{
+				multimedia := &schema.MultimediaAttachment{
 					Images: []string{"https://example.com/image.png"},
 				}
 
@@ -487,12 +477,7 @@ var _ = Describe("AgentJobService", func() {
 
 			It("should save and load jobs with multimedia from file", func() {
 				params := map[string]string{}
-				multimedia := &struct {
-					Images []string
-					Videos []string
-					Audios []string
-					Files  []string
-				}{
+				multimedia := &schema.MultimediaAttachment{
 					Images: []string{"https://example.com/image.png"},
 					Videos: []string{"https://example.com/video.mp4"},
 				}
@@ -547,12 +532,7 @@ var _ = Describe("AgentJobService", func() {
 			})
 
 			It("should handle URLs correctly", func() {
-				multimedia := &struct {
-					Images []string
-					Videos []string
-					Audios []string
-					Files  []string
-				}{
+				multimedia := &schema.MultimediaAttachment{
 					Images: []string{"https://example.com/image.png"},
 					Videos: []string{"http://example.com/video.mp4"},
 				}
@@ -567,12 +547,7 @@ var _ = Describe("AgentJobService", func() {
 			})
 
 			It("should handle data URIs correctly", func() {
-				multimedia := &struct {
-					Images []string
-					Videos []string
-					Audios []string
-					Files  []string
-				}{
+				multimedia := &schema.MultimediaAttachment{
 					Images: []string{"data:image/png;base64,iVBORw0KG"},
 					Videos: []string{"data:video/mp4;base64,AAAAIGZ0eXBpc29t"},
 				}
@@ -589,12 +564,7 @@ var _ = Describe("AgentJobService", func() {
 			It("should handle base64 strings (will be converted during execution)", func() {
 				// Base64 strings without data URI prefix should be stored as-is
 				// They will be converted to data URIs during execution
-				multimedia := &struct {
-					Images []string
-					Videos []string
-					Audios []string
-					Files  []string
-				}{
+				multimedia := &schema.MultimediaAttachment{
 					Images: []string{"iVBORw0KGgoAAAANSUhEUgAAAAEAAAABCAYAAAAfFcSJAAAADUlEQVR42mNk+M9QDwADhgGAWjR9awAAAABJRU5ErkJggg=="},
 				}
 
@@ -608,14 +578,8 @@ var _ = Describe("AgentJobService", func() {
 			})
 
 			It("should handle empty multimedia arrays", func() {
-				multimedia := &struct {
-					Images []string
-					Videos []string
-					Audios []string
-					Files  []string
-				}{
+				multimedia := &schema.MultimediaAttachment{
 					Images: []string{""},
-					Videos: []string{},
 				}
 
 				jobID, err := service.ExecuteJob(taskID, map[string]string{}, "test", multimedia)