Initial commit

2026-06-17 01:26:44 +02:00
commit 51c56a6f30
65 changed files with 3731 additions and 0 deletions
@@ -0,0 +1,8 @@
+GOOGLE_API_KEY=AIzaSyD6deyt0d3yp88sLbFaadJKxcOg-5cGy-A
+BUFFER_ACCESS_TOKEN=EygpqeH802evGWSXqJ9EDSL_OpVBAFOBZqoVzw6OS9u
+MAKE_WEBHOOK_URL=optional_make_webhook
+TELEGRAM_BOT_TOKEN=your_bot_token
+TELEGRAM_CHAT_ID=your_chat_id
+
+# Higgsfield Credentials
+HIGGSFIELD_API_KEY=792d35ce-5429-4056-a2b4-5b20bf74cf69
@@ -0,0 +1,5 @@
+.env
+.venv/
+__pycache__/
+*.pyc
+.DS_Store
@@ -0,0 +1,25 @@
+FROM python:3.11-slim
+
+# Install system dependencies
+RUN apt-get update && apt-get install -y \
+    ffmpeg \
+    libmagic1 \
+    && rm -rf /var/lib/apt/lists/*
+
+WORKDIR /app
+
+# Aggiungiamo src al PYTHONPATH così gli import funzionano da ovunque
+ENV PYTHONPATH=/app/src
+
+# Copy requirements and install
+COPY requirements.txt .
+RUN pip install --no-cache-dir -r requirements.txt
+
+# Copy the rest of the application
+COPY . .
+
+# Expose port for Streamlit
+EXPOSE 8501
+
+# Avvio di Streamlit puntando alla nuova cartella src
+CMD ["streamlit", "run", "src/app.py", "--server.port=8501", "--server.address=0.0.0.0"]
@@ -0,0 +1,43 @@
+# 📊 Guida alla Sincronizzazione Real-Time dei Crediti Higgsfield AI
+
+Questa guida spiega come abilitare il tracciamento automatico e in tempo reale del consumo crediti del tuo abbonamento Higgsfield AI (Pro Plan) all'interno della dashboard Streamlit. 
+
+Seguendo le tue indicazioni di sicurezza, **il sistema non effettua scansioni del tuo browser Chrome né accede al portachiavi di sistema (Chrome Safe Storage)**. La sincronizzazione si basa unicamente sul token di sessione configurato nel file `.env`.
+
+---
+
+## 🛠️ Cosa è stato fatto
+1. **Automazione Completa**: Rimosso il form manuale per regolare il budget dei crediti dal cruscotto Streamlit. Ora i crediti si allineano da soli online.
+2. **Aggiornamento REST Real-Time**: Implementato un client REST ufficiale (`HiggsClient`) in `src/app.py` che, ad ogni avvio o refresh del cruscotto, interroga in tempo reale i server di Higgsfield per mostrare la percentuale e i crediti rimanenti effettivi.
+3. **Indicatori Grafici Premium**:
+   * **🟢 SINCRONIZZATO ONLINE**: Indica che il token è valido e i crediti mostrati provengono in diretta dal tuo account, con data e ora dell'ultimo allineamento.
+   * **🟡 CONTATORE LOCALE (CACHE)**: Compare se il token non è impostato nel file `.env`, indicandoti come procedere.
+   * **🔴 ERRORE DI CONNESSIONE**: Compare in caso di token scaduto o problemi temporanei di connessione, mantenendo visibile l'ultima cache per stabilità.
+
+---
+
+## 🔑 Come configurare la sincronizzazione automatica
+
+Per sbloccare il monitoraggio automatico dei crediti reali, devi aggiungere il tuo token di sessione all'interno del file di configurazione ambientale. Segui questi passi semplici:
+
+1. Apri il browser ed effettua l'accesso sul sito ufficiale **[higgsfield.ai](https://higgsfield.ai)**.
+2. Premi `F12` (oppure clicca col tasto destro sulla pagina e seleziona **Ispeziona**) per aprire gli Strumenti per Sviluppatori del browser.
+3. Spostati nella scheda **Application** (su Google Chrome) o **Storage** (su Firefox/Safari).
+4. Nel menu laterale sinistro, espandi la voce **Cookies** e seleziona `https://higgsfield.ai`.
+5. Cerca nell'elenco la riga con il nome **`__session`**.
+6. Fai doppio clic sul valore del cookie per selezionarlo interamente e **copialo** (è una lunga stringa alfanumerica che inizia con `eyJ...`).
+7. Apri il file [**.env**](file:///Users/davidfrassi/SRC/agenti/agenzia/.env) situato nella cartella principale del progetto.
+8. Aggiungi la seguente variabile in fondo al file incollando il token copiato:
+   ```env
+   HIGGSFIELD_TOKEN=eyJ...[incolla_qui_il_tuo_token]
+   ```
+9. Salva il file `.env`.
+
+---
+
+## 🚀 Risultato Atteso
+
+Una volta salvata la variabile nel file `.env`, ricarica il cruscotto Streamlit. Il pannello in basso dedicato ai crediti:
+* Rileverà automaticamente il token.
+* Mostrerà il badge verde **🟢 SINCRONIZZATO ONLINE** con la data e l'ora dell'aggiornamento.
+* Mostrerà i tuoi reali crediti rimanenti online e la percentuale esatta consumata, in perfetto allineamento con la dashboard ufficiale di Higgsfield!
@@ -0,0 +1,37 @@
+# Roadmap Implementazione: Redazione Social Multi-Agente
+
+Questo documento elenca i passi necessari per costruire il sistema di automazione social.
+
+## Fase 1: Setup Ambiente e Dati (Dockerized)
+- [ ] Inizializzare un progetto con **Docker** e **docker-compose**.
+- [ ] Creare la struttura delle cartelle persistenti (volumi):
+    - `data/images/artista_n`: per le foto.
+    - `data/audio/artista_n`: per i brani .mp3 da analizzare.
+- [ ] Creare un file `config.json` con i profili dei 4 cantanti e i link Spotify/Distrokid.
+- [ ] Preparare il `Dockerfile` con le dipendenze (Python, FFmpeg per audio, ecc.).
+
+## Fase 2: Sviluppo del Database delle Immagini
+- [ ] Creare uno script che scansiona la cartella immagini.
+- [ ] Implementare una logica di "tracking" per segnare le immagini già usate (per evitare ripetizioni).
+- [ ] (Opzionale) Usare un modello Vision per pre-analizzare le immagini e salvarne una descrizione.
+
+## Fase 3: Sviluppo degli Agenti
+- [ ] **Agente Audio Analyst**: Utilizza modelli AI multimodali (es. Gemini) per ascoltare il brano .mp3 e dedurre genere, mood, bpm e temi principali.
+- [ ] **Agente Asset Manager**: Seleziona l'immagine non usata dal folder specifico dell'artista.
+- [ ] **Agente Copywriter**: Genera testi basandosi sull'output dell'Audio Analyst (non più su generi predefiniti).
+- [ ] **Agente Visual-Harmonizer**: Incrocia l'analisi audio con l'estetica dell'immagine scelta.
+- [ ] **Agente Hashtag**: Genera tag basati sul brano analizzato e sul visual.
+- [ ] **Agente Supervisore**: Verifica finale della coerenza tra audio, immagine e link.
+
+## Fase 4: Orchestrazione e Approvazione
+- [ ] Configurare il grafo di **LangGraph** per collegare gli agenti.
+- [ ] Creare una semplice interfaccia (es. **Streamlit**) per visualizzare la proposta del post e cliccare su "Approva" o "Rifiuta".
+- [ ] Integrare un sistema di notifiche (es. Telegram Bot) per avvisarti quando un post è pronto.
+
+## Fase 5: Distribuzione (Post reale)
+- [ ] Integrare le API dei social media o un aggregatore (es. **Buffer API** o un webhook verso **Make.com**).
+- [ ] Testare il flusso completo con un post di prova.
+
+## Fase 6: Manutenzione
+- [ ] Implementare un log dei post pubblicati.
+- [ ] Sistema di alert se le immagini in una cartella stanno per finire.
@@ -0,0 +1,85 @@
+{
+  "artists": [
+    {
+      "id": "Veronica Intorcia",
+      "name": "Veronica Intorcia",
+      "spotify_url": "https://open.spotify.com/intl-it/album/2LXeQfqTYA9SyYFz776pj5?si=k0caFWXaSHW63FLmrzifIg",
+      "distrokid_url": "https://distrokid.com/hyperfollow/veronika33/midnight-meridian-2",
+      "schedule_time": "09:00",
+      "drive_audio_url": "https://drive.google.com/file/d/1IfFSCPYuBBBCRaLerQYe30lrzcIjhqiI/view?usp=sharing",
+      "drive_images_url": "https://drive.google.com/drive/folders/1MlO5TkCGrnUNoNFzepM6NtnBnAjhyiHp?usp=drive_link",
+      "drive_videos_url": "https://drive.google.com/drive/folders/1oKRPFYZgusUtk0F9RWiC6a-Su8e4kZY4?usp=drive_link",
+      "buffer_token": "d0uh1tvuD4Y_VDJkVDlDDYZSGHOxVWDnJlLCvNoBiYO",
+      "narrative_style": "sile narrativo da cantnate pop figa stile dualipa",
+      "social_tag": "@veronika_officialpage",
+      "soul_id": "soul_veronica_intorcia",
+      "drive_starting_photos_url": "https://drive.google.com/drive/folders/1FAFyCaIInW5qGkv_pzviqLZKUbaSCIf-?usp=drive_link"
+    },
+    {
+      "id": "Cinzia Carreri",
+      "name": "Cinzia Carreri",
+      "spotify_url": "https://open.spotify.com/intl-it/track/0Z5xASrrUMx60kBXyqzTfe?si=f672e645783f4fd5",
+      "distrokid_url": "https://distrokid.com/hyperfollow/odette4/mirror-of-life-2",
+      "schedule_time": "10:30",
+      "drive_audio_url": "https://drive.google.com/file/d/1BVn0MlYzolR9HxwZGlsAC8YmIJGBbW8y/view?usp=drive_link",
+      "drive_images_url": "https://drive.google.com/drive/folders/1M7mD43vGKoBV_TsSjEDrN6GCsliQrvn1?usp=drive_link",
+      "drive_videos_url": "https://drive.google.com/drive/folders/1jB_BJH2QKh_XvrZz5qjTuDS65jQ_NJE6?usp=drive_link",
+      "buffer_token": "d0uh1tvuD4Y_VDJkVDlDDYZSGHOxVWDnJlLCvNoBiYO",
+      "narrative_style": "lo stile narrativo deve essere un po' sognatore da amanti del genere Enya e colonne sonore di film fantasy. Scrivi il post in inglese",
+      "social_tag": "@odette_officialpage",
+      "soul_id": "soul_cinzia_carreri"
+    },
+    {
+      "id": "Martina Zerjial",
+      "name": "Martina Zerjial",
+      "spotify_url": "https://open.spotify.com/intl-it/track/4jlDBu3SV24FUrH7YjTRzt?si=72142d6360e34370",
+      "distrokid_url": "https://distrokid.com/hyperfollow/elaya/saltwater-kisses",
+      "schedule_time": "14:00",
+      "drive_audio_url": "https://drive.google.com/file/d/13vZtqfVQ7G3Rjta1DvsYaAicEI1J__Ix/view?usp=drive_link",
+      "drive_images_url": "https://drive.google.com/drive/folders/1jkA5ApSNTJ68mEdHPmaVkBSa-CDJ0ihe?usp=drive_link",
+      "drive_videos_url": "https://drive.google.com/drive/folders/1Fk85iijYBAuznGMv_LZN018jvQOkH6M6?usp=drive_link",
+      "buffer_token": "d0uh1tvuD4Y_VDJkVDlDDYZSGHOxVWDnJlLCvNoBiYO",
+      "narrative_style": "stile narrativo da popstar che fa genere retrowave ma anche funky moderno, in stile the kolors",
+      "social_tag": "@elaya_officialpage",
+      "soul_id": "soul_martina_zerjial"
+    },
+    {
+      "id": "Ambra Manca",
+      "name": "Ambra Manca",
+      "spotify_url": "",
+      "distrokid_url": " ",
+      "schedule_time": "18:00",
+      "drive_audio_url": "https://drive.google.com/file/d/1ScgXJmnniTK8QeFZlB_qgCrywVaR5-_J/view?usp=drive_link",
+      "drive_images_url": "https://drive.google.com/drive/folders/1OXYnJNILkzVrlXNcMUbZBcKMrE9KT1ho?usp=drive_link",
+      "drive_videos_url": "",
+      "buffer_token": "d0uh1tvuD4Y_VDJkVDlDDYZSGHOxVWDnJlLCvNoBiYO",
+      "narrative_style": "stile narrativo figo da cantante funky americano, con contaminazione alla jamiroquai",
+      "social_tag": "@jaderaya_official",
+      "soul_id": "soul_ambra_manca"
+    },
+    {
+      "id": "StereoComics",
+      "name": "StereoComics",
+      "spotify_url": "",
+      "distrokid_url": "",
+      "schedule_time": "18:00",
+      "drive_audio_url": "https://drive.google.com/file/d/14RMoKsiupSQVj2MEzbXgNBzKZ3FPLSVQ/view?usp=sharing",
+      "drive_images_url": "https://drive.google.com/drive/folders/1aaReM9YqG8JhXMQKtZuFqWE9RbvwvsRp?usp=drive_link",
+      "drive_videos_url": "",
+      "buffer_token": "4OwhtCHLJ61Ld659qeQS1qd0MVFypTY3TtP1Wa9AKSA",
+      "narrative_style": "lo stile narrativo deve essere un po' canzonatorio, ironico, da nerd, che fa citazione degli aneddoti degli autori originari del brano, usa un linguaggio giovanile e accattivante",
+      "social_tag": "@spaziosigle",
+      "soul_id": "soul_stereocomics"
+    }
+  ],
+  "settings": {
+    "daily_post_count": 1,
+    "platforms": [
+      "instagram",
+      "facebook",
+      "tiktok",
+      "x",
+      "youtube"
+    ]
+  }
+}
@@ -0,0 +1,51 @@
+[2026-05-05 21:05:20] INVIO POST a Canale: 69f9e7e35c4c051afa116a9e
+[2026-05-05 21:05:22] RISPOSTA BUFFER: {"post": {"id": "69fa5b92508055e61464dc15"}}
+[2026-05-05 21:05:22] INVIO POST a Canale: 69f9ea855c4c051afa117baa
+[2026-05-05 21:05:22] ERRORE HTTP 400: {"errors":[{"message":"Variable \"$input\" got invalid value { type: \"post\" } at \"input.metadata.instagram\"; Field \"shouldShareToFeed\" of required type \"Boolean!\" was not provided.","locations":[{"line":2,"column":23}],"extensions":{"code":"BAD_USER_INPUT"}}]}
+
+[2026-05-05 21:07:05] INVIO POST a Canale: 69f9e7e35c4c051afa116a9e
+[2026-05-05 21:07:06] RISPOSTA BUFFER: {"post": {"id": "69fa5bfaf2feba8c0c48e58f"}}
+[2026-05-05 21:07:06] INVIO POST a Canale: 69f9ea855c4c051afa117baa
+[2026-05-05 21:07:07] RISPOSTA BUFFER: {"post": {"id": "69fa5bfbf2feba8c0c48e5a9"}}
+[2026-05-05 21:09:19] INVIO POST a Canale: 69f9e7e35c4c051afa116a9e
+[2026-05-05 21:09:20] RISPOSTA BUFFER: {"post": {"id": "69fa5c80f2feba8c0c48e835"}}
+[2026-05-05 21:09:20] INVIO POST a Canale: 69f9ea855c4c051afa117baa
+[2026-05-05 21:09:21] RISPOSTA BUFFER: {"post": {"id": "69fa5c81508055e61464e273"}}
+[2026-05-05 21:37:02] INVIO POST a Canale: 69f9e7e35c4c051afa116a9e
+[2026-05-05 21:37:04] RISPOSTA BUFFER: {"post": {"id": "69fa6300f2feba8c0c492100"}}
+[2026-05-05 21:37:04] INVIO POST a Canale: 69f9ea855c4c051afa117baa
+[2026-05-05 21:37:05] RISPOSTA BUFFER: {"post": {"id": "69fa6301f2feba8c0c492127"}}
+[2026-05-05 21:38:31] INVIO POST a Canale: 69f9e7e35c4c051afa116a9e
+[2026-05-05 21:38:33] RISPOSTA BUFFER: {"post": {"id": "69fa6359508055e614652001"}}
+[2026-05-05 21:38:33] INVIO POST a Canale: 69f9ea855c4c051afa117baa
+[2026-05-05 21:38:35] RISPOSTA BUFFER: {"post": {"id": "69fa635bf2feba8c0c4922ab"}}
+[2026-05-05 21:41:27] INVIO POST a Canale: 69f9e7e35c4c051afa116a9e
+[2026-05-05 21:41:27] ERRORE HTTP 400: {"errors":[{"message":"Variable \"$input\" got invalid value { channelId: \"69f9e7e35c4c051afa116a9e\", text: \"*Un volo elettrico verso il nulla ✨*\\n\\nLuci notturne, riflessi dorati e il bisogno viscerale di lasciarsi tutto alle spalle. \\\"Midnight Meridian\\\" è la mia fuga urbana, una catarsi tra synth brillanti e ritmi che non lasciano scampo. Se senti la necessità di perderti nel movimento per ritrovare te stessa, questa è la tua nuova colonna sonora. Lascia che le ombre prendano vita e abbandonati a questa danza liberatoria. Ascolta ora: https://distrokid.com/hyperfollow/veronika33/midnight-meridian-2\\n\\n#MidnightMeridian #Veronika33 #Synthwave #ElectronicMusic #NewMusic #UrbanVibes #NightLife #Catharsis #Synthpop #DanceMusic #MusicRelease #MidnightVibes #IndependentArtist #ElectroPop #NightDrive\", mode: \"shareNow\", assets: { images: [Array] }, metadata: { tiktok: [Object] } }; Field \"schedulingType\" of required type \"SchedulingType!\" was not provided.","locations":[{"line":2,"column":23}],"extensions":{"code":"BAD_USER_INPUT"}}]}
+
+[2026-05-05 21:41:27] INVIO POST a Canale: 69f9ea855c4c051afa117baa
+[2026-05-05 21:41:27] ERRORE HTTP 400: {"errors":[{"message":"Variable \"$input\" got invalid value { channelId: \"69f9ea855c4c051afa117baa\", text: \"*Un volo elettrico verso il nulla ✨*\\n\\nLuci notturne, riflessi dorati e il bisogno viscerale di lasciarsi tutto alle spalle. \\\"Midnight Meridian\\\" è la mia fuga urbana, una catarsi tra synth brillanti e ritmi che non lasciano scampo. Se senti la necessità di perderti nel movimento per ritrovare te stessa, questa è la tua nuova colonna sonora. Lascia che le ombre prendano vita e abbandonati a questa danza liberatoria. Ascolta ora: https://distrokid.com/hyperfollow/veronika33/midnight-meridian-2\\n\\n#MidnightMeridian #Veronika33 #Synthwave #ElectronicMusic #NewMusic #UrbanVibes #NightLife #Catharsis #Synthpop #DanceMusic #MusicRelease #MidnightVibes #IndependentArtist #ElectroPop #NightDrive\", mode: \"shareNow\", assets: { images: [Array] }, metadata: { instagram: [Object] } }; Field \"schedulingType\" of required type \"SchedulingType!\" was not provided.","locations":[{"line":2,"column":23}],"extensions":{"code":"BAD_USER_INPUT"}}]}
+
+[2026-05-05 21:42:30] INVIO POST a Canale: 69f9e7e35c4c051afa116a9e
+[2026-05-05 21:42:41] RISPOSTA BUFFER: {"post": {"id": "69fa6447f2feba8c0c492ab3"}}
+[2026-05-05 21:42:41] INVIO POST a Canale: 69f9ea855c4c051afa117baa
+[2026-05-05 21:42:52] RISPOSTA BUFFER: {"post": {"id": "69fa6452f2feba8c0c492ae3"}}
+[2026-05-06 00:27:00] INVIO POST a Canale: 69f9e7e35c4c051afa116a9e
+[2026-05-06 00:27:01] ERRORE HTTP 400: {"errors":[{"message":"Variable \"$input\" got invalid value { video: { url: \"https://h.uguu.se/iBZbfOTe.mp4\" } } at \"input.assets\"; Field \"video\" is not defined by type \"AssetsInput\". [Suggestion hidden]?","locations":[{"line":2,"column":23}],"extensions":{"code":"BAD_USER_INPUT"}}]}
+
+[2026-05-06 00:27:01] INVIO POST a Canale: 69f9ea855c4c051afa117baa
+[2026-05-06 00:27:01] ERRORE HTTP 400: {"errors":[{"message":"Variable \"$input\" got invalid value { video: { url: \"https://h.uguu.se/iBZbfOTe.mp4\" } } at \"input.assets\"; Field \"video\" is not defined by type \"AssetsInput\". [Suggestion hidden]?","locations":[{"line":2,"column":23}],"extensions":{"code":"BAD_USER_INPUT"}}]}
+
+[2026-05-06 00:28:05] INVIO POST a Canale: 69f9e7e35c4c051afa116a9e
+[2026-05-06 00:28:19] RISPOSTA BUFFER: {"post": {"id": "69fa8b19e6fd3862f225c85c"}}
+[2026-05-06 00:28:19] INVIO POST a Canale: 69f9ea855c4c051afa117baa
+[2026-05-06 00:28:32] RISPOSTA BUFFER: {"post": {"id": "69fa8b25e6fd3862f225c8e1"}}
+[2026-05-12 23:52:39] INVIO POST a Canale: 69f9e7e35c4c051afa116a9e
+[2026-05-12 23:52:39] RISPOSTA BUFFER: {"message": "Channel not found"}
+[2026-05-12 23:52:39] INVIO POST a Canale: 69f9ea855c4c051afa117baa
+[2026-05-12 23:52:40] RISPOSTA BUFFER: {"message": "Channel not found"}
+[2026-05-13 00:09:28] Nessun Profile ID fornito, avvio auto-discovery...
