AI Social Media Content and Management

def generate_social_post(topic, platform, brand_voice, audience, model="glm"): """Generate platform-optimized social media content.""" platform_specs = { "twitter": {"max_chars": 280, "style": "concise, punchy, thread-friendly", "hashtags": "2-3 relevant"}, "linkedin": {"max_chars": 3000, "style": "professional, thought-leadership", "hashtags": "3-5 relevant"}, "instagram": {"max_chars": 2200, "style": "visual-first, storytelling, emoji-friendly", "hashtags": "15-20 relevant"}, "facebook": {"max_chars": 63206, "style": "conversational, community-focused", "hashtags": "2-3 relevant"}, "tiktok": {"max_chars": 2200, "style": "trendy, hook-driven, youth-friendly", "hashtags": "3-5 trending"}, "threads": {"max_chars": 500, "style": "casual, conversational, text-focused", "hashtags": "minimal"} } spec = platform_specs.get(platform, platform_specs["twitter"]) prompt = f"""Create a {platform} post about: {topic} Brand Voice: {brand_voice} Target Audience: {audience} Platform Requirements: - Max length: {spec['max_chars']} characters - Style: {spec['style']} - Hashtags: {spec['hashtags']} Requirements: - Hook attention in first 1-2 lines - Include clear call-to-action - Optimize for {platform} algorithm (engagement-driving) - Match brand voice consistently - Include relevant hashtags - Suggest visual/video concept - If platform supports threads/carousels, suggest structure Also generate: 1. The main post text 2. 2-3 alternative versions (A/B test options) 3. Recommended posting time 4. Visual suggestion 5. Engagement prediction (high/medium/low) with reasoning Output as JSON.""" return call_llm_api(prompt, temperature= 0.6, max_tokens=1200, response_format="json")

def repurpose_content(source_content, source_format, target_platforms, model="deepseek"): """Repurpose a single content piece for multiple platforms.""" prompt = f"""Repurpose the following {source_format} for multiple social media platforms. Original Content: {source_content[:5000]} Target Platforms: {', '.join(target_platforms)} For each platform, provide: 1. PLATFORM-SPECIFIC VERSION: - Formatted for the platform's conventions - Optimized length - Appropriate tone adjustments - Hashtag strategy 2. CONTENT ANGLE: - How to frame the same content differently for this audience - What to emphasize vs. de-emphasize - Hook variation 3. VISUAL/FORMAT SUGGESTION: - Image carousel description - Video script outline (if applicable) - Infographic key points 4. POSTING STRATEGY: - Best time to post - Whether to post simultaneously or stagger - Cross-promotion approach Platforms to cover: {', '.join(target_platforms)} Output as JSON with one entry per platform.""" return call_llm_api(prompt, temperature=0.5, max_tokens=2000, response_format="json")

def generate_comment_reply(comment_text, post_context, brand_voice, sentiment, model="qwen"): """Generate appropriate reply to social media comment.""" prompt = f"""Generate a reply to the following social media comment. Original Post Context: {post_context} Comment to Reply To: {comment_text} Detected Sentiment: {sentiment} Brand Voice: {brand_voice} Guidelines: - Match brand voice consistently - Address the specific point in the comment - Be authentic and human-sounding (not robotic) - For complaints: acknowledge, apologize, offer resolution - For praise: thank genuinely, add value - For questions: answer accurately, provide next steps - For trolls/spam: flag for moderation, do not engage - Keep it concise (platform-appropriate length) - Use emojis if brand voice permits (sparingly) - Never be defensive or argumentative - Include CTA when natural (visit link, DM us, etc.) Generate 2-3 reply options with different approaches. Also flag if this comment requires human escalation (sensitive legal/PR issue).""" return call_llm_api(prompt, temperature=0.5, max_tokens=600) def triage_social_mentions(mentions_batch, brand_keywords, model="deepseek"): """Triage and categorize social media mentions.""" prompt = f"""Triage the following social media mentions for response priority. Brand Keywords: {', '.join(brand_keywords)} Mentions: {chr(10).join([f"{i+1}. [{m['platform']}] @{m['author']}: {m['text']}" for i, m in enumerate(mentions_batch)])} For each mention: 1. PRIORITY: URGENT | HIGH | MEDIUM | LOW | IGNORE - URGENT: Complaints from influencers, viral negative, legal issues, safety concerns - HIGH: Customer complaints, product issues, partnership inquiries - MEDIUM: General questions, feature requests, neutral mentions - LOW: Casual mentions, off-topic, positive but no response needed - IGNORE: Spam, bots, irrelevant 2. CATEGORY: COMPLAINT | QUESTION | PRAISE | INQUIRY | COMPETITOR_MENTION | SPAM | OTHER 3. SENTIMENT: POSITIVE | NEUTRAL | NEGATIVE | MIXED 4. RECOMMENDED_ACTION: RESPOND_NOW | RESPOND_SOON | MONITOR | IGNORE | ESCALATE 5. SUGGESTED_REPLY_APPROACH: Brief description of how to handle 6. INFLUENCER_FLAG: Is this from an account with high follower count or industry influence? Output as JSON array.""" return call_llm_api(prompt, temperature=0.2, max_tokens=1500, response_format="json")

