#432 App: AI Content Generator (content_generator)

## Overview Generate marketing content, blog posts, and social media content using LLM with human review workflow. ## App Metadata - **App Name:** content_generator - **Publisher:** platform - **Version:** 1.0.0 - **Priority:** High - **Complexity:** Medium - **Phase:** 2 (Core business value) ## Actions ### 1. generate_blog_post **Description:** Generate a blog post from topic and guidelines **Parameters:** - topic (str, required): Blog post topic - tone (str, default='professional'): Writing tone (professional|casual|technical|friendly) - length (str, default='medium'): Post length (short:300w|medium:800w|long:1500w) - target_audience (str, optional): Audience description - keywords (list, optional): SEO keywords to include - outline (list, optional): Specific sections to include - include_meta (bool, default=True): Generate meta description and title **Returns:** { "title": "...", "content": "...", "meta_description": "...", "word_count": 850, "keywords_used": ["k1", "k2"], "sections": ["intro", "section1", ...], "model_used": "qwen3-vl" } ### 2. generate_email_campaign **Description:** Generate marketing email content **Parameters:** - campaign_type (str, required): Type (newsletter|promotional|announcement|welcome|followup) - product_or_topic (str, required): What the email is about - tone (str, default='friendly'): Email tone - call_to_action (str, required): Desired CTA - recipient_segment (str, optional): Target audience segment - personalization_fields (list, optional): Fields like {{name}}, {{company}} **Returns:** { "subject_lines": ["Option 1", "Option 2", "Option 3"], "preview_text": "...", "body_html": "...", "body_text": "...", "cta_button_text": "..." } ### 3. generate_social_posts **Description:** Generate social media content for multiple platforms **Parameters:** - topic (str, required): Content topic - platforms (list, default=['twitter','linkedin']): Target platforms - tone (str, default='engaging'): Post tone - include_hashtags (bool, default=True): Add relevant hashtags - variations (int, default=3): Number of variations per platform - link (str, optional): Link to include **Returns:** { "twitter": [{"text": "...", "hashtags": [...], "char_count": 240}, ...], "linkedin": [{"text": "...", "hashtags": [...]}], "instagram": [{"caption": "...", "hashtags": [...]}] } ### 4. review_and_send **Description:** Send generated content for human review via email **Parameters:** - content (dict, required): Generated content to review - reviewer_email (str, required): Reviewer email - content_type (str, required): Type of content (blog|email|social) - approval_required (bool, default=True): Require explicit approval - edit_link (str, optional): Link to edit the content **Returns:** {"sent": true, "review_id": "rev_123"} ## Example Workflow - Weekly Blog Post builder = WorkflowBuilder(name='weekly_blog', version='1.0.0') builder.task( task_id='generate_post', function='apps.platform.content_generator.generate_blog_post', kwargs={ 'topic': '{{topic}}', 'tone': 'professional', 'length': 'medium', 'keywords': ['{{keyword1}}', '{{keyword2}}'], 'target_audience': 'Small business owners' }, result_key='blog_post' ) builder.task( task_id='send_for_review', function='apps.platform.content_generator.review_and_send', kwargs={ 'content': '{{blog_post}}', 'reviewer_email': 'marketing@example.com', 'content_type': 'blog' } ) ## LLM Prompt Templates Store prompts in configuration for easy tuning: - blog_post_prompt: System prompt for blog generation - email_campaign_prompt: System prompt for email content - social_post_prompt: System prompt for social media ## Implementation Notes - Use existing LLM tool (engine/tools/llm.py) - Temperature settings: 0.7 for creative, 0.3 for factual - Implement content length estimation and enforcement - Add plagiarism/quality checks in future version - Support multiple LLM providers (Ollama, OpenAI, Anthropic) ## Configuration Schema { "llm_model": "qwen3-vl:235b-instruct-cloud", "default_temperature": 0.7, "max_retries": 2, "content_guidelines": "Company style guide text...", "prohibited_topics": [] } ## Secrets Config None (LLM config from environment) ## Dependencies - engine/tools/llm.py - Email sending (SMTP) - Optional: approval_chain app for review workflow