+[2026-05-13 00:09:29] ERRORE recupero profili: {"errors":[{"message":"Cannot query field \"profiles\" on type \"Query\".","locations":[{"line":3,"column":11}],"extensions":{"code":"GRAPHQL_VALIDATION_FAILED"}}]}
+
+[2026-05-13 00:09:48] Nessun Profile ID fornito, avvio auto-discovery...
+[2026-05-13 00:09:48] ERRORE recupero profili: {"errors":[{"message":"Cannot query field \"profiles\" on type \"Query\".","locations":[{"line":3,"column":11}],"extensions":{"code":"GRAPHQL_VALIDATION_FAILED"}}]}
+
@@ -0,0 +1,5 @@
+{
+  "total_monthly": 1000,
+  "remaining": 910,
+  "generated_this_month": 90
+}
@@ -0,0 +1,138 @@
+
+--- Tue May 12 16:02:34 2026 ---
+❌ Errore critico nel task generate (StereoComics): Failed to retrieve file url:
+
+	Cannot retrieve the public link of the file. You may need to change
+	the permission to 'Anyone with the link', or have had many accesses.
+	Check FAQ in https://github.com/wkentaro/gdown?tab=readme-ov-file#faq.
+
+You may still be able to access the file from the browser:
+
+	https://drive.google.com/uc?id=14RMoKsiupSQVj2MEzbXgNBzKZ3FPLSVQ
+
+but Gdown can't. Please check connections and permissions.
+Traceback (most recent call last):
+  File "/usr/local/lib/python3.11/site-packages/gdown/download.py", line 301, in download
+    url = get_url_from_gdrive_confirmation(res.text)
+          ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
+  File "/usr/local/lib/python3.11/site-packages/gdown/download.py", line 65, in get_url_from_gdrive_confirmation
+    raise FileURLRetrievalError(
+gdown.exceptions.FileURLRetrievalError: Cannot retrieve the public link of the file. You may need to change the permission to 'Anyone with the link', or have had many accesses. Check FAQ in https://github.com/wkentaro/gdown?tab=readme-ov-file#faq.
+
+During handling of the above exception, another exception occurred:
+
+Traceback (most recent call last):
+  File "/app/src/background_runner.py", line 71, in <module>
+    run_generate(args.artist_id, mode=args.mode)
+  File "/app/src/background_runner.py", line 43, in run_generate
+    results = agents.run_for_artist(artist, mode=mode)
+              ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
+  File "/app/src/agents.py", line 285, in run_for_artist
+    audio_state = self.analyze_audio_node(initial_state)
+                  ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
+  File "/app/src/agents.py", line 49, in analyze_audio_node
+    gdown.download(url, output=path, quiet=True)
+  File "/usr/local/lib/python3.11/site-packages/gdown/download.py", line 312, in download
+    raise FileURLRetrievalError(message)
+gdown.exceptions.FileURLRetrievalError: Failed to retrieve file url:
+
+	Cannot retrieve the public link of the file. You may need to change
+	the permission to 'Anyone with the link', or have had many accesses.
+	Check FAQ in https://github.com/wkentaro/gdown?tab=readme-ov-file#faq.
+
+You may still be able to access the file from the browser:
+
+	https://drive.google.com/uc?id=14RMoKsiupSQVj2MEzbXgNBzKZ3FPLSVQ
+
+but Gdown can't. Please check connections and permissions.
+
+
+--- Tue May 12 20:58:46 2026 ---
+❌ Errore critico nel task generate (Veronica Intorcia): (sqlite3.OperationalError) table drafts has no column named focus_points
+[SQL: INSERT INTO drafts (artist_id, title, caption, hashtags, image_path, image_paths, image_url, video_url, video_path, audio_analysis, focus_points, status, created_at) VALUES (?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?)]
+[parameters: ('Veronica Intorcia', "Perdersi nell'Emozione", 'Lasciati trasportare dalle vibrazioni ipnotiche di "Lost In The Emotion". Una fusione magistrale tra Deep House e Melodic House, dove voci soul incon ... (151 characters truncated) ... ofisticata o vuole accendere l\'energia della notte. Ascolta ora il nuovo brano qui: https://distrokid.com/hyperfollow/veronika33/midnight-meridian-2', '#DeepHouse #MelodicHouse #ElectronicMusic #LostInTheEmotion #HouseMusic #SoulfulHouse #NewMusicAlert #HypnoticBeats #MusicJourney #NightVibes #MelodicTechno #IntrospectiveVibes #EuphoricMusic #DanceMusic #NewRelease', '', None, None, '', None, "Immergiti nelle profondità ipnotiche di 'Lost In The Emotion', una magistrale fusione di Deep House e Melodic House. Con le sue voci soul, il ritmo i ... (166 characters truncated) ... are un mood sofisticato o accendere l'energia notturna, è un viaggio sonoro accattivante, progettato per risuonare profondamente con gli ascoltatori.", None, 'pending', '2026-05-12 20:58:46.021420')]
+(Background on this error at: https://sqlalche.me/e/20/e3q8)
+Traceback (most recent call last):
+  File "/usr/local/lib/python3.11/site-packages/sqlalchemy/engine/base.py", line 1967, in _exec_single_context
+    self.dialect.do_execute(
+  File "/usr/local/lib/python3.11/site-packages/sqlalchemy/engine/default.py", line 952, in do_execute
+    cursor.execute(statement, parameters)
+sqlite3.OperationalError: table drafts has no column named focus_points
+
+The above exception was the direct cause of the following exception:
+
+Traceback (most recent call last):
+  File "/app/src/background_runner.py", line 72, in <module>
+    run_generate(args.artist_id, mode=args.mode)
+  File "/app/src/background_runner.py", line 46, in run_generate
+    db.save_draft(
+  File "/app/src/database.py", line 131, in save_draft
+    session.commit()
+  File "/usr/local/lib/python3.11/site-packages/sqlalchemy/orm/session.py", line 2030, in commit
+    trans.commit(_to_root=True)
+  File "<string>", line 2, in commit
+  File "/usr/local/lib/python3.11/site-packages/sqlalchemy/orm/state_changes.py", line 137, in _go
+    ret_value = fn(self, *arg, **kw)
+                ^^^^^^^^^^^^^^^^^^^^
+  File "/usr/local/lib/python3.11/site-packages/sqlalchemy/orm/session.py", line 1311, in commit
+    self._prepare_impl()
+  File "<string>", line 2, in _prepare_impl
+  File "/usr/local/lib/python3.11/site-packages/sqlalchemy/orm/state_changes.py", line 137, in _go
+    ret_value = fn(self, *arg, **kw)
+                ^^^^^^^^^^^^^^^^^^^^
+  File "/usr/local/lib/python3.11/site-packages/sqlalchemy/orm/session.py", line 1286, in _prepare_impl
+    self.session.flush()
+  File "/usr/local/lib/python3.11/site-packages/sqlalchemy/orm/session.py", line 4331, in flush
+    self._flush(objects)
+  File "/usr/local/lib/python3.11/site-packages/sqlalchemy/orm/session.py", line 4466, in _flush
+    with util.safe_reraise():
+  File "/usr/local/lib/python3.11/site-packages/sqlalchemy/util/langhelpers.py", line 121, in __exit__
+    raise exc_value.with_traceback(exc_tb)
+  File "/usr/local/lib/python3.11/site-packages/sqlalchemy/orm/session.py", line 4427, in _flush
+    flush_context.execute()
+  File "/usr/local/lib/python3.11/site-packages/sqlalchemy/orm/unitofwork.py", line 466, in execute
+    rec.execute(self)
+  File "/usr/local/lib/python3.11/site-packages/sqlalchemy/orm/unitofwork.py", line 642, in execute
+    util.preloaded.orm_persistence.save_obj(
+  File "/usr/local/lib/python3.11/site-packages/sqlalchemy/orm/persistence.py", line 93, in save_obj
+    _emit_insert_statements(
+  File "/usr/local/lib/python3.11/site-packages/sqlalchemy/orm/persistence.py", line 1233, in _emit_insert_statements
+    result = connection.execute(
+             ^^^^^^^^^^^^^^^^^^^
+  File "/usr/local/lib/python3.11/site-packages/sqlalchemy/engine/base.py", line 1419, in execute
+    return meth(
+           ^^^^^
+  File "/usr/local/lib/python3.11/site-packages/sqlalchemy/sql/elements.py", line 527, in _execute_on_connection
+    return connection._execute_clauseelement(
+           ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
+  File "/usr/local/lib/python3.11/site-packages/sqlalchemy/engine/base.py", line 1641, in _execute_clauseelement
+    ret = self._execute_context(
+          ^^^^^^^^^^^^^^^^^^^^^^
+  File "/usr/local/lib/python3.11/site-packages/sqlalchemy/engine/base.py", line 1846, in _execute_context
+    return self._exec_single_context(
+           ^^^^^^^^^^^^^^^^^^^^^^^^^^
+  File "/usr/local/lib/python3.11/site-packages/sqlalchemy/engine/base.py", line 1986, in _exec_single_context
+    self._handle_dbapi_exception(
+  File "/usr/local/lib/python3.11/site-packages/sqlalchemy/engine/base.py", line 2363, in _handle_dbapi_exception
+    raise sqlalchemy_exception.with_traceback(exc_info[2]) from e
+  File "/usr/local/lib/python3.11/site-packages/sqlalchemy/engine/base.py", line 1967, in _exec_single_context
+    self.dialect.do_execute(
+  File "/usr/local/lib/python3.11/site-packages/sqlalchemy/engine/default.py", line 952, in do_execute
+    cursor.execute(statement, parameters)
+sqlalchemy.exc.OperationalError: (sqlite3.OperationalError) table drafts has no column named focus_points
+[SQL: INSERT INTO drafts (artist_id, title, caption, hashtags, image_path, image_paths, image_url, video_url, video_path, audio_analysis, focus_points, status, created_at) VALUES (?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?)]
+[parameters: ('Veronica Intorcia', "Perdersi nell'Emozione", 'Lasciati trasportare dalle vibrazioni ipnotiche di "Lost In The Emotion". Una fusione magistrale tra Deep House e Melodic House, dove voci soul incon ... (151 characters truncated) ... ofisticata o vuole accendere l\'energia della notte. Ascolta ora il nuovo brano qui: https://distrokid.com/hyperfollow/veronika33/midnight-meridian-2', '#DeepHouse #MelodicHouse #ElectronicMusic #LostInTheEmotion #HouseMusic #SoulfulHouse #NewMusicAlert #HypnoticBeats #MusicJourney #NightVibes #MelodicTechno #IntrospectiveVibes #EuphoricMusic #DanceMusic #NewRelease', '', None, None, '', None, "Immergiti nelle profondità ipnotiche di 'Lost In The Emotion', una magistrale fusione di Deep House e Melodic House. Con le sue voci soul, il ritmo i ... (166 characters truncated) ... are un mood sofisticato o accendere l'energia notturna, è un viaggio sonoro accattivante, progettato per risuonare profondamente con gli ascoltatori.", None, 'pending', '2026-05-12 20:58:46.021420')]
+(Background on this error at: https://sqlalche.me/e/20/e3q8)
+
+
+--- Thu May 28 09:50:25 2026 ---
+❌ Errore critico nel task higgsfield (Veronica Intorcia): name 'subprocess' is not defined
+Traceback (most recent call last):
+  File "/Users/davidfrassi/SRC/agenti/agenzia/src/background_runner.py", line 136, in <module>
+    run_higgsfield(args.artist_id, args.prompt, dry_run=args.dry_run)
+    ~~~~~~~~~~~~~~^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
+  File "/Users/davidfrassi/SRC/agenti/agenzia/src/background_runner.py", line 114, in run_higgsfield
+    res = subprocess.run(cmd, capture_output=True, text=True)
+          ^^^^^^^^^^
+NameError: name 'subprocess' is not defined. Did you forget to import 'subprocess'?
+
@@ -0,0 +1,21 @@
+services:
+  web:
+    container_name: redazione_web
+    build: .
+    volumes:
+      - .:/app
+    ports:
+      - "8501:8501"
+    env_file:
+      - .env
+    command: streamlit run src/app.py --server.port=8501 --server.address=0.0.0.0
+
+  scheduler:
+    container_name: redazione_scheduler
+    build: .
