- Dashboard: track saved ideas in React state (Set), compare against
suggestion texts to show "✓ Salvata" persistently
- Backend: append 'Z' to saved_at ISO timestamp so JS parses as UTC
(fixes "2h fa" bug for UTC+2 users)
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Backend:
- Suggestions cached in DB per user, regenerated only after 24h
- ?force=true parameter to regenerate on demand
- New endpoints: GET/POST/DELETE /content/ideas for saved ideas
- POST /content/ideas/{id}/mark-used to track usage
Frontend:
- Dashboard: suggestions loaded from cache, not regenerated on every visit
- Dashboard: "Salva idea" button on each suggestion card
- Dashboard: "Dammi altri suggerimenti" CTA to force regeneration
- Dashboard: removed "Piani Attivi" stat card
- SavedIdeas page: list saved ideas, add new, delete, generate from idea
- Sidebar: added "Idee" nav item after "Contenuti"
- App.jsx: added /ideas route
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Mobile UX:
- index.css: comprehensive mobile media queries — headings scale down,
touch targets enforced, grid-2col-mobile collapse class, tablet breakpoint
- ContentArchive/ContentPage: grid minmax uses min(100%, Npx) to prevent
overflow on small screens
- CharacterForm: visual style + rules editor grids collapse on mobile
- Dashboard: stat cards grid mobile-safe
- Layout: better nav touch targets, footer responsive gap
Phase C — One-Click Generation:
- Backend: GET /api/content/suggestions endpoint — LLM generates 3 topic
ideas based on character profile and avoids repeating recent posts
- Dashboard: "Suggerimenti per oggi" section loads suggestions on mount,
each card links to /content with prefilled topic + character
- ContentPage: reads ?topic= and ?character= URL params, auto-fills form
and auto-triggers generation (one-click flow from Dashboard)
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
- Approve action saves post as reference example in character's content_rules
- Keep last 5 approved examples per character (auto-rotating)
- Inject last 3 approved examples as few-shot in LLM system prompt
- Lock YouTube/TikTok hashtag profile tabs for Freemium users (Pro only)
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Backend:
- Character model: add brand_voice, target_audience, business_goals,
products_services, content_rules (JSON do/dont), hashtag_profiles (JSON)
- Content generation: inject full character context into LLM system prompt
(voice, audience, goals, products, rules)
- Hashtag generation: merge always-on tags from profile with AI-generated tags
- Schema: update CharacterBase and CharacterUpdate with new fields
Frontend:
- CharacterForm: new sections "Identità e Voce", "Regole Contenuti",
"Profili Hashtag" with dedicated editors
- RulesEditor: do/don't list with add/remove
- HashtagProfileEditor: per-platform tabs, fixed hashtags + max generated count
- All fields loaded on edit, saved on submit
DB migration: 6 new columns added to characters table
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Backend:
- Add batch_id column to Post model (UUID, groups posts from same generation)
- Set batch_id in /generate endpoint for all posts in same request
Frontend:
- ContentArchive: group posts by batch_id into single cards with platform tabs
- Character name at top, platform tabs below, status badge, text preview
- Click platform tab to switch between variants of same content
- ConfirmModal: render via React portal to document.body for true fullscreen overlay
- Add box-shadow and higher z-index for better visual separation
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Backend:
- /generate now returns array of posts (one per platform selected)
- Each post generated with platform-specific LLM prompt and char limits
- Monthly counter incremented by number of platforms
Frontend:
- ConfirmModal: reusable Editorial Fresh modal replaces ugly browser confirm()
- ContentPage: platform tabs when multiple posts, switch between variants
- ContentPage: generatedPosts array state replaces single generated
- ContentArchive: uses ConfirmModal for delete confirmation
- Platform chips filtered by plan (Freemium: IG/FB only)
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
- Add 'brief' field to GenerateContentRequest schema
- Pass brief from router to generate_post_text service
- Inject brief as mandatory instructions in LLM prompt with highest priority
- Return structured error when LLM provider/API key not configured
- Show dedicated warning banner with link to Settings when API key missing
Fixes: content ignoring editorial brief, unhelpful API key error messages
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
- Redesign Settings: Testi, Immagini, Video, Voiceover — sezioni separate
- Ogni sezione ha dropdown provider + API key + campo opzionale modello
- Opzione "Personalizzato" con campo Base URL libero per qualsiasi servizio
- LLM: aggiunto OpenRouter + provider custom OpenAI-compatible
- Backend: OpenAICompatibleProvider unifica OpenAI/OpenRouter/custom
- Router content: passa llm_base_url a get_llm_provider
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