def analyze_social_performance(metrics_data, content_calendar, model="deepseek"): """Analyze social media performance and generate recommendations.""" prompt = f"""Analyze the following social media performance data and provide strategic recommendations. Performance Metrics (last 30 days): {chr(10).join([f"- {m['platform']}: Posts: {m['posts']}, Reach: {m['reach']:,}, Engagement: {m['engagement_rate']}%, Clicks: {m['clicks']:,}, Shares: {m['shares']:,}, Comments: {m['comments']:,}" for m in metrics_data])} Top Performing Content: {chr(10).join([f"- {p['platform']}: {p['topic']} | Engagement: {p['engagement']} | Format: {p['format']}" for p in content_calendar['top_performers']])} Low Performing Content: {chr(10).join([f"- {p['platform']}: {p['topic']} | Engagement: {p['engagement']} | Format: {p['format']}" for p in content_calendar['low_performers']])} Provide: 1. PERFORMANCE SUMMARY: - Overall health assessment - Platform-by-platform comparison - Trend vs. previous period 2. CONTENT INSIGHTS: - What content types perform best - Optimal posting times by platform - Best-performing topics/themes - Format effectiveness (video, image, carousel, text) 3. AUDIENCE INSIGHTS: - Engagement patterns - Growth opportunities - Audience sentiment trends 4. COMPETITIVE POSITION (if data available): - Benchmark vs. industry averages - Share of voice assessment 5. STRATEGIC RECOMMENDATIONS: - Top 5 actions for next month - Content calendar adjustments - Platform investment priorities - Budget allocation suggestions 6. RISK ALERTS: - Any declining metrics requiring attention - Negative sentiment trends - Competitive threats Output as structured report.""" return call_llm_api(prompt, temperature=0.3, max_tokens=2500)

def generate_content_calendar(theme, platforms, frequency, duration_weeks, model="glm"): """Generate strategic content calendar.""" prompt = f"""Create a strategic social media content calendar. Campaign Theme: {theme} Platforms: {', '.join(platforms)} Posting Frequency: {frequency} Duration: {duration_weeks} weeks Generate a week-by-week calendar with: 1. CONTENT PILLARS (3-5 recurring themes) 2. WEEKLY BREAKDOWN: For each day with a scheduled post: - Platform - Content topic/angle - Format (single image, carousel, video, text, poll, etc.) - Posting time (with timezone) - Caption theme - Hashtag strategy - CTA objective (awareness, engagement, traffic, conversion) 3. CONTENT MIX: - Educational: X% - Entertaining: X% - Promotional: X% - Community/UGC: X% - Behind-the-scenes: X% 4. CROSS-PLATFORM STRATEGY: - How content adapts per platform - Cross-promotion plan - Platform-specific campaigns 5. KEY DATES: - Industry events to reference - Trending topics to leverage - Company milestones to highlight Output as structured JSON (week -> day -> post details).""" return call_llm_api(prompt, temperature=0.5, max_tokens=3000, response_format="json")

def analyze_influencer_profile(profile_data, brand_fit_criteria, model="deepseek"): """Analyze influencer profile for brand partnership potential.""" prompt = f"""Analyze the following influencer profile for brand partnership fit. Influencer Profile: - Handle: @{profile_data['handle']} - Platform: {profile_data['platform']} - Followers: {profile_data['followers']:,} - Niche: {profile_data['niche']} - Engagement Rate: {profile_data['engagement_rate']}% - Content Style: {profile_data['content_style']} - Recent Content Themes: {', '.join(profile_data['recent_themes'])} - Audience Demographics: {profile_data.get('demographics', 'Unknown')} - Previous Brand Partnerships: {', '.join(profile_data.get('brand_partnerships', []))} Brand Fit Criteria: {chr(10).join([f"- {k}: {v}" for k, v in brand_fit_criteria.items()])} Provide: 1. PARTNERSHIP FIT SCORE: 0-100 2. AUDIENCE ALIGNMENT: - Demographic match - Interest overlap - Geographic relevance 3. AUTHENTICITY ASSESSMENT: - Genuine engagement vs. bot activity indicators - Content quality and consistency - Brand partnership history (over-commercialized?) 4. RISK FACTORS: - Controversial content history - Competitor associations - Engagement authenticity concerns 5. PARTNERSHIP RECOMMENDATION: - COLLABORATION_TYPE: Sponsored post, affiliate, ambassador, giveaway - Expected ROI range - Suggested campaign concept - Negotiation starting point 6. ALTERNATIVES: 2-3 similar influencers if this one is not ideal Output as JSON.""" return call_llm_api(prompt, temperature=0.3, max_tokens=1500, response_format="json")