+    volumes:
+      - .:/app
+      - ./videogenerati_test:/app/videogenerati
+    env_file:
+      - .env
+    command: python src/scheduler_service.py
@@ -0,0 +1,6 @@
+docker stop $(docker ps -aq) 2>/dev/null; \
+docker rm $(docker ps -aq) 2>/dev/null; \
+docker rmi -f $(docker images -aq) 2>/dev/null; \
+docker network prune -f; \
+docker volume prune -f; \
+docker system prune -a --volumes -f
@@ -0,0 +1,31 @@
+import os
+import requests
+import json
+from dotenv import load_dotenv
+
+load_dotenv()
+token = os.getenv("BUFFER_ACCESS_TOKEN")
+url = "https://api.buffer.com/graphql"
+headers = {"Authorization": f"Bearer {token}", "Content-Type": "application/json"}
+
+query = """
+query {
+  __type(name: "PostInputMetaData") {
+    name
+    inputFields {
+      name
+      type {
+        name
+        kind
+        ofType {
+          name
+          kind
+        }
+      }
+    }
+  }
+}
+"""
+
+resp = requests.post(url, headers=headers, json={'query': query})
+print(json.dumps(resp.json(), indent=2))
@@ -0,0 +1,48 @@
+#!/bin/bash
+
+# ==============================================================================
+# Script: list_soul_ids.sh
+# Descrizione: Si collega a Higgsfield AI tramite il token predisposto in .env
+#              e scarica l'elenco dei Soul ID dei personaggi con il loro nome.
+# ==============================================================================
+
+# Cambia directory sul percorso del progetto
+cd "$(dirname "$0")"
+
+echo "===================================================================="
+echo "⚡ Higgsfield Character Discovery Utility"
+echo "===================================================================="
+
+# 1. Carica le variabili dal file .env se presente
+if [ -f .env ]; then
+  # Esporta le variabili escludendo i commenti
+  export $(grep -v '^#' .env | xargs)
+  echo "✅ File .env caricato con successo."
+else
+  echo "⚠️ Attenzione: File .env non trovato. Verranno usate le variabili di sistema."
+fi
+
+# 2. Verifica la presenza di Python e dell'ambiente virtuale
+if [ -f ".venv/bin/python3" ]; then
+  PYTHON_EXE=".venv/bin/python3"
+  echo "✅ Utilizzo dell'ambiente virtuale locale (.venv)."
+else
+  PYTHON_EXE="python3"
+  echo "ℹ️ Ambiente virtuale locale non trovato. Utilizzo del python di sistema."
+fi
+
+# 3. Avvia lo script di recupero e formattazione
+echo "📡 Connessione a Higgsfield AI in corso..."
+echo ""
+
+$PYTHON_EXE -m src.list_higgsfield_characters
+
+EXIT_CODE=$?
+echo ""
+if [ $EXIT_CODE -eq 0 ]; then
+  echo "✅ Ricerca completata con successo."
+else
+  echo "❌ Errore durante il recupero dei dati da Higgsfield."
+fi
+echo "===================================================================="
+exit $EXIT_CODE
@@ -0,0 +1,14 @@
+langchain
+langgraph
+langchain-google-genai
+google-generativeai
+streamlit
+gdown
+python-dotenv
+Pillow
+sqlalchemy
+pydantic
+ffmpeg-python
+python-magic
+python-telegram-bot
+apscheduler
@@ -0,0 +1,343 @@
+import os
+import json
+from typing import TypedDict, List
+from langgraph.graph import StateGraph, END
+from audio_analyzer import AudioAnalyzer
+from database import Database
+from dotenv import load_dotenv
+import google.generativeai as genai
+
+load_dotenv()
+
+class AgentState(TypedDict):
+    artist_id: str
+    artist_name: str
+    spotify_url: str
+    distrokid_url: str
+    audio_analysis: str
+    image_path: str
+    image_url: str
+    video_url: str # NUOVO
+    image_description: str
+    climax_start_sec: int # NUOVO
+    bpm: int # NUOVO
+    title: str
+    caption: str
+    hashtags: str
+    review_status: str
+    narrative_style: str
+    social_tag: str # UNIFICATO
+
+class SocialAgents:
+    def __init__(self):
+        self.db = Database()
+        self.analyzer = AudioAnalyzer()
+        genai.configure(api_key=os.getenv("GOOGLE_API_KEY"))
+        self.model = genai.GenerativeModel('gemini-3-flash-preview')
+
+    def analyze_audio_node(self, state: AgentState):
+        audio = self.db.get_audio_for_artist(state['artist_id'])
+        if audio:
+            path = audio.file_path
+            if not os.path.exists(path) and audio.source_url:
+                os.makedirs("data/cache", exist_ok=True)
+                import gdown
+                # Forza estensione mp3 se il path non ce l'ha
+                if not path.endswith(('.mp3', '.wav', '.ogg')):
+                    path += ".mp3"
+                
+                print(f"📥 Download audio da Drive: {audio.source_url}...")
+                url = f"https://drive.google.com/uc?id={audio.source_url}"
+                gdown.download(url, output=path, quiet=True)
+            else:
+                print(f"✅ Audio già in cache: {path}")
+                
+            # Specifichiamo il mime_type se necessario per l'analyzer
+            raw_analysis = self.analyzer.analyze_audio(path)
+            
+            # PULIZIA IMMEDIATA per risparmiare spazio
+            if os.path.exists(path): os.remove(path)
+
+            try:
+                # Usa regex per estrarre solo il blocco JSON
+                import re
+                match = re.search(r'\{.*\}', raw_analysis, re.DOTALL)
+                if match:
+                    clean_json = match.group(0)
+                else:
+                    clean_json = raw_analysis.replace("```json", "").replace("```", "").strip()
+                
+                data = json.loads(clean_json)
+                return {
+                    "audio_analysis": data.get("analysis", raw_analysis),
+                    "climax_start_sec": data.get("climax_start_sec", 0),
+                    "bpm": data.get("bpm", 120)
+                }
+            except:
+                return {"audio_analysis": raw_analysis, "climax_start_sec": 0, "bpm": 120}
+        return {"audio_analysis": "Nessun brano trovato.", "bpm": 120}
+
+    def select_image_node(self, state: AgentState):
+        image = self.db.get_unused_image(state['artist_id'])
+        if image:
+            path = image.file_path
+            # Se il file non esiste localmente ma abbiamo l'ID Drive, lo scarichiamo in cache
+            if not os.path.exists(path) and image.source_url:
+                os.makedirs("data/cache", exist_ok=True)
+                import gdown
+                print(f"📥 Scaricando in cache: {path}")
+                try:
+                    url = f"https://drive.google.com/uc?id={image.source_url}"
+                    gdown.download(url, output=path, quiet=True)
+                except Exception as e:
+                    return {"caption": None, "error": f"Errore download immagine: {e}"}
+            else:
+                print(f"✅ Immagine già in cache: {path}")
+
+            if not os.path.exists(path):
+                return {"caption": None, "error": f"File non trovato in cache dopo il download: {path}"}
+
+            # Upload file to Gemini
+            img_data = genai.upload_file(path)
+            
+            # Wait for processing
+            while img_data.state.name == "PROCESSING":
+                import time
+                time.sleep(2)
+                img_data = genai.get_file(img_data.name)
+
+            response = self.model.generate_content(["Descrivi questa immagine per un post social musicale.", img_data])
+            
+            # Segna come usata per non ripeterla mai più
+            self.db.mark_as_used(image.id)
+            
+            # PULIZIA IMMEDIATA per risparmiare spazio
+            if os.path.exists(path): os.remove(path)
+            
+            return {"image_path": path, "image_description": response.text, "image_url": image.source_url}
+        return {"image_path": "", "image_description": "Nessuna immagine disponibile.", "image_url": None}
+
+    def write_caption_node(self, state: AgentState):
+        is_video = "video" if state.get("video_url") else "foto"
+        style_instructions = state.get("narrative_style", "scrivi un post accattivante")
+        
+        prompt = f"""
+        Sei un Social Manager esperto per l'artista {state['artist_name']}.
+        Il tuo obiettivo è scrivere un post che spacca seguendo queste ISTRUZIONI DI STILE:
+        ---
+        {style_instructions}
+        ---
+
+        REQUISITI AGGIUNTIVI:
+        - Tag Social: Inserisci il SOCIAL TAG dell'artista all'interno del CORPO del post in modo naturale.
+        - Connessione: Collega ironicamente quello che si vede nella {is_video} con il testo del brano.
+        - Conoscenza Nerd: Se le istruzioni lo richiedono, inserisci aneddoti o curiosità plausibili sugli autori originali.
+
+        CONTESTO:
+        ARTISTA: {state['artist_name']}
+        SOCIAL TAG: {state.get('social_tag', '')}
+        ANALISI AUDIO (Brano): {state['audio_analysis']}
+        DESCRIZIONE VISIVA ({is_video}): {state['image_description']}
+        LINK DI DESTINAZIONE: {state['distrokid_url']}
+        
+        FORMATO RISPOSTA (RESTITUISCI SOLO QUESTO):
+        TITOLO: [Titolo corto e accattivante]
+        CORPO: [Testo del post seguendo lo stile richiesto, termina con il link {state['distrokid_url']}]
+        """
+        response = self.model.generate_content(prompt).text
+        
+        # Parsing semplice
+        title = ""
+        body = ""
+        if "TITOLO:" in response and "CORPO:" in response:
+            parts = response.split("CORPO:")
+            title = parts[0].replace("TITOLO:", "").strip()
+            body = parts[1].strip()
+        else:
+            body = response # fallback
+
+        return {"title": title, "caption": body}
+
+    def generate_hashtags_node(self, state: AgentState):
+        prompt = f"Genera 15 hashtag per questo post: {state['caption']}. Rispondi SOLO con gli hashtag, senza testo aggiuntivo o consigli."
+        response = self.model.generate_content(prompt).text
+        # Rimuove eventuali testi residui fuori dagli hashtag
+        tags = [word for word in response.split() if word.startswith("#")]
+        return {"hashtags": " ".join(tags)}
+
+    def build_workflow(self):
+        workflow = StateGraph(AgentState)
+        
+        workflow.add_node("analyze_audio", self.analyze_audio_node)
+        workflow.add_node("select_image", self.select_image_node)
+        workflow.add_node("write_caption", self.write_caption_node)
+        workflow.add_node("generate_hashtags", self.generate_hashtags_node)
+        workflow.add_node("select_mixed_assets", self.select_mixed_assets_node)
+        workflow.add_node("generate_video", self.generate_video_node)
+        
+        workflow.set_entry_point("analyze_audio")
+        workflow.add_edge("analyze_audio", "select_mixed_assets")
+        workflow.add_edge("select_mixed_assets", "write_caption")
+        workflow.add_edge("write_caption", "generate_video")
+        workflow.add_edge("generate_video", "generate_hashtags")
+        workflow.add_edge("generate_hashtags", END)
+        
+        return workflow.compile()
+    def select_mixed_assets_node(self, state: AgentState):
+        assets = self.db.get_mixed_assets(state['artist_id'], total=12)
+        if not assets:
+             return {"image_path": "", "image_description": "Nessun asset disponibile."}
+        
+        # 1. Seleziona una copertina significativa (Preferibilmente immagine MAI USATA)
+        cover = next((a for a in assets if a.file_type == 'image' and not a.is_used), assets[0])
+        # Altre immagini
+        others = [a for a in assets if a.id != cover.id]
+        sorted_assets = [cover] + others
+        
+        # Marcatura immediata della cover per non riusarla
+        self.db.mark_as_used(cover.id)
+        
+        asset_paths = []
+        os.makedirs("data/cache", exist_ok=True)
+        import gdown
+        
+        for asset in sorted_assets:
+            path = asset.file_path
+            if not os.path.exists(path) and asset.source_url:
+                try:
+                    url = f"https://drive.google.com/uc?id={asset.source_url}"
+                    gdown.download(url, output=path, quiet=True)
+                    asset_paths.append(path)
+                except: continue
+            elif os.path.exists(path):
+                asset_paths.append(path)
+
+        # 3. AI DESCRIZIONE E FACE DETECTION per Pan-Zoom intelligente
+        focus_points = []
+        desc = "Un montaggio video dinamico."
+        try:
+            print(f"👁️ AI: Ricerca volti (Pinpoint) e descrizione per {len(asset_paths)} asset...")
+            image_files = []
+            for p in asset_paths:
+                if p.lower().endswith(('.jpg', '.jpeg', '.png')):
+                    image_files.append(genai.upload_file(p))
+            
+            if image_files:
+                # Chiediamo sia la descrizione che le coordinate in un colpo solo
+                prompt = "Analyze these images for a high-end music video. " \
+                         "1. For the first image (the cover), give a short poetic description. " \
+                         "2. For ALL images, identify the EXACT center coordinates (x, y as percentage 0-100) of the main subject's face. " \
+                         "If multiple people are present, pick the most attractive girl. " \
+                         "Return ONLY JSON: {'description': '...', 'focus': [{'x': val, 'y': val}, ...]}"
+                
+                response = self.model.generate_content(image_files + [prompt])
+                clean_json = response.text.replace('```json', '').replace('```', '').strip()
+                data = json.loads(clean_json)
+                desc = data.get('description', desc)
+                focus_points = data.get('focus', [])
+                print(f"✅ AI: Analisi completata. Focus points trovati: {len(focus_points)}")
+        except Exception as e:
+            print(f"⚠️ AI Regia fallita (uso default): {e}")
+            focus_points = [None] * len(asset_paths)
+
+        return {
+            "image_path": json.dumps(asset_paths),
+            "focus_points": json.dumps(focus_points),
+            "image_description": desc,
+            "image_url": cover.source_url 
+        }
+
+    def generate_video_node(self, state: AgentState):
+        from video_generator import VideoGenerator
+        from gdown import download
+        
+        audio = self.db.get_audio_for_artist(state['artist_id'])
+        if audio and state['image_path']:
+            vg = VideoGenerator()
+            audio_path = audio.file_path
+            if not os.path.exists(audio_path):
+                download(id=audio.source_url, output=audio_path, quiet=True)
+            
+            try:
+                # Se image_path è una lista JSON (caso video)
+                try:
+                    image_paths = json.loads(state['image_path'])
+                except:
+                    image_paths = [state['image_path']]
+                
+                print(f"🎬 AVVIO MONTAGGIO VIDEO: {len(image_paths)} asset + {audio_path} (Start: {state['climax_start_sec']}s, BPM: {state.get('bpm', 120)})")
+                focus_points = json.loads(state.get('focus_points', '[]'))
+                local_video = vg.generate_video(image_paths, audio_path, state['title'], start_time=state['climax_start_sec'], bpm=state.get('bpm', 120), focus_points=focus_points)
+                
+                print(f"🚀 Caricamento video su host temporaneo...")
+                video_url = vg.upload_to_ephemeral(local_video)
+                print(f"✅ Video caricato: {video_url}")
+                
+                # PULIZIA IMMEDIATA per risparmiare spazio
+                if os.path.exists(local_video): os.remove(local_video)
+                # La miniatura locale la teniamo solo se serve a Streamlit, ma qui la cancelliamo 
+                # perché abbiamo detto No-Space e useremo l'URL cloud se possibile.
+                # In realtà vg.generate_thumbnail crea un .jpg che potremmo voler tenere, 
+                # ma se l'utente vuole zero spazio, cancelliamo tutto.
+                
+                return {"video_url": video_url, "video_path": local_video}
+            except Exception as e:
+                print(f"❌ ERRORE CRITICO GENERAZIONE VIDEO: {e}")
+            finally:
+                # PULIZIA ASSET ORIGINALI (Sempre, anche se fallisce)
+                try:
+                    image_paths = json.loads(state['image_path'])
+                    for p in image_paths:
+                        if os.path.exists(p): os.remove(p)
+                except: pass
+        
+        return {"video_url": ""}
+
+    def run_for_artist(self, artist_data, mode="both"):
+        """Esegue i workflow. Mode: 'photo', 'video', 'both'"""
+        results = []
+        
+        # 1. Analisi Audio (comune)
+        initial_state = {
+            "artist_id": artist_data['id'],
+            "artist_name": artist_data['name'],
+            "spotify_url": artist_data['spotify_url'],
+            "distrokid_url": artist_data['distrokid_url'],
+            "audio_analysis": "",
+            "image_path": "",
+            "image_url": "",
+            "video_url": "",
+            "image_description": "",
+            "climax_start_sec": 0,
+            "bpm": 120,
+            "title": "",
+            "caption": "",
+            "hashtags": "",
+            "review_status": "pending",
+            "narrative_style": artist_data.get('narrative_style', ''),
+            "social_tag": artist_data.get('social_tag', '')
+        }
+        
+        audio_state = self.analyze_audio_node(initial_state)
+        initial_state.update(audio_state)
+        
+        # 2. Generazione BOZZA FOTO
+        if mode in ["photo", "both"]:
+            print(f"📸 Avvio Workflow FOTO per {artist_data['name']}...")
+            photo_state = initial_state.copy()
+            photo_state.update(self.select_image_node(photo_state))
+            photo_state.update(self.write_caption_node(photo_state))
+            photo_state.update(self.generate_hashtags_node(photo_state))
+            results.append(photo_state)
+        
+        # 3. Generazione BOZZA VIDEO
+        if mode in ["video", "both"]:
+            print(f"🎬 Avvio Workflow VIDEO per {artist_data['name']}...")
+            video_state = initial_state.copy()
+            video_state.update(self.select_mixed_assets_node(video_state))
+            video_state.update(self.write_caption_node(video_state))
+            video_state.update(self.generate_video_node(video_state))
+            video_state.update(self.generate_hashtags_node(video_state))
+            results.append(video_state)
+
+        return results
@@ -0,0 +1,643 @@
+import streamlit as st
+import json
+import os
+import sys
+import subprocess
+import glob
+import signal
+from database import Database
+from buffer_publisher import BufferPublisher
+from PIL import Image
+from dotenv import load_dotenv
+
+# Carichiamo le variabili d'ambiente (.env) prima di tutto
+load_dotenv()
+
+st.set_page_config(page_title="Redazione Social Multi-Agente", layout="wide")
+
+# Inizializziamo il database e il publisher
+db = Database()
+publisher = BufferPublisher()
+
+# Carichiamo la configurazione
+with open('config.json', 'r') as f:
+    config = json.load(f)
+
+artist_names = {a['id']: a['name'] for a in config['artists']}
+artist_ids = list(artist_names.keys())
+
+def load_credits_cache():
+    # Valori di default
+    cache_data = {"total_monthly": 1000, "remaining": 910, "generated_this_month": 90, "last_sync_status": "no_token"}
+    credits_file = "data/cache/credits.json"
+    try:
+        os.makedirs("data/cache", exist_ok=True)
+        if os.path.exists(credits_file):
+            with open(credits_file, 'r') as f:
+                loaded = json.load(f)
+                if isinstance(loaded, dict):
+                    cache_data.update(loaded)
+    except:
+        pass
+
+    # Se c'è HIGGSFIELD_TOKEN nel file .env, facciamo la sincronizzazione in tempo reale online
+    token = os.getenv("HIGGSFIELD_TOKEN")
+    if token and token.strip() and token.startswith("eyJ"):
+        import datetime
+        try:
+            from higgsfield_cli.client import HiggsClient
+            client = HiggsClient(token=token)
+            wallet = client.get_wallet()
+            
+            total = wallet.total_credits
+            remaining = int(wallet.credits_display)
+            generated = max(0, total - remaining)
+            
+            now_str = datetime.datetime.now().strftime("%Y-%m-%d %H:%M:%S")
+            cache_data.update({
+                "total_monthly": total,
+                "remaining": remaining,
+                "generated_this_month": generated,
+                "last_sync_status": "success",
+                "last_sync_time": now_str,
+                "sync_error": None
+            })
+            save_credits_cache(cache_data)
+        except Exception as e:
+            now_str = datetime.datetime.now().strftime("%Y-%m-%d %H:%M:%S")
+            cache_data.update({
+                "last_sync_status": "failed",
+                "last_sync_time": now_str,
+                "sync_error": str(e)
+            })
+            # Non sovrascriviamo i valori reali passati in caso di errore di rete/timeout temporaneo, aggiorniamo solo lo stato del sync
+            save_credits_cache(cache_data)
+    else:
+        cache_data["last_sync_status"] = "no_token"
+        
+    return cache_data
+
+def save_credits_cache(data):
+    try:
+        os.makedirs("data/cache", exist_ok=True)
+        credits_file = "data/cache/credits.json"
+        with open(credits_file, 'w') as f:
+            json.dump(data, f, indent=2)
+    except:
+        pass
+
+def get_running_tasks():
+    """Ritorna la lista dei task attualmente in corso (basandosi sui file .lock)"""
+    locks = glob.glob("data/tasks/*.lock")
+    tasks = []
+    for lock in locks:
+        name = os.path.basename(lock).replace(".lock", "")
+        tasks.append(name)
+    return tasks
+
+def create_manual_lock(task_name, artist_id="all"):
+    """Crea un file di lock manualmente per feedback immediato nella UI"""
+    os.makedirs("data/tasks", exist_ok=True)
+    lock_file = f"data/tasks/{task_name}_{artist_id.replace(' ', '_')}.lock"
+    with open(lock_file, "w") as f:
+        f.write("starting")
+    return lock_file
+
+def manage_task(task_id, action):
+    """Gestisce un task in background: stop, restart o delete lock"""
+    lock_path = f"data/tasks/{task_id}.lock"
+    if not os.path.exists(lock_path):
+        return
+
+    pid_str = open(lock_path).read().strip()
+    
+    # 1. Stop Process if exists
+    if pid_str.isdigit():
+        pid = int(pid_str)
+        try:
+            os.kill(pid, signal.SIGTERM)
+            print(f"Inviato SIGTERM a PID {pid}")
+        except ProcessLookupError:
+            pass
+        except Exception as e:
+            st.error(f"Errore chiusura processo: {e}")
+
+    # 2. Perform Action
+    if action == "stop" or action == "remove":
+        if os.path.exists(lock_path):
+            os.remove(lock_path)
+    
+    if action == "restart":
+        # Determina i parametri dal task_id (es: generate_Ambra_Manca)
+        parts = task_id.split("_")
+        task_type = parts[0] # sync o generate
+        artist_id = "_".join(parts[1:]) if len(parts) > 1 else "all"
+        artist_id = artist_id.replace("_", " ") # Ripristina spazi
+        
+        if os.path.exists(lock_path):
+            os.remove(lock_path)
+            
+        create_manual_lock(task_type, artist_id)
+        cmd = [sys.executable, "src/background_runner.py", "--task", task_type]
+        if artist_id != "all":
+            cmd += ["--artist_id", artist_id]
+        subprocess.Popen(cmd)
+
+# ============================================================
+# SIDEBAR: Selezione Artista + Configurazione
+# ============================================================
+st.sidebar.header("🎤 Seleziona Artista")
+
+selected_artist_idx = st.sidebar.selectbox(
+    "Artista",
+    options=range(len(config['artists'])),
+    format_func=lambda i: config['artists'][i]['name'],
+    label_visibility="collapsed"
+)
+
+# Rilevamento cambio artista per pulizia stato (Previene ghosting)
+if "last_artist_idx" not in st.session_state:
+    st.session_state.last_artist_idx = selected_artist_idx
+
+if st.session_state.last_artist_idx != selected_artist_idx:
+    st.session_state.last_artist_idx = selected_artist_idx
+    # Pulizia totale per evitare che i checkbox o le foto rimangano in trasparenza
+    st.session_state.clear()
+    st.rerun()
+
+selected_artist = config['artists'][selected_artist_idx]
+selected_artist_id = selected_artist['id']
+selected_artist_name = selected_artist['name']
+
+st.sidebar.markdown("---")
+
+# Pannello di Configurazione
+with st.sidebar.expander(f"⚙️ Configurazione {selected_artist_name}", expanded=False):
+    with st.form(key=f"form_edit_{selected_artist_idx}"):
+        new_name = st.text_input("Nome Artista", selected_artist.get('name', ''))
+        new_spotify = st.text_input("Spotify URL", selected_artist.get('spotify_url', ''))
+        new_distrokid = st.text_input("DistroKid URL", selected_artist.get('distrokid_url', ''))
+        new_time = st.text_input("Orario Pubblicazione (es. 09:00)", selected_artist.get('schedule_time', ''))
+        new_audio = st.text_input("Cartella Drive Audio (URL)", selected_artist.get('drive_audio_url', ''))
+        new_img = st.text_input("Cartella Drive Immagini (URL)", selected_artist.get('drive_images_url', ''))
+        new_vid = st.text_input("Cartella Drive Video (URL)", selected_artist.get('drive_videos_url', ''))
+        new_buffer_token = st.text_input("Buffer Access Token (Specifico)", selected_artist.get('buffer_token', ''), type="password")
+        new_style = st.text_area("Stile Narrativo (Istruzioni AI)", selected_artist.get('narrative_style', ""), height=100)
+        new_social_tag = st.text_input("Social Tag (es. @artista)", selected_artist.get('social_tag', ''))
+        new_soul_id = st.text_input("Higgsfield Soul ID 2.0", selected_artist.get('soul_id', ''))
+        new_starting_photos = st.text_input("Cartella Drive Foto Partenza / Modelli (URL)", selected_artist.get('drive_starting_photos_url', ''))
+
+        
+        if st.form_submit_button("💾 Salva Modifiche"):
+            config['artists'][selected_artist_idx]['name'] = new_name
+            config['artists'][selected_artist_idx]['spotify_url'] = new_spotify
+            config['artists'][selected_artist_idx]['distrokid_url'] = new_distrokid
+            config['artists'][selected_artist_idx]['schedule_time'] = new_time
+            config['artists'][selected_artist_idx]['drive_audio_url'] = new_audio
+            config['artists'][selected_artist_idx]['drive_images_url'] = new_img
+            config['artists'][selected_artist_idx]['drive_videos_url'] = new_vid
+            config['artists'][selected_artist_idx]['buffer_token'] = new_buffer_token
+            config['artists'][selected_artist_idx]['narrative_style'] = new_style
+            config['artists'][selected_artist_idx]['social_tag'] = new_social_tag
+            config['artists'][selected_artist_idx]['soul_id'] = new_soul_id
+            config['artists'][selected_artist_idx]['drive_starting_photos_url'] = new_starting_photos
+
+            
+            # Rimuovo i vecchi campi se esistenti per pulizia
+            for old_key in ['instagram_tag', 'tiktok_tag', 'instagram_profile_id', 'tiktok_profile_id']:
+                config['artists'][selected_artist_idx].pop(old_key, None)
+            
+            with open('config.json', 'w') as f:
+                json.dump(config, f, indent=2)
+            st.success(f"Configurazione di {new_name} salvata!")
+            st.rerun()
+
+st.sidebar.markdown("---")
+
+running_tasks = get_running_tasks()
+
+if running_tasks:
+    st.sidebar.warning("⏳ Task in background:")
+    for task_id in running_tasks:
+        col_t, col_stop, col_rel, col_del = st.sidebar.columns([4, 1, 1, 1])
+        col_t.caption(f"• {task_id.replace('_', ' ')}")
+        
+        if col_stop.button("🛑", key=f"stop_{task_id}", help="Ferma"):
+            manage_task(task_id, "stop")
+            st.rerun()
+        if col_rel.button("🔄", key=f"rel_{task_id}", help="Riavvia"):
+            manage_task(task_id, "restart")
+            st.rerun()
+        if col_del.button("🗑️", key=f"del_{task_id}", help="Rimuovi Lock"):
+            manage_task(task_id, "remove")
+            st.rerun()
+
+if st.sidebar.button("🔄 Sincronizza Asset da Drive"):
+    if any(t.startswith("sync") for t in running_tasks):
+        st.sidebar.error("Sincronizzazione già in corso!")
+    else:
+        create_manual_lock("sync")
+        subprocess.Popen([sys.executable, "src/background_runner.py", "--task", "sync"])
+        st.sidebar.success("✅ Sincronizzazione avviata in background!")
+        st.rerun()
+
+if st.sidebar.button("🗑️ Svuota Cache Locale"):
+    import shutil
+    try:
+        shutil.rmtree("data/cache")
+        os.makedirs("data/cache", exist_ok=True)
+        st.sidebar.success("✅ Cache svuotata con successo!")
+        st.session_state.clear()
+        st.rerun()
+    except Exception as e:
+        st.sidebar.error(f"Errore durante la pulizia: {e}")
+
+# ============================================================
+# AREA PRINCIPALE: Generazione e Gestione Bozze
+# ============================================================
+with st.container(key=f"main_content_{selected_artist_id}"):
+    st.title(f"🚀 Redazione: {selected_artist_name}")
+
+    # Sezione Generazione Rapida
+    st.subheader("✨ Nuova Generazione")
+    col_gen1, col_gen2 = st.columns(2)
+
+    gen_task_id_photo = f"generate_{selected_artist_id.replace(' ', '_')}_photo"
+    gen_task_id_video = f"generate_{selected_artist_id.replace(' ', '_')}_video"
+
+    # 1. Post Immagine (Leggero)
+    if gen_task_id_photo in running_tasks:
+        col_gen1.info("📸 Post Foto in corso...")
+    elif col_gen1.button("📸 Post Immagine (Solo 1 Foto)", use_container_width=True, key=f"btn_p_{selected_artist_id}"):
+        create_manual_lock("generate", f"{selected_artist_id}_photo")
+        subprocess.Popen([sys.executable, "src/background_runner.py", "--task", "generate", "--artist_id", selected_artist_id, "--mode", "photo"])
+        st.rerun()
+
+    # 2. Video Mix (Più pesante)
+    if gen_task_id_video in running_tasks:
+        col_gen2.info("🎬 Video Mix in corso...")
+    elif col_gen2.button("🎬 Video Mix (12 Asset)", use_container_width=True, key=f"btn_v_{selected_artist_id}"):
+        create_manual_lock("generate", f"{selected_artist_id}_video")
+        subprocess.Popen([sys.executable, "src/background_runner.py", "--task", "generate", "--artist_id", selected_artist_id, "--mode", "video"])
+        st.rerun()
+
+    st.markdown("---")
+    st.subheader("🎨 Generazione Foto Higgsfield AI (Soul ID)")
+    
+    gen_task_id_higgsfield = f"higgsfield_{selected_artist_id.replace(' ', '_')}_both"
+    
+    col_h1, col_h2 = st.columns([3, 1])
+    hf_prompt = col_h1.text_input("Prompt Stile Estetico", "A beautiful editorial studio portrait, highly detailed, cinematic studio lighting, professional photography, 8k resolution, crisp details", key=f"hf_p_{selected_artist_id}")
+    hf_dry_run = col_h2.checkbox("Simulazione (Dry-run)", value=False, key=f"hf_dr_{selected_artist_id}")
+    
+    if gen_task_id_higgsfield in running_tasks:
+        st.info("🎨 Generazione Higgsfield AI in corso in background...")
+    else:
+        if st.button("🚀 Avvia Generazione Higgsfield AI per tutti i modelli", use_container_width=True, type="primary", key=f"btn_hf_{selected_artist_id}"):
+            # Check parameters
+            m_url = selected_artist.get("drive_starting_photos_url")
+            if not m_url:
+                st.error("Errore: Assicurati di aver configurato la 'Cartella Foto Partenza' nella barra laterale!")
+            else:
+                create_manual_lock("higgsfield", f"{selected_artist_id}_both")
+                cmd = [
+                    sys.executable, "src/background_runner.py", 
+                    "--task", "higgsfield", 
+                    "--artist_id", selected_artist_id,
+                    "--prompt", hf_prompt
+                ]
+                if hf_dry_run:
+                    cmd.append("--dry_run")
+                subprocess.Popen(cmd)
+                st.success("✅ Generazione Higgsfield avviata in background!")
+                st.rerun()
+
+    st.markdown("---")
+
+# ============================================================
+# BOZZE PENDENTI (filtrate per artista selezionato)
+# ============================================================
+st.subheader("📝 Bozze Pendenti")
+drafts = db.get_pending_drafts(selected_artist_id)
+
+if not drafts:
+    st.info(f"Nessuna bozza per {selected_artist_name}. Clicca su 'Genera Nuovo Post' per iniziare.")
+else:
+    # Callback per seleziona/deseleziona tutto
+    if "current_draft_ids" not in st.session_state:
+        st.session_state.current_draft_ids = []
+    st.session_state.current_draft_ids = [d.id for d in drafts]
+
+    def toggle_all_drafts():
+        val = st.session_state.select_all_drafts_chk
+        for d_id in st.session_state.current_draft_ids:
+            st.session_state[f"sel_{d_id}"] = val
+
+    # Pre-calcolo delle bozze selezionate per evitare NameError nelle azioni bulk
+    selected_drafts = [d for d in drafts if st.session_state.get(f"sel_{d.id}", False)]
+
+    # Azioni di Gruppo
+    with st.expander("🛠️ Azioni di Gruppo", expanded=False):
+        col_btn1, col_btn2 = st.columns(2)
+        if col_btn1.button("🚀 PUBBLICA TUTTI I SELEZIONATI"):
+            if selected_drafts:
+                with st.spinner(f"Pubblicazione di {len(selected_drafts)} post..."):
+                    for d in selected_drafts:
+                        t_val = st.session_state.get(f"t_{d.id}", d.title)
+                        c_val = st.session_state.get(f"c_{d.id}", d.caption)
+                        h_val = st.session_state.get(f"h_{d.id}", d.hashtags)
+                        db.update_draft(d.id, title=t_val, caption=c_val, hashtags=h_val)
+                        full_text = f"*{t_val}*\n\n{c_val}\n\n{h_val}"
+                        
+                        # Recupera dati specifici artista (Token e Profile IDs)
+                        artist_data = next((a for a in config['artists'] if a['id'] == d.artist_id), None)
+                        artist_token = artist_data.get("buffer_token") if artist_data else None
+                        
+                        # Passiamo profile_ids vuoti per attivare l'auto-discovery nel publisher
+                        profile_ids = []
+                        publisher.publish(profile_ids, full_text, d.image_path, d.video_url, d.image_paths, artist_token=artist_token)
+                        db.mark_as_published(d.id)
+                st.success(f"Pubblicati {len(selected_drafts)} post!")
+                st.rerun()
+
+        if col_btn2.button("🗑️ ELIMINA TUTTI I SELEZIONATI"):
+            if selected_drafts:
+                for d in selected_drafts:
+                    db.delete_draft(d.id)
+                st.warning(f"Eliminati {len(selected_drafts)} post.")
+                st.rerun()
+        
+        st.checkbox("Seleziona Tutte le Bozze", key="select_all_drafts_chk", on_change=toggle_all_drafts)
+
+    for draft in drafts:
+        with st.expander(f"📌 {draft.title or 'Bozza senza titolo'}", expanded=len(drafts) == 1):
+            c1, c2 = st.columns([1, 2])
+
+            with c1:
+                st.checkbox("Seleziona per Azione di Gruppo", key=f"sel_{draft.id}")
+
+                # ANTEPRIMA VIDEO (Priorità assoluta se esiste)
+                if draft.video_url:
+                    st.video(draft.video_url)
+                    st.success("🎬 Video 'CapCut-style' generato!")
+                
+                # Anteprima Immagine (Metodo Base64 Infallibile)
+                display_url = draft.image_url
+                if not display_url and draft.image_path:
+                    m_rec = db.get_media_by_path(draft.image_path)
+                    if m_rec: display_url = m_rec.source_url
+                
+                if display_url:
+                    try:
+                        import requests, base64
+                        # Usiamo un URL di Drive che restituisce direttamente l'immagine
+                        thumb_url = f"https://drive.google.com/thumbnail?id={display_url}&sz=w600"
+                        response = requests.get(thumb_url, timeout=5)
+                        if response.status_code == 200:
+                            b64_img = base64.b64encode(response.content).decode()
+                            st.markdown(f'<img src="data:image/jpeg;base64,{b64_img}" style="width:100%; border-radius: 10px; margin-bottom: 10px;">', unsafe_allow_html=True)
+                        else:
+                            # Fallback se il server non raggiunge Google
+                            st.markdown(f'<img src="https://lh3.googleusercontent.com/u/0/d/{display_url}=w600" style="width:100%; border-radius: 10px;">', unsafe_allow_html=True)
+                    except:
+                        st.warning("🖼️ Errore caricamento (Cloud)")
+                elif draft.image_path and os.path.exists(draft.image_path):
+                    try:
+                        import base64
+                        with open(draft.image_path, "rb") as img_file:
+                            b64_img = base64.b64encode(img_file.read()).decode()
+                        st.markdown(f'<img src="data:image/jpeg;base64,{b64_img}" style="width:100%; border-radius: 10px; margin-bottom: 10px;">', unsafe_allow_html=True)
+                    except Exception as e:
+                        st.image(draft.image_path)
+                elif draft.image_paths:
+
+                    try:
+                        paths = json.loads(draft.image_paths)
+                        m_rec = db.get_media_by_path(paths[0])
+                        if m_rec and m_rec.source_url:
+                            img_url = f"https://lh3.googleusercontent.com/u/0/d/{m_rec.source_url}=w600"
+                            st.markdown(f'<img src="{img_url}" style="width:100%; border-radius: 10px;">', unsafe_allow_html=True)
+                    except: pass
+                elif not draft.video_url:
+                    st.warning("⚠️ Anteprima non disponibile")
+
+            with c2:
+                edited_title = st.text_input("Titolo", draft.title, key=f"t_{draft.id}")
+                edited_caption = st.text_area("Testo", draft.caption, height=150, key=f"c_{draft.id}")
+                edited_tags = st.text_input("Hashtag", draft.hashtags, key=f"h_{draft.id}")
+
+                if draft.audio_analysis:
+                    with st.expander("🎵 Pitch per Playlist Spotify (Generato dall'AI)", expanded=True):
+                        st.info(draft.audio_analysis)
+
+                if st.button("🤖 Autocompleta Testi e Pitch con AI (Ascolta Audio)", key=f"ai_{draft.id}", use_container_width=True):
+                    with st.spinner("Ascolto audio e generazione testi in corso..."):
+                        from agents import SocialAgents
+                        agents = SocialAgents()
+                        audio = db.get_audio_for_artist(draft.artist_id)
+                        if not audio:
+                            st.error("Nessun audio trovato per questo artista.")
+                        else:
+                            state = {"artist_id": draft.artist_id, "artist_name": artist_names.get(draft.artist_id), "distrokid_url": ""}
+                            if artist_data: 
+                                state["distrokid_url"] = artist_data.get("distrokid_url", "")
+                                state["narrative_style"] = artist_data.get("narrative_style", "")
+                                state["social_tag"] = artist_data.get("social_tag", "")
+
+                            analysis_res = agents.analyze_audio_node(state)
+                            state.update(analysis_res)
+                            state["image_description"] = "Un bellissimo post per i social."
+
+                            cap_res = agents.write_caption_node(state)
+                            state.update(cap_res)
+                            tag_res = agents.generate_hashtags_node(state)
+
+                            db.update_draft(
+                                draft.id,
+                                title=cap_res.get("title", ""),
+                                caption=cap_res.get("caption", ""),
+                                hashtags=tag_res.get("hashtags", ""),
+                                audio_analysis=analysis_res.get("audio_analysis", "")
+                            )
+                            st.rerun()
+
+                if st.button(f"✅ Salva Modifiche e Pubblica", key=f"pub_{draft.id}", type="primary", use_container_width=True):
+                    db.update_draft(draft.id, title=edited_title, caption=edited_caption, hashtags=edited_tags)
+                    full_text = f"*{edited_title}*\n\n{edited_caption}\n\n{edited_tags}"
+                    
+                    # Recupera dati specifici artista (Token e Profile IDs)
+                    artist_data = next((a for a in config['artists'] if a['id'] == draft.artist_id), None)
+                    artist_token = artist_data.get("buffer_token") if artist_data else None
+                    
+                    # Passiamo profile_ids vuoti per attivare l'auto-discovery nel publisher
+                    profile_ids = []
+                    res = publisher.publish(profile_ids, full_text, draft.image_path, draft.video_url, draft.image_paths, artist_token=artist_token)
+                    if res['success']:
+                        db.mark_as_published(draft.id)
+                        st.success("Inviato a Buffer!")
+                        st.rerun()
+                    else:
+                        st.error(f"Errore: {res['error']}")
+                if st.button("🗑️ Elimina Bozza", key=f"del_single_{draft.id}", type="secondary", use_container_width=True):
+                    db.delete_draft(draft.id)
+                    st.rerun()
+
+            # Sezione Sostituzione Media / Generazione Video Manuale
+            st.markdown("---")
+            st.write("📸 **Sostituisci Media / Genera Video (Opzionale)**")
+            st.caption("Seleziona una o più foto/video recenti per generare un video animato.")
+
+            recent_media = db.get_recent_media(draft.artist_id, limit=10)
+            selected_images = []
+            cols_img = st.columns(5)
+            for i, media in enumerate(recent_media):
+                with cols_img[i % 5]:
+                    img_to_show = None
+                    if os.path.exists(media.file_path):
+                        img_to_show = media.file_path
+                    elif media.source_url:
+                        # MINIATURA DIRETTA DA GOOGLE DRIVE (Metodo più robusto)
+                        img_to_show = f"https://drive.google.com/thumbnail?id={media.source_url}&sz=w400"
+                    
+                    if img_to_show:
+                        if media.file_type == 'video':
+                            st.markdown(f"🎥 **Video**")
+                        
+                        try:
+                            # TENTATIVO CLOUD-FIRST (Lookup ID da database)
+                            m_id = media.source_url
+                            if not m_id:
+                                m_rec = db.get_media_by_path(media.file_path)
+                                if m_rec: m_id = m_rec.source_url
+                            
+                            if m_id:
+                                try:
+                                    import requests, base64
+                                    thumb_url = f"https://drive.google.com/thumbnail?id={m_id}&sz=w200"
+                                    response = requests.get(thumb_url, timeout=3)
+                                    if response.status_code == 200:
+                                        b64_img = base64.b64encode(response.content).decode()
+                                        st.markdown(f'<img src="data:image/jpeg;base64,{b64_img}" style="width:100%; border-radius: 5px;">', unsafe_allow_html=True)
+                                    else:
+                                        st.markdown(f'<img src="https://lh3.googleusercontent.com/u/0/d/{m_id}=w200" style="width:100%; border-radius: 5px;">', unsafe_allow_html=True)
+                                except:
+                                    st.caption("🖼️ Cloud N/A")
+                            else:
+                                st.caption("🖼️ No ID")
+                        except Exception:
+                            st.caption("🖼️ Error")
+                        
+                        if st.checkbox("Scegli", key=f"chk_m_{draft.id}_{media.id}", label_visibility="collapsed"):
+                            selected_images.append(media.file_path)
+
+            if st.button("🎬 Genera Video Animato (CapCut style) con i media selezionati", key=f"gen_{draft.id}"):
+                if not selected_images:
+                    st.error("Seleziona almeno un media!")
+                else:
+                    with st.spinner("Generazione video in corso..."):
+                        # Sincronizza/Scarica media
+                        final_local_media = []
+                        import requests
+                        for img_p in selected_images:
+                            if not os.path.exists(img_p):
+                                m_rec = db.get_media_by_path(img_p)
+                                if m_rec and m_rec.source_url:
+                                    try:
+                                        dl_url = f"https://drive.google.com/uc?id={m_rec.source_url}"
+                                        resp = requests.get(dl_url, timeout=15)
+                                        if resp.status_code == 200:
+                                            os.makedirs(os.path.dirname(img_p), exist_ok=True)
+                                            with open(img_p, 'wb') as f:
+                                                f.write(resp.content)
+                                            final_local_media.append(img_p)
+                                    except: pass
+                            else:
+                                final_local_media.append(img_p)
+
+                        from video_generator import VideoGenerator
+                        vg = VideoGenerator()
+                        audio = db.get_audio_for_artist(draft.artist_id)
+                        if not audio:
+                            st.error("Nessun file audio trovato per questo artista!")
+                        else:
+                            audio_path = audio.file_path
+                            if not os.path.exists(audio_path) and audio.source_url:
+                                try:
+                                    dl_url = f"https://drive.google.com/uc?id={audio.source_url}"
+                                    resp = requests.get(dl_url, timeout=30)
+                                    if resp.status_code == 200:
+                                        os.makedirs(os.path.dirname(audio_path), exist_ok=True)
+                                        with open(audio_path, 'wb') as f:
+                                            f.write(resp.content)
+                                except: pass
+
+                            try:
+                                # Percorso ASSOLUTO interno a Docker per garantire il sync con il volume
+                                out_base = "/app/data/outputs"
+                                os.makedirs(out_base, exist_ok=True)
+                                
+                                safe_artist = draft.artist_id.replace(" ", "")
+                                out_path = f"{out_base}/{safe_artist}_{draft.id}.mp4"
+                                
+                                vid_path = vg.generate_video(final_local_media, audio_path, edited_title, start_time=0, output_path=out_path)
+                                thumb_path = vg.generate_thumbnail(vid_path)
+                                video_url = vg.upload_to_ephemeral(vid_path)
+                                if not video_url:
+                                    st.warning("Upload remoto fallito. Il video è disponibile solo localmente.")
+                                db.update_draft(draft.id, video_url=video_url, video_path=vid_path)
+                                st.rerun()
+                            except Exception as e:
+                                st.error(f"Errore generazione video: {e}")
+
+# ============================================================
+# STATO CONSUMO CREDITI HIGGSFIELD AI (In basso alla console)
+# ============================================================
+st.markdown("---")
+st.subheader("📊 Stato Consumo Crediti Higgsfield AI")
+
+credits_data = load_credits_cache()
+total = credits_data.get("total_monthly", 1000)
+remaining = credits_data.get("remaining", 1000)
+consumed = total - remaining
+
+# Visualizzazione badge di stato del sync online
+sync_status = credits_data.get("last_sync_status", "no_token")
+sync_time = credits_data.get("last_sync_time", "Mai")
+sync_err = credits_data.get("sync_error", "")
+
+if sync_status == "success":
+    st.markdown(f"<span style='background-color:#d4ff00; color:#000000; padding:4px 8px; border-radius:4px; font-weight:bold; font-size:12px;'>🟢 SINCRONIZZATO ONLINE</span> &nbsp;&nbsp;<span style='font-size:12px; color:#888888;'>Ultimo aggiornamento automatico: {sync_time}</span>", unsafe_allow_html=True)
+elif sync_status == "failed":
+    st.markdown(f"<span style='background-color:#ff4b4b; color:#ffffff; padding:4px 8px; border-radius:4px; font-weight:bold; font-size:12px;'>🔴 ERRORE DI CONNESSE</span> &nbsp;&nbsp;<span style='font-size:12px; color:#888888;'>Impossibile allineare i dati online ({sync_err}). Utilizzo cache del {sync_time}</span>", unsafe_allow_html=True)
+else:
+    st.markdown("<span style='background-color:#ffaa00; color:#000000; padding:4px 8px; border-radius:4px; font-weight:bold; font-size:12px;'>🟡 CONTATORE LOCALE (CACHE)</span> &nbsp;&nbsp;<span style='font-size:12px; color:#888888;'>Token .env non configurato</span>", unsafe_allow_html=True)
+    st.info("💡 **Allineamento Online in Tempo Reale**: Per allineare automaticamente il consumo effettivo con il tuo account Higgsfield AI (Pro Plan) senza usare Chrome Safe Storage, aggiungi la variabile `HIGGSFIELD_TOKEN=eyJ...` nel file `.env` (copiabile dal cookie `__session` della console browser di higgsfield.ai).")
+
+# Calculate percentage
+pct_remaining = (remaining / total * 100) if total > 0 else 0.0
+
+# Render elegant columns
+col_c1, col_c2, col_c3 = st.columns(3)
+col_c1.metric("Crediti Rimanenti", f"{remaining} crediti", f"{pct_remaining:.1f}% rimasti", delta_color="normal")
+col_c2.metric("Crediti Consumati", f"{consumed} crediti", f"{100 - pct_remaining:.1f}% consumati", delta_color="inverse")
+col_c3.metric("Budget Mensile", f"{total} crediti", help="Questo valore è allineato in tempo reale con il tuo abbonamento online")
+
+# Beautiful, customized progress bar matching Higgsfield's neon lime styling
+progress_html = f"""
+<div style="font-weight: bold; font-size: 16px; margin-top: 15px; margin-bottom: 5px; font-family: sans-serif;">
+    ⚡ {pct_remaining:.0f}% credits left
+</div>
+<div style="background-color: #333333; border-radius: 10px; width: 100%; height: 16px; overflow: hidden; margin-bottom: 25px; border: 1px solid #222222; box-shadow: inset 0 1px 3px rgba(0,0,0,0.2);">
+    <div style="background: linear-gradient(90deg, #d4ff00 0%, #bfff00 100%); width: {pct_remaining:.1f}%; height: 100%; transition: width 0.5s ease-in-out;">
+    </div>
+</div>
+"""
+st.markdown(progress_html, unsafe_allow_html=True)
+
+# ============================================================
+# AUTO-REFRESH (Solo se ci sono task in corso)
+# ============================================================
+if running_tasks:
+    import time
+    time.sleep(5)
+    st.rerun()
@@ -0,0 +1,54 @@
+import os
+import google.generativeai as genai
+from dotenv import load_dotenv
+
+load_dotenv()
+
+class AudioAnalyzer:
+    def __init__(self, api_key=None):
+        self.api_key = api_key or os.getenv("GOOGLE_API_KEY")
+        if self.api_key:
+            genai.configure(api_key=self.api_key)
+            self.model = genai.GenerativeModel('gemini-2.5-flash')
+        else:
+            self.model = None
+
+    def analyze_audio(self, file_path):
+        if not self.model:
+            return "API Key non configurata."
+        
+        if not os.path.exists(file_path):
+            return f"File non trovato: {file_path}"
+
+        try:
+            # Rilevamento mime-type per evitare errori
+            mime_type = "audio/mpeg"
+            if file_path.endswith(".wav"): mime_type = "audio/wav"
+            elif file_path.endswith(".ogg"): mime_type = "audio/ogg"
+            
+            # Upload file to Gemini con mime_type esplicito
+            print(f"🎵 Caricamento audio su Gemini: {file_path}...")
+            audio_file = genai.upload_file(path=file_path, mime_type=mime_type)
+            
+            # Wait for processing
+            while audio_file.state.name == "PROCESSING":
+                import time
+                print("...elaborazione audio in corso su Gemini...")
+                time.sleep(2)
+                audio_file = genai.get_file(audio_file.name)
+            
+            print("🧠 Analisi audio in corso con AI...")
+
+            prompt = """
+        Analizza questo brano musicale e rispondi in formato JSON con queste chiavi:
+        1. "climax_start_sec": (intero) Il secondo esatto in cui inizia la parte più energica o il ritornello (es. 45). ASSICURATI che in questo punto l'audio sia chiaramente udibile (evita silenzi iniziali o intro troppo lunghe).
+        2. "bpm": (intero) I battiti per minuto del brano (es. 128). Se non sei sicuro, fornisci una stima accurata basata sul ritmo.
+        3. "analysis": Scrivi un "Pitch" accattivante e professionale (max 3-4 frasi) pronto da inviare ai curatori delle playlist di Spotify, mettendo in luce il genere, il mood, le vibes e i punti di forza del brano.
+        
+        Rispondi SOLO ed ESCLUSIVAMENTE con il JSON puro. Non includere blocchi markdown come ```json o ```, inizia direttamente con { e finisci con }.
+        """
+            
+            response = self.model.generate_content([prompt, audio_file])
+            return response.text
+        except Exception as e:
+            return f"Errore durante l'analisi audio: {str(e)}"
@@ -0,0 +1,146 @@
+import os
+import sys
+import json
+import time
+import argparse
+from database import Database
+from agents import SocialAgents
+
+def create_lock(task_name, artist_id="all"):
+    os.makedirs("data/tasks", exist_ok=True)
+    lock_file = f"data/tasks/{task_name}_{artist_id.replace(' ', '_')}.lock"
+    with open(lock_file, "w") as f:
+        f.write(str(os.getpid()))
+    return lock_file
+
+def remove_lock(lock_file):
+    if os.path.exists(lock_file):
+        os.remove(lock_file)
+
+def run_sync():
+    print("🚀 Avvio Sincronizzazione in background...")
+    import cloud_sync
+    cloud_sync.sync_from_drive()
+    print("✅ Sincronizzazione completata.")
+
+def run_generate(artist_id, mode="both"):
+    db = Database()
+    agents = SocialAgents()
+    
+    with open('config.json', 'r') as f:
+        config = json.load(f)
+        artist = next((a for a in config['artists'] if a['id'] == artist_id), None)
+    
+    if not artist:
+        print(f"❌ Artista non trovato: {artist_id}")
+        return
+
+    print(f"🤖 Avvio generazione ({mode}) per {artist['name']}...")
+    # Sincronizzazione preventiva (leggera)
+    import cloud_sync
+    cloud_sync.sync_from_drive()
+    
+    results = agents.run_for_artist(artist, mode=mode)
+    for res in results:
+        if 'error' not in res:
+            db.save_draft(
+                artist_id=artist_id,
+                title=res.get('title', 'Nuovo Post'),
+                caption=res.get('caption', ''),
+                hashtags=res.get('hashtags', ''),
+                image_path=res.get('image_path'),
+                image_url=res.get('image_url'),
+                video_url=res.get('video_url'),
+                audio_analysis=res.get('audio_analysis'),
+                video_path=res.get('video_path'),
+                focus_points=res.get('focus_points')
+            )
+    print(f"✅ Generazione completata per {artist['name']}.")
+
+def run_higgsfield(artist_id, prompt, dry_run=False):
+    print(f"🎨 Avvio task Higgsfield per {artist_id}...")
+    with open('config.json', 'r') as f:
+        config = json.load(f)
+        artist = next((a for a in config['artists'] if a['id'] == artist_id), None)
+    
+    if not artist:
+        print(f"❌ Artista non trovato: {artist_id}")
+        return
+        
+    models_url = artist.get('drive_starting_photos_url')
+    if not models_url:
+        print(f"❌ Nessuna cartella Drive Foto Partenza configurata per {artist_id}")
+        return
+        
+    soul_id = artist.get('soul_id')
+    if not soul_id:
+        print(f"❌ Nessun Higgsfield Soul ID configurato per {artist_id}")
+        return
+        
+    # 1. Download starting photos from Drive
+    import cloud_sync
+    import gdown
+    
+    fid = cloud_sync.get_drive_id(models_url)
+    local_dir = f"data/starting_photos/{artist_id.replace(' ', '_')}"
+    os.makedirs(local_dir, exist_ok=True)
+    
+    print(f"📥 Controllo cartella Drive per modelli/foto partenza ({fid})...")
+    files = cloud_sync.list_files_in_public_folder(fid)
+    if not files:
+        print(f"⚠️ Nessun file trovato nella cartella Drive: {models_url}")
+    
+    for fid_sub, fname in files:
+        if fname.lower().endswith(('.jpg', '.jpeg', '.png', '.webp')):
+            local_path = f"{local_dir}/{fname}"
+            if not os.path.exists(local_path):
+                print(f"📥 Scaricamento foto partenza: {fname}...")
+                url = f"https://drive.google.com/uc?id={fid_sub}"
+                try:
+                    gdown.download(url, output=local_path, quiet=True)
+                except Exception as dl_err:
+                    print(f"⚠️ Errore download {fname}: {dl_err}")
+                    
+    # 2. Run generate_higgsfield_photos.py via subprocess
+    cmd = [
+        sys.executable, "-m", "src.generate_higgsfield_photos",
+        "--dir", local_dir,
+        "--prompt", prompt
+    ]
+    if dry_run:
+        cmd.append("--dry-run")
+        
+    print(f"🚀 Esecuzione script generatore: {' '.join(cmd)}")
+    res = subprocess.run(cmd, capture_output=True, text=True)
+    print(res.stdout)
+    if res.stderr:
+        print(f"⚠️ Stderr: {res.stderr}")
+    print("✅ Task Higgsfield completato.")
+
+if __name__ == "__main__":
+    parser = argparse.ArgumentParser()
+    parser.add_argument("--task", choices=["sync", "generate", "higgsfield"], required=True)
+    parser.add_argument("--artist_id", default="all")
+    parser.add_argument("--mode", choices=["photo", "video", "both"], default="both")
+    parser.add_argument("--prompt", default="A beautiful editorial studio portrait, highly detailed, cinematic studio lighting, professional photography, 8k resolution, crisp details")
+    parser.add_argument("--dry_run", action="store_true")
+    args = parser.parse_args()
+
+    lock = create_lock(args.task, f"{args.artist_id}_{args.mode}")
+    try:
+        if args.task == "sync":
+            run_sync()
+        elif args.task == "generate":
+            run_generate(args.artist_id, mode=args.mode)
+        elif args.task == "higgsfield":
+            run_higgsfield(args.artist_id, args.prompt, dry_run=args.dry_run)
+    except Exception as e:
+        import traceback
+        error_msg = f"❌ Errore critico nel task {args.task} ({args.artist_id}): {str(e)}\n{traceback.format_exc()}"
+        print(error_msg)
+        os.makedirs("data/tasks", exist_ok=True)
+        with open("data/tasks/error.log", "a") as f:
+            f.write(f"\n--- {time.ctime()} ---\n{error_msg}\n")
+    finally:
+        remove_lock(lock)
+
@@ -0,0 +1,153 @@
+import os
+import requests
+import json
+from datetime import datetime
+from dotenv import load_dotenv
+
+load_dotenv()
+
+class BufferPublisher:
+    def __init__(self, token=None):
+        self.token = token or os.getenv("BUFFER_ACCESS_TOKEN")
+        self.graphql_url = "https://api.buffer.com/graphql"
+        self.log_file = "data/buffer_logs.txt"
+
+    def log(self, message):
+        timestamp = datetime.now().strftime("%Y-%m-%d %H:%M:%S")
+        os.makedirs("data", exist_ok=True)
+        with open(self.log_file, "a") as f:
+            f.write(f"[{timestamp}] {message}\n")
+
+    def get_profiles(self):
+        query = """
+        query {
+          profiles {
+            id
+            type
+            service
+          }
+        }
+        """
+        headers = {
+            "Authorization": f"Bearer {self.token}",
+            "Content-Type": "application/json"
+        }
+        try:
+            response = requests.post(self.graphql_url, headers=headers, json={'query': query})
+            if response.status_code == 200:
+                data = response.json()
+                profiles = data.get('data', {}).get('profiles', [])
+                # Filtriamo per i servizi che ci interessano (Instagram e TikTok)
+                return [p['id'] for p in profiles if p['service'] in ['instagram', 'tiktok']]
+            else:
+                self.log(f"ERRORE recupero profili: {response.text}")
+                return []
+        except Exception as e:
+            self.log(f"ECCEZIONE recupero profili: {e}")
+            return []
+
+    def publish(self, profile_ids, text, image_path, video_url=None, image_paths=None, artist_token=None):
+        # Garantisce l'uso del token specifico o il fallback al globale per ogni singola chiamata
+        self.token = artist_token or os.getenv("BUFFER_ACCESS_TOKEN")
+        
+        # Se non ci sono profile_ids, li cerchiamo automaticamente
+        if not profile_ids:
+            self.log("Nessun Profile ID fornito, avvio auto-discovery...")
+            profile_ids = self.get_profiles()
+            if not profile_ids:
+                return {"success": False, "error": "Nessun canale social trovato per questo account Buffer."}
+            self.log(f"Canali trovati automaticamente: {len(profile_ids)}")
+
+        return self.publish_via_graphql(profile_ids, text, image_path, video_url, image_paths)
+
+    def publish_via_graphql(self, profile_ids, text, image_path, video_url=None, image_paths=None):
+        from database import Database, Media
+        db = Database()
+        session = db.Session()
+        
+        # Prepariamo gli asset: se c'è un video, ha la precedenza
+        assets_payload = None
+        if video_url:
+            assets_payload = {"videos": [{"url": video_url}]} 
+        elif image_paths:
+            # Gestione Carousel (lista di percorsi locali)
+            if isinstance(image_paths, str):
+                try: image_paths = json.loads(image_paths)
+                except: image_paths = [image_paths]
+            
+            image_urls = []
+            for path in image_paths:
+                media = session.query(Media).filter_by(file_path=path).first()
+                if media and media.source_url:
+                    image_urls.append({"url": f"https://lh3.googleusercontent.com/d/{media.source_url}=s1080"})
+            
+            if image_urls:
+                assets_payload = {"images": image_urls}
+        else:
+            media = session.query(Media).filter_by(file_path=image_path).first()
+            if media and media.source_url:
+                direct_link = f"https://lh3.googleusercontent.com/d/{media.source_url}=s1080"
+                assets_payload = {"images": [{"url": direct_link}]}
+        
+        results = []
+        for channel_id in profile_ids:
+            mutation = """
+            mutation ($input: CreatePostInput!) {
+              createPost(input: $input) {
+                ... on PostActionSuccess { post { id } }
+                ... on MutationError { message }
+              }
+            }
+            """
+            
+            # Input base per la PUBBLICAZIONE DIRETTA
+            post_input = {
+                "channelId": channel_id,
+                "text": text,
+                "schedulingType": "automatic", # Questo campo è obbligatorio per l'API anche se condividiamo subito
+                "mode": "shareNow" # Ignora la coda e pubblica ISTANTANEAMENTE
+            }
+            
+            if assets_payload:
+                post_input["assets"] = assets_payload
+            
+            metadata = {}
+            if channel_id == "69f9ea855c4c051afa117baa": # Instagram
+                metadata["instagram"] = {
+                    "type": "post",
+                    "shouldShareToFeed": True # CAMPO MANCANTE RIPRISTINATO
+                }
+            elif channel_id == "69f9e7e35c4c051afa116a9e": # TikTok
+                metadata["tiktok"] = {"title": text[:50]}
+            
+            if metadata:
+                post_input["metadata"] = metadata
+
+            headers = {
+                "Authorization": f"Bearer {self.token}",
+                "Content-Type": "application/json"
+            }
+
+            self.log(f"INVIO POST a Canale: {channel_id}")
+            response = requests.post(self.graphql_url, headers=headers, json={
+                'query': mutation, 
+                'variables': {"input": post_input}
+            })
+            
+            if response.status_code == 200:
+                data = response.json()
+                res = data.get('data', {}).get('createPost', {})
+                self.log(f"RISPOSTA BUFFER: {json.dumps(res)}")
+                
+                if 'message' in res:
+                    results.append({"success": False, "error": res['message']})
+                else:
+                    results.append({"success": True, "id": res.get('post', {}).get('id')})
+            else:
+                self.log(f"ERRORE HTTP {response.status_code}: {response.text}")
+                results.append({"success": False, "error": response.text})
+        
+        session.close()
+        if any(r['success'] for r in results):
+            return {"success": True, "details": results}
+        return {"success": False, "error": results[0].get('error', 'Errore')}
@@ -0,0 +1,120 @@
+import os
+import json
+import requests
+import re
+import subprocess
+from database import Database
+
+db = Database()
+
+def get_drive_id(url):
+    match = re.search(r'/file/d/([a-zA-Z0-9_-]+)', url)
+    if match: return match.group(1)
+    match = re.search(r'id=([a-zA-Z0-9_-]+)', url)
+    if match: return match.group(1)
+    match = re.search(r'folders/([a-zA-Z0-9_-]+)', url)
+    if match: return match.group(1)
+    return None
+
+def list_files_in_public_folder(folder_id):
+    url = f"https://drive.google.com/embeddedfolderview?id={folder_id}"
+    files_found = []
+    try:
+        response = requests.get(url, timeout=15)
+        if response.status_code == 200:
+            entries = re.findall(r'id="entry-([a-zA-Z0-9_-]+)".*?class="flip-entry-title">([^<]+)</div>', response.text, re.DOTALL)
+            for fid, fname in entries:
+                if fid not in [f[0] for f in files_found]:
+                    files_found.append((fid, fname))
+    except Exception as e:
+        print(f"Errore Drive: {e}")
+    return files_found
+
+def generate_video_thumbnail(video_fid, output_path):
+    """Genera una miniatura da un video di Google Drive usando FFmpeg"""
+    if os.path.exists(output_path): 
+        print(f"⏩ Miniatura già esistente: {output_path}")
+        return True
+    
+    print(f"🎬 Generazione miniatura per FID: {video_fid}...")
+    # URL di download diretto
+    url = f"https://drive.google.com/uc?id={video_fid}&export=download"
+    
+    # Comando FFmpeg per estrarre un frame a 1 secondo
+    # Usiamo parametri per velocizzare l'apertura dello stream
+    cmd = [
+        'ffmpeg', '-ss', '00:00:01', '-i', url, 
+        '-frames:v', '1', '-q:v', '2', 
+        '-vf', 'scale=320:-1', 
+        output_path, '-y'
+    ]
+    
+    try:
+        # Timeout per evitare blocchi infiniti su file troppo grandi o link protetti
+        res = subprocess.run(cmd, capture_output=True, timeout=25)
+        if res.returncode != 0:
+            print(f"⚠️ FFmpeg ha restituito un errore (probabile file protetto o troppo grande): {video_fid}")
+            return False
+        return os.path.exists(output_path)
+    except subprocess.TimeoutExpired:
+        print(f"⏳ Timeout FFmpeg per {video_fid} - saltato.")
+        return False
+    except Exception as e:
+        print(f"❌ Errore inaspettato thumbnail: {e}")
+        return False
+
+def sync_from_drive():
+    if not os.path.exists('config.json'): return
+    
+    # Cartella cache unica
+    cache_dir = "data/cache"
+    os.makedirs(cache_dir, exist_ok=True)
+
+    with open('config.json', 'r') as f:
+        config = json.load(f)
+
+    for artist in config['artists']:
+        artist_id = artist['id']
+        
+        # Audio Discovery
+        audio_url = artist.get('drive_audio_url')
+        if audio_url:
+            fid = get_drive_id(audio_url)
+            if "/file/d/" in audio_url:
+                db.add_media(f"{cache_dir}/audio_{fid}", 'audio', artist_id, source_url=fid)
+            elif fid:
+                for fid_sub, fname in list_files_in_public_folder(fid):
+                    db.add_media(f"{cache_dir}/{fname}", 'audio', artist_id, source_url=fid_sub)
+
+        # Images Discovery
+        images_url = artist.get('drive_images_url')
+        if images_url:
+            fid = get_drive_id(images_url)
+            if "/file/d/" in images_url:
+                db.add_media(f"{cache_dir}/image_{fid}", 'image', artist_id, source_url=fid)
+            elif fid:
+                for fid_sub, fname in list_files_in_public_folder(fid):
+                    db.add_media(f"{cache_dir}/{fname}", 'image', artist_id, source_url=fid_sub)
+
+        # Videos Discovery
+        videos_url = artist.get('drive_videos_url')
+        if videos_url:
+            fid = get_drive_id(videos_url)
+            if not fid:
+                print(f"⚠️ URL Video non valido per {artist_id}: {videos_url}")
+            elif "/file/d/" in videos_url:
+                # Nessuna miniatura locale per risparmiare spazio
+                db.add_media(f"{cache_dir}/video_{fid}", 'video', artist_id, source_url=fid)
+            else:
+                v_files = list_files_in_public_folder(fid)
+                if not v_files:
+                    print(f"ℹ️ Nessun video trovato nella cartella di {artist_id}")
+                for fid_sub, fname in v_files:
+                    # Registriamo solo il metadata, la miniatura sarà caricata 'on-the-fly' dalla UI
+                    db.add_media(f"{cache_dir}/{fname}", 'video', artist_id, source_url=fid_sub)
+        else:
+            print(f"ℹ️ Nessuna cartella video configurata per {artist_id}")
+
+if __name__ == "__main__":
+    sync_from_drive()
+    print("Sincro Cache completata.")
@@ -0,0 +1,239 @@
+import datetime
+from sqlalchemy import create_engine, Column, Integer, String, Boolean, DateTime, Text
+from sqlalchemy.ext.declarative import declarative_base
+from sqlalchemy.orm import sessionmaker
+
+Base = declarative_base()
+
+class Media(Base):
+    __tablename__ = 'media'
+    
+    id = Column(Integer, primary_key=True)
+    file_path = Column(String, unique=True, nullable=False)
+    file_type = Column(String)  # 'image' or 'audio'
+    artist_id = Column(String)
+    is_used = Column(Boolean, default=False)
+    description = Column(Text)
+    source_url = Column(String)
+    thumbnail_path = Column(String) # NUOVO
+    created_at = Column(DateTime, default=datetime.datetime.utcnow)
+    last_used_at = Column(DateTime)
+
+class Draft(Base):
+    __tablename__ = 'drafts'
+    id = Column(Integer, primary_key=True)
+    artist_id = Column(String)
+    title = Column(String)
+    caption = Column(Text)
+    hashtags = Column(Text)
+    image_path = Column(String)
+    image_paths = Column(Text) # JSON list of local paths
+    image_url = Column(String) # NUOVA COLONNA PER ANTEPRIMA DIRETTA
+    video_url = Column(String)
+    video_path = Column(String) # NUOVO
+    audio_analysis = Column(Text)
+    focus_points = Column(Text) # NUOVO: Coordinate JSON per la regia AI
+    status = Column(String, default='pending') # pending, approved, published
+    created_at = Column(DateTime, default=datetime.datetime.utcnow)
+
+class Database:
+    def __init__(self, db_url="sqlite:///data/redazione.db"):
+        self.engine = create_engine(db_url)
+        Base.metadata.create_all(self.engine)
+        self.Session = sessionmaker(bind=self.engine)
+
+    def add_media(self, file_path, file_type, artist_id, description=None, source_url=None, thumbnail_path=None):
+        session = self.Session()
+        try:
+            media = session.query(Media).filter_by(file_path=file_path).first()
+            if not media:
+                media = Media(
+                    file_path=file_path, 
+                    file_type=file_type, 
+                    artist_id=artist_id,
+                    description=description,
+                    source_url=source_url,
+                    thumbnail_path=thumbnail_path
+                )
+                session.add(media)
+                session.commit()
+            elif thumbnail_path and not media.thumbnail_path:
+                media.thumbnail_path = thumbnail_path
+                session.commit()
+            return media
+        finally:
+            session.close()
+
+    def get_unused_image(self, artist_id):
+        from sqlalchemy.sql.expression import func
+        session = self.Session()
+        try:
+            return session.query(Media).filter_by(
+                artist_id=artist_id, 
+                file_type='image', 
+                is_used=False
+            ).order_by(func.random()).first()
+        finally:
+            session.close()
+
+    def get_media_by_artist(self, artist_id, file_type):
+        session = self.Session()
+        try:
+            return session.query(Media).filter_by(
+                artist_id=artist_id, 
+                file_type=file_type
+            ).first()
+        finally:
+            session.close()
+
+    def get_audio_for_artist(self, artist_id):
+        return self.get_media_by_artist(artist_id, 'audio')
+
+    def mark_as_used(self, media_id):
+        session = self.Session()
+        try:
+            media = session.query(Media).get(media_id)
+            if media:
+                media.is_used = True
+                media.last_used_at = datetime.datetime.utcnow()
+                session.commit()
+        finally:
+            session.close()
+
+    def get_audio_for_artist(self, artist_id):
+        # We assume one main audio track per session or the latest added
+        session = self.Session()
+        try:
+            return session.query(Media).filter_by(
+                artist_id=artist_id, 
+                file_type='audio'
+            ).order_by(Media.created_at.desc()).first()
+        finally:
+            session.close()
+
+    def save_draft(self, artist_id, title, caption, hashtags, image_path, audio_analysis, image_url=None, video_url=None, image_paths=None, video_path=None, focus_points=None):
+        session = self.Session()
+        try:
+            draft = Draft(
+                artist_id=artist_id,
+                title=title,
+                caption=caption,
+                hashtags=hashtags,
+                image_path=image_path,
+                image_paths=image_paths,
+                image_url=image_url,
+                video_url=video_url,
+                video_path=video_path,
+                audio_analysis=audio_analysis,
+                focus_points=focus_points # SALVATAGGIO
+            )
+            session.add(draft)
+            session.commit()
+            return draft.id
+        finally:
+            session.close()
+
+    def get_recent_media(self, artist_id, limit=10):
+        session = self.Session()
+        try:
+            return session.query(Media).filter(
+                Media.artist_id == artist_id,
+                Media.file_type.in_(['image', 'video'])
+            ).order_by(Media.created_at.desc()).limit(limit).all()
+        finally:
+            session.close()
+
+    def get_mixed_assets(self, artist_id, total=12):
+        from sqlalchemy.sql.expression import func
+        session = self.Session()
+        try:
+            # 1. Seleziona una copertina (preferibilmente immagine mai usata)
+            cover = session.query(Media).filter_by(
+                artist_id=artist_id, 
+                file_type='image', 
+                is_used=False
+            ).order_by(func.random()).first()
+            
+            if not cover:
+                cover = session.query(Media).filter_by(
+                    artist_id=artist_id, 
+                    file_type='image'
+                ).order_by(func.random()).first()
+            
+            if not cover:
+                # Se proprio non ci sono immagini, prendi un video come cover
+                cover = session.query(Media).filter_by(
+                    artist_id=artist_id,
+                    file_type='video'
+                ).order_by(func.random()).first()
+
+            if not cover:
+                return []
+
+            # 2. Seleziona un mix di altre immagini e video
+            others = session.query(Media).filter(
+                Media.artist_id == artist_id, 
+                Media.file_type.in_(['image', 'video']),
+                Media.id != cover.id
+            ).order_by(func.random()).limit(total - 1).all()
+            
+            return [cover] + others
+        finally:
+            session.close()
+
+    def get_pending_drafts(self, artist_id=None):
+        session = self.Session()
+        try:
+            query = session.query(Draft).filter_by(status='pending')
+            if artist_id:
+                query = query.filter_by(artist_id=artist_id)
+            return query.all()
+        finally:
+            session.close()
+    def get_media_by_path(self, file_path):
+        session = self.Session()
+        try:
+            return session.query(Media).filter_by(file_path=file_path).first()
+        finally:
+            session.close()
+
+    def delete_draft(self, draft_id):
+        session = self.Session()
+        try:
+            draft = session.query(Draft).filter_by(id=draft_id).first()
+            if draft:
+                session.delete(draft)
+                session.commit()
+        finally:
+            session.close()
+
+    def update_draft(self, draft_id, title=None, caption=None, hashtags=None, image_path=None, status=None, audio_analysis=None, video_url=None, video_path=None, image_paths=None, focus_points=None):
+        session = self.Session()
+        try:
+            draft = session.query(Draft).get(draft_id)
+            if draft:
+                if title is not None: draft.title = title
+                if caption is not None: draft.caption = caption
+                if hashtags is not None: draft.hashtags = hashtags
+                if image_path is not None: draft.image_path = image_path
+                if status is not None: draft.status = status
+                if audio_analysis is not None: draft.audio_analysis = audio_analysis
+                if video_url is not None: draft.video_url = video_url
+                if video_path is not None: draft.video_path = video_path
+                if image_paths is not None: draft.image_paths = image_paths
+                if focus_points is not None: draft.focus_points = focus_points # AGGIORNAMENTO
+                session.commit()
+                return True
+            return False
+        finally:
+            session.close()
+
+    def mark_as_published(self, draft_id):
+        session = self.Session()
+        try:
+            draft = session.query(Draft).filter_by(id=draft_id).first()
+            if draft:
+                draft.status = 'published'
+                session.commit()
+        finally:
+            session.close()
@@ -0,0 +1,86 @@
+import os
+import json
+import sys
+from dotenv import load_dotenv
+
+# Aggiungiamo src al path
+sys.path.append(os.path.join(os.getcwd(), 'src'))
+
+from database import Database
+from agents import SocialAgents
+from buffer_publisher import BufferPublisher
+
+load_dotenv()
+
+def force_test():
+    db = Database()
+    agents = SocialAgents()
+    publisher = BufferPublisher()
+    
+    # 1. Carichiamo i dati di Veronica
+    with open('config.json', 'r') as f:
+        config = json.load(f)
+        veronica = next(a for a in config['artists'] if "Veronica" in a['name'])
+    
+    print(f"--- FASE 1: GENERAZIONE PER {veronica['name']} ---")
+    result = agents.run_for_artist(veronica)
+    
+    if 'error' in result:
+        print(f"Errore Generazione: {result['error']}")
+        return
+
+    print(f"Post Generato: {result['title']}")
+    
+    # 2. Salvataggio bozza locale
+    draft_id = db.save_draft(
+        artist_id=veronica['id'],
+        title=result['title'],
+        caption=result['caption'],
+        hashtags=result['hashtags'],
+        image_path=result['image_path'],
+        audio_analysis=result['audio_analysis']
+    )
+    
+    # 3. PUBBLICAZIONE SU BUFFER
+    print(f"\n--- FASE 2: INVIO A BUFFER ---")
+    profile_ids = ["69f9e7e35c4c051afa116a9e", "69f9ea855c4c051afa117baa"]
+    full_text = f"*{result['title']}*\n\n{result['caption']}\n\n{result['hashtags']}"
+    
+    pub_res = publisher.publish(profile_ids, full_text, result['image_path'])
+    print(f"Risultato Pubblicazione: {pub_res}")
+    
+    # 4. VERIFICA REALE SU BUFFER
+    print(f"\n--- FASE 3: VERIFICA BOZZE SU BUFFER ---")
+    token = os.getenv("BUFFER_ACCESS_TOKEN")
+    url = "https://api.buffer.com/graphql"
+    headers = {"Authorization": f"Bearer {token}", "Content-Type": "application/json"}
+    
+    # Chiediamo i post in stato 'draft' per Instagram
+    query = """
+    query {
+      channels(input: {ids: ["69f9ea855c4c051afa117baa"]}) {
+        name
+        posts(input: {state: draft}) {
+          totalCount
+          nodes {
+            id
+            text
+          }
+        }
+      }
+    }
+    """
+    
+    resp = requests.post(url, headers=headers, json={'query': query})
+    if resp.status_code == 200:
+        data = resp.json()
+        drafts = data['data']['channels'][0]['posts']
+        print(f"Bozze totali su Instagram: {drafts['totalCount']}")
+        for node in drafts['nodes']:
+            print(f"- Bozza trovata ID: {node['id']} | Testo: {node['text'][:50]}...")
+    else:
+        print(f"Errore verifica: {resp.text}")
+
+if __name__ == "__main__":
+    import requests # Lo importiamo qui per sicurezza
+    force_test()
@@ -0,0 +1,401 @@
+import os
+import sys
+import json
+import argparse
+import subprocess
+import re
+import requests
+from pathlib import Path
+from dotenv import load_dotenv
+
+# Add the directory containing this script to sys.path to allow imports from src
+current_dir = Path(__file__).resolve().parent
+if str(current_dir) not in sys.path:
+    sys.path.insert(0, str(current_dir))
+
+# Load environment variables
+load_dotenv()
+
+def normalize_string(s):
+    """Normalize a string to lowercase and remove non-alphanumeric characters for fuzzy matching."""
+    if not s:
+        return ""
+    return re.sub(r'[^a-z0-9]', '', s.lower())
+
+def load_artists_config(config_path="config.json"):
+    """Load artists configuration from config.json."""
+    try:
+        with open(config_path, 'r', encoding='utf-8') as f:
+            config = json.load(f)
+            return config.get("artists", [])
+    except Exception as e:
+        print(f"❌ Error loading config.json: {e}", file=sys.stderr)
+        return []
+
+def match_artist(file_path, artists):
+    """
+    Fuzzy match a file path/filename to an artist in config.json.
+    Returns the matched artist dict, or None if no match found.
+    """
+    path_obj = Path(file_path)
+    # Check parent directory name and filename
+    search_space = f"{path_obj.parent.name} {path_obj.name}"
+    normalized_search = normalize_string(search_space)
+    
+    # Try exact or substring matches
+    for artist in artists:
+        artist_id = artist.get("id", "")
+        artist_name = artist.get("name", "")
+        social_tag = artist.get("social_tag", "")
+        
+        norm_id = normalize_string(artist_id)
+        norm_name = normalize_string(artist_name)
+        norm_tag = normalize_string(social_tag)
+        
+        # Check if artist name/id is in the search space
+        if (norm_id and norm_id in normalized_search) or \
+           (norm_name and norm_name in normalized_search) or \
+           (norm_tag and norm_tag in normalized_search):
+            return artist
+            
+    # Fuzzy match check (e.g. if singer first name matches)
+    for artist in artists:
+        artist_name = artist.get("name", "")
+        first_name = artist_name.split()[0] if artist_name else ""
+        norm_first = normalize_string(first_name)
+        if norm_first and norm_first in normalized_search:
+            return artist
+            
+    return None
+
+def run_higgsfield_cli_generate(prompt, soul_id, image_path, output_path):
+    """
+    Run Higgsfield generation using the Higgsfield CLI via subprocess.
+    """
+    print(f"🎬 [CLI] Generating AI photo for Soul ID: {soul_id}...")
+    try:
+        # First, upload the reference image if provided
+        upload_cmd = ["higgsfield", "upload", str(image_path)]
+        print(f"   Uploading reference image: {' '.join(upload_cmd)}")
+        upload_res = subprocess.run(upload_cmd, capture_output=True, text=True, check=True)
+        
+        # Extract UUID from upload response
+        # Standard UUID regex
+        uuids = re.findall(r'[0-9a-f]{8}-[0-9a-f]{4}-[0-9a-f]{4}-[0-9a-f]{4}-[0-9a-f]{12}', upload_res.stdout, re.IGNORECASE)
+        if not uuids:
+            print(f"   ❌ No UUID found in upload output: {upload_res.stdout}")
+            return False
+            
+        uuid = uuids[0]
+        print(f"   Uploaded successfully. Reference UUID: {uuid}")
+        
+        # Create generation job
+        gen_cmd = [
+            "higgsfield", "generate", "create", "soul_v2",
+            "--prompt", prompt,
+            "--soul-id", soul_id,
+            "--image-id", uuid,
+            "--wait"
+        ]
+        print(f"   Running generation: {' '.join(gen_cmd)}")
+        gen_res = subprocess.run(gen_cmd, capture_output=True, text=True, check=True)
+        
+        # Find output URL or download path
+        # If the CLI downloads automatically or provides a URL, we capture it
+        urls = re.findall(r'https?://[^\s]+', gen_res.stdout)
+        if urls:
+            url = urls[0]
+            print(f"   Generation completed. Output URL: {url}")
+            # Download file
+            resp = requests.get(url, timeout=30)
+            if resp.status_code == 200:
+                with open(output_path, "wb") as f:
+                    f.write(resp.content)
+                print(f"   ✅ Saved generated photo to: {output_path}")
+                return True
+        else:
+            print(f"   ❌ Could not extract output URL from generation log: {gen_res.stdout}")
+            return False
+            
+    except subprocess.CalledProcessError as e:
+        print(f"   ❌ CLI command failed: {e.stderr}")
+        return False
+    except Exception as e:
+        print(f"   ❌ CLI generation error: {e}")
+        return False
+
+def run_higgsfield_api_generate(prompt, model_to_use, soul_id, image_path, output_path, api_key, api_secret=None):
+    """
+    Run Higgsfield generation using direct REST API requests via the official SDK.
+    """
+    import higgsfield_client
+    
+    if soul_id:
+        print(f"🔌 [API] Generating AI photo using model '{model_to_use}' for Soul ID: {soul_id}...")
+    else:
+        print(f"🔌 [API] Generating AI photo using fallback model '{model_to_use}'...")
+        
+    try:
+        # Set environment variables for the SDK
+        os.environ["HF_API_KEY"] = api_key
+        if api_secret:
+            os.environ["HF_API_SECRET"] = api_secret
+            
+        print("   Initializing Higgsfield SDK Client...")
+        client = higgsfield_client.SyncClient()
+        
+        # 1. Upload starting photo
+        print(f"   Uploading starting photo: {image_path}...")
+        public_url = client.upload_file(image_path)
+        print(f"   Uploaded successfully. URL: {public_url}")
+        
+        # 2. Submit Generation Job
+        if model_to_use == "flux-2":
+            arguments = {
+                "prompt": prompt,
+                "aspect_ratio": "1:1",
+                "resolution": "2k",
+                "input_images": [public_url]
+            }
+        else:
+            # Fallback/Default for soul or others if soul_id is present
+            # We use the 'soul' endpoint on the platform API
+            arguments = {
+                "prompt": prompt,
+                "aspect_ratio": "1:1",
+                "style_id": "realistic", # default style_id required by the API
+                "character_id": soul_id,
+                "input_images": [public_url]
+            }
+            
+        print(f"   Submitting job to endpoint '{model_to_use}'...")
+        controller = client.submit(
+            application=model_to_use,
+            arguments=arguments
+        )
+        request_id = controller.request_id
+        print(f"   Job submitted successfully. Request ID: {request_id}. Polling for completion...")
+        
+        # 3. Poll for completion
+        for status in controller.poll_request_status(delay=3.0):
+            # Print status to stdout for log visibility
+            print(f"   Job status: {status}")
+            
+        # 4. Download result
+        print("   Retrieving completed job data...")
+        result = controller.get()
+        
+        # In flux-2/soul, completed output is stored in 'images' list or 'outputs' list
+        images = result.get("images") or result.get("outputs") or []
+        output_url = result.get("output_url") or (images[0].get("url") if images else None)
+        
+        if not output_url:
+            print(f"   ❌ Job completed but no output URL found in response: {result}")
+            return False
+            
+        print(f"   ✅ Generation complete! Downloading from: {output_url}")
+        resp = requests.get(output_url, timeout=30)
+        if resp.status_code == 200:
+            with open(output_path, "wb") as f:
+                f.write(resp.content)
+            print(f"   ✅ Saved generated photo to: {output_path}")
+            return True
+        else:
+            print(f"   ❌ Download failed with status: {resp.status_code}")
+            return False
+            
+    except Exception as e:
+        print(f"   ❌ REST API generation error: {e}")
+        import traceback
+        traceback.print_exc()
+        return False
+def deduct_credits(amount=1):
+    try:
+        cache_dir = Path("data/cache")
+        cache_dir.mkdir(parents=True, exist_ok=True)
+        credits_file = cache_dir / "credits.json"
+        
+        # Load existing or create
+        if credits_file.exists():
+            with open(credits_file, 'r') as f:
+                data = json.load(f)
+        else:
+            data = {"total_monthly": 1000, "remaining": 1000, "generated_this_month": 0}
+            
+        data["remaining"] = max(0, data.get("remaining", 1000) - amount)
+        data["generated_this_month"] = data.get("generated_this_month", 0) + amount
+        
+        with open(credits_file, 'w') as f:
+            json.dump(data, f, indent=2)
+            
+        print(f"   📊 Local credits updated: {data['remaining']}/{data['total_monthly']} remaining ({data['generated_this_month']} generated)")
+    except Exception as e:
+        print(f"   ⚠️ Could not update credits cache: {e}")
+
+def main():
+    parser = argparse.ArgumentParser(description="Batch generate AI photos using Higgsfield AI with Soul ID character consistency.")
+    parser.add_argument("--dir", required=True, help="Directory containing starting photos.")
+    parser.add_argument("--soul-ids", help="JSON string mapping artist names/IDs to Higgsfield Soul IDs.")
+    parser.add_argument("--output-dir", default="data/images/generated_photos", help="Output directory.")
+    parser.add_argument("--prompt", default="A beautiful editorial studio portrait, highly detailed, cinematic studio lighting, professional photography, 8k resolution, crisp details", help="Prompt override for visual style.")
+    parser.add_argument("--dry-run", action="store_true", help="Simulate the execution and check mappings without making API/CLI requests.")
+    
+    args = parser.parse_args()
+    
+    input_dir = Path(args.dir)
+    if not input_dir.is_dir():
+        print(f"❌ Error: {input_dir} is not a valid directory.", file=sys.stderr)
+        sys.exit(1)
+        
+    output_dir = Path(args.output_dir)
+    output_dir.mkdir(parents=True, exist_ok=True)
+    
+    # Load artists from config
+    artists = load_artists_config()
+    if not artists:
+        print("⚠️ Warning: No artists found in config.json.", file=sys.stderr)
+        
+    # Load custom soul IDs mapping
+    custom_soul_ids = {}
+    if args.soul_ids:
+        try:
+            custom_soul_ids = json.loads(args.soul_ids)
+            print(f"ℹ️ Loaded custom Soul ID overrides: {custom_soul_ids}")
+        except Exception as e:
+            print(f"❌ Error parsing --soul-ids JSON: {e}", file=sys.stderr)
+            sys.exit(1)
+            
+    # Compile final artist -> soul_id mapping
+    artist_soul_map = {}
+    for artist in artists:
+        name = artist.get("name", "")
+        artist_id = artist.get("id", "")
+        # Priority: 1. CLI Override by Name, 2. CLI Override by ID, 3. config.json soul_id
+        soul_id = custom_soul_ids.get(name) or custom_soul_ids.get(artist_id) or artist.get("soul_id")
+        if soul_id:
+            artist_soul_map[name] = soul_id
+            
+    print(f"ℹ️ Active Singer Soul ID mapping: {artist_soul_map}")
+    
+    # Scan input directory for images
+    supported_extensions = ['.png', '.jpg', '.jpeg', '.webp']
+    image_files = [f for f in input_dir.rglob("*") if f.suffix.lower() in supported_extensions]
+    
+    if not image_files:
+        print(f"⚠️ No starting photos found in {input_dir}.")
+        sys.exit(0)
+        
+    print(f"🔍 Found {len(image_files)} starting photos. Beginning processing...")
+    
+    # Retrieve credentials
+    api_key = os.getenv("HIGGSFIELD_API_KEY")
+    api_secret = os.getenv("HIGGSFIELD_SECRET") # optional
+    
+    # Determine generation method
+    use_api = bool(api_key and api_key != "your_higgsfield_api_key_here")
+    
+    # If explicitly in dry-run mode
+    if args.dry_run:
+        print("\n🧪 [DRY-RUN] Simulating Higgsfield AI generation pipeline. No real API calls will be made.")
+    elif not use_api:
+        print("ℹ️ HIGGSFIELD_API_KEY is not set or is placeholder. Using CLI-based generation.")
+        # Check if higgsfield CLI is installed
+        try:
+            res = subprocess.run(["which", "higgsfield"], capture_output=True, text=True)
+            if res.returncode != 0:
+                print("⚠️ Warning: Higgsfield CLI ('higgsfield') not found on system path.")
+                print("   Automatically falling back to DRY-RUN simulation mode.")
+                args.dry_run = True
+        except Exception as e:
+            print("⚠️ Warning: Error checking for Higgsfield CLI.")
+            print("   Automatically falling back to DRY-RUN simulation mode.")
+            args.dry_run = True
+            
+    success_count = 0
+    failure_count = 0
+    
+    for img_path in image_files:
+        print(f"\n📸 Processing photo: {img_path.name}")
+        
+        # 1. Match artist
+        matched_artist = match_artist(img_path, artists)
+        if not matched_artist:
+            print(f"   ⚠️ Skipping: Could not match image path/name to any known artist in config.json.")
+            failure_count += 1
+            continue
+            
+        artist_name = matched_artist.get("name")
+        print(f"   Matched Artist: {artist_name}")
+        
+        # 2. Retrieve Soul ID & Check if valid
+        soul_id = artist_soul_map.get(artist_name)
+        is_placeholder = bool(soul_id and (soul_id.startswith("soul_") or soul_id == "your_soul_id_here"))
+        
+        if not soul_id or is_placeholder:
+            print(f"   ℹ️ No valid Soul ID configured for artist '{artist_name}'. Falling back to flux-2 model!")
+            model_to_use = "flux-2"
+            soul_id = None
+        else:
+            model_to_use = "soul"
+            
+        # 3. Generate output file name
+        out_filename = f"gen_{artist_name.replace(' ', '_')}_{img_path.stem}.png"
+        out_path = output_dir / out_filename
+        
+        # 4. Generate AI Photo
+        if args.dry_run:
+            print(f"   🧪 [DRY-RUN] Would generate AI photo:")
+            print(f"     - Starting photo: {img_path.absolute()}")
+            print(f"     - Model to use: {model_to_use}")
+            print(f"     - Soul ID reference: {soul_id}")
+            print(f"     - Style prompt: {args.prompt}")
+            print(f"     - Output destination: {out_path.absolute()}")
+            success = True
+            # Simulate DB save in dry-run
+            print(f"   🧪 [DRY-RUN] Would register in webapp database as Draft for {artist_name}")
+        elif use_api:
+            success = run_higgsfield_api_generate(args.prompt, model_to_use, soul_id, img_path, out_path, api_key, api_secret)
+        else:
+            # CLI fallback doesn't support model argument in our signature, so use original
+            success = run_higgsfield_cli_generate(args.prompt, soul_id, img_path, out_path)
+            
+        if success:
+            success_count += 1
+            if not args.dry_run:
+                try:
+                    from database import Database
+                    db_instance = Database()
+                    
+                    if model_to_use == "flux-2":
+                        analysis_text = "Immagine generata tramite Higgsfield AI utilizzando il modello Flux-2 con riferimento visivo."
+                        hashtags_text = "#higgsfield #flux #consistentcharacter"
+                    else:
+                        analysis_text = f"Immagine generata tramite Higgsfield AI utilizzando il Soul ID: {soul_id}."
+                        hashtags_text = "#higgsfield #soul #consistentcharacter"
+                        
+                    draft_id = db_instance.save_draft(
+                        artist_id=matched_artist.get("id"),
+                        title=f"AI Photo - {artist_name}",
+                        caption=f"Ecco un nuovo post generato con l'intelligenza artificiale per {artist_name}! #music #ai",
+                        hashtags=hashtags_text,
+                        image_path=str(out_path),
+                        audio_analysis=analysis_text
+                    )
+                    print(f"   💾 Registered in webapp database as Draft ID: {draft_id}")
+                    deduct_credits(1)
+                except Exception as db_err:
+                    print(f"   ⚠️ Could not register in webapp database: {db_err}")
+        else:
+            failure_count += 1
+            
+    print(f"\n📊 Processing complete!")
+    print(f"   Successfully processed/simulated: {success_count} photos")
+    print(f"   Failed or skipped: {failure_count} photos")
+    if args.dry_run:
+        print(f"   🧪 Dry-run simulation completed successfully. No files were written to: {output_dir.absolute()}")
+    else:
+        print(f"   Generated photos are saved in: {output_dir.absolute()}")
+
+
+if __name__ == "__main__":
+    main()
@@ -0,0 +1,35 @@
+import json
+import asyncio
+from agents import SocialAgents
+from database import Database
+
+async def generate():
+    db = Database()
+    agents = SocialAgents()
+    
+    with open('config.json', 'r') as f:
+        config = json.load(f)
+        artists = config['artists']
+
+    for artist in artists:
+        try:
+            print(f"Generazione per {artist['name']}...")
+            result = agents.run_for_artist(artist)
+            
+            if result['caption']:
+                db.save_draft(
+                    artist_id=artist['id'],
+                    title=result['title'],
+                    caption=result['caption'],
+                    hashtags=result['hashtags'],
+                    image_path=result['image_path'],
+                    audio_analysis=result['audio_analysis']
+                )
+                print(f"✅ Bozza salvata per {artist['name']}")
+            else:
+                print(f"⚠️ Nessun contenuto generato per {artist['name']} (controlla se ci sono file)")
+        except Exception as e:
+            print(f"❌ Errore per {artist['name']}: {str(e)}")
+
+if __name__ == "__main__":
+    asyncio.run(generate())
@@ -0,0 +1,36 @@
+import os
+import requests
+from dotenv import load_dotenv
+
+load_dotenv()
+
+def inspect_asset_detail():
+    token = os.getenv("BUFFER_ACCESS_TOKEN")
+    url = "https://api.buffer.com/graphql"
+    headers = {
+        "Authorization": f"Bearer {token}",
+        "Content-Type": "application/json"
+    }
+    
+    query = """
+    query {
+      __type(name: "ImageAssetInput") {
+        inputFields {
+          name
+        }
+      }
+    }
+    """
+    
+    response = requests.post(url, headers=headers, json={'query': query})
+    if response.status_code == 200:
+        data = response.json()
+        fields = data['data']['__type']['inputFields']
+        print("\n--- CAMPI DI IMAGEASSETINPUT ---")
+        for f in fields:
+            print(f"Campo: {f['name']}")
+    else:
+        print(f"Errore {response.status_code}")
+
+if __name__ == "__main__":
+    inspect_asset_detail()
@@ -0,0 +1,28 @@
+import os
+import requests
+import json
+from dotenv import load_dotenv
+
+load_dotenv()
+token = os.getenv("BUFFER_ACCESS_TOKEN")
+url = "https://api.buffer.com/graphql"
+headers = {"Authorization": f"Bearer {token}", "Content-Type": "application/json"}
+
+query = """
+query {
+  __type(name: "AssetsInput") {
+    inputFields { 
+      name 
+      type { 
+        name 
+        kind 
+        ofType { name kind } 
+      } 
+    }
+  }
+}
+
+"""
+
+resp = requests.post(url, headers=headers, json={'query': query})
+print(json.dumps(resp.json(), indent=2))
@@ -0,0 +1,45 @@
+import os
+import requests
+from dotenv import load_dotenv
+
+load_dotenv()
+
+def inspect_metadata():
+    token = os.getenv("BUFFER_ACCESS_TOKEN")
+    url = "https://api.buffer.com/graphql"
+    headers = {
+        "Authorization": f"Bearer {token}",
+        "Content-Type": "application/json"
+    }
+    
+    # Cerchiamo il tipo del campo 'metadata'
+    query = """
+    query {
+      __type(name: "CreatePostInput") {
+        inputFields {
+          name
+          type {
+            name
+            kind
+            ofType {
+              name
+              kind
+            }
+          }
+        }
+      }
+    }
+    """
+    
+    response = requests.post(url, headers=headers, json={'query': query})
+    if response.status_code == 200:
+        data = response.json()
+        fields = data['data']['__type']['inputFields']
+        for f in fields:
+            if f['name'] == 'metadata':
+                print(f"Tipo di metadata: {f['type']['name'] or f['type']['ofType']['name']}")
+    else:
+        print(f"Errore {response.status_code}")
+
+if __name__ == "__main__":
+    inspect_metadata()
@@ -0,0 +1,39 @@
+import os
+import requests
+from dotenv import load_dotenv
+
+load_dotenv()
+
+def inspect_enums():
+    token = os.getenv("BUFFER_ACCESS_TOKEN")
+    url = "https://api.buffer.com/graphql"
+    headers = {
+        "Authorization": f"Bearer {token}",
+        "Content-Type": "application/json"
+    }
+    
+    query = """
+    query {
+      sched: __type(name: "SchedulingType") {
+        enumValues { name }
+      }
+      mode: __type(name: "ShareMode") {
+        enumValues { name }
+      }
+    }
+    """
+    
+    response = requests.post(url, headers=headers, json={'query': query})
+    if response.status_code == 200:
+        data = response.json()
+        print("\n--- VALORI SCHEDULING ---")
+        for v in data['data']['sched']['enumValues']:
+            print(v['name'])
+        print("\n--- VALORI SHARE MODE ---")
+        for v in data['data']['mode']['enumValues']:
+            print(v['name'])
+    else:
+        print(f"Errore {response.status_code}")
+
+if __name__ == "__main__":
+    inspect_enums()
@@ -0,0 +1,52 @@
+import os
+import requests
+from dotenv import load_dotenv
+
+load_dotenv()
+
+def inspect_enum():
+    token = os.getenv("BUFFER_ACCESS_TOKEN")
+    url = "https://api.buffer.com/graphql"
+    headers = {
+        "Authorization": f"Bearer {token}",
+        "Content-Type": "application/json"
+    }
+    
+    # Cerchiamo il tipo del campo 'type' in InstagramPostMetadataInput
+    query = """
+    query {
+      __type(name: "InstagramPostMetadataInput") {
+        inputFields {
+          name
+          type {
+            name
+            kind
+          }
+        }
+      }
+    }
+    """
+    
+    response = requests.post(url, headers=headers, json={'query': query})
+    data = response.json()
+    for f in data['data']['__type']['inputFields']:
+        if f['name'] == 'type':
+            print(f"Tipo del campo 'type': {f['type']['name']}")
+            # Ora ispezioniamo quell'ENUM
+            enum_query = f"""
+            query {{
+              __type(name: "{f['type']['name']}") {{
+                enumValues {{
+                  name
+                }}
+              }}
+            }}
+            """
+            res_enum = requests.post(url, headers=headers, json={'query': enum_query})
+            data_enum = res_enum.json()
+            print("Valori possibili:")
+            for val in data_enum['data']['__type']['enumValues']:
+                print(f"- {val['name']}")
+
+if __name__ == "__main__":
+    inspect_enum()
@@ -0,0 +1,51 @@
+import os
+import requests
+from dotenv import load_dotenv
+
+load_dotenv()
+
+def introspect():
+    token = os.getenv("BUFFER_ACCESS_TOKEN")
+    url = "https://api.buffer.com/graphql"
+    headers = {
+        "Authorization": f"Bearer {token}",
+        "Content-Type": "application/json"
+    }
+    
+    # Query per elencare tutte le mutation disponibili
+    query = """
+    query {
+      __schema {
+        mutationType {
+          fields {
+            name
+            args {
+              name
+              type {
+                name
+                kind
+              }
+            }
+          }
+        }
+      }
+    }
+    """
+    
+    response = requests.post(url, headers=headers, json={'query': query})
+    if response.status_code == 200:
+        data = response.json()
+        if 'errors' in data:
+            print(f"Errore: {data['errors']}")
+        else:
+            fields = data['data']['__schema']['mutationType']['fields']
+            print("\n--- MUTATION DISPONIBILI SU BUFFER ---")
+            for f in fields:
+                print(f"Nome: {f['name']}")
+                for arg in f['args']:
+                    print(f"  Arg: {arg['name']} | Type: {arg['type']['name']} ({arg['type']['kind']})")
+    else:
+        print(f"Errore {response.status_code}: {response.text}")
+
+if __name__ == "__main__":
+    introspect()
@@ -0,0 +1,40 @@
+import os
+import requests
+from dotenv import load_dotenv
+
+load_dotenv()
+
+def introspect_payload():
+    token = os.getenv("BUFFER_ACCESS_TOKEN")
+    url = "https://api.buffer.com/graphql"
+    headers = {
+        "Authorization": f"Bearer {token}",
+        "Content-Type": "application/json"
+    }
+    
+    query = """
+    query {
+      __type(name: "PostActionPayload") {
+        fields {
+          name
+          type {
+            name
+            kind
+          }
+        }
+      }
+    }
+    """
+    
+    response = requests.post(url, headers=headers, json={'query': query})
+    if response.status_code == 200:
+        data = response.json()
+        fields = data['data']['__type']['fields']
+        print("\n--- CAMPI DI POSTACTIONPAYLOAD ---")
+        for f in fields:
+            print(f"Campo: {f['name']} | Tipo: {f['type']['name']}")
+    else:
+        print(f"Errore {response.status_code}")
+
+if __name__ == "__main__":
+    introspect_payload()
@@ -0,0 +1,56 @@
+import os
+import requests
+from dotenv import load_dotenv
+
+load_dotenv()
+
+def list_profiles():
+    token = os.getenv("BUFFER_ACCESS_TOKEN")
+    if not token:
+        print("Errore: BUFFER_ACCESS_TOKEN non trovato nel file .env")
+        return
+
+    # Nuova URL per GraphQL API
+    url = "https://api.buffer.com/graphql"
+    headers = {
+        "Authorization": f"Bearer {token}",
+        "Content-Type": "application/json"
+    }
+    
+    # Query GraphQL per ottenere l'account e le organizzazioni
+    query = """
+    query {
+      account {
+        organizations {
+          id
+          name
+          channels {
+            id
+            service
+            name
+          }
+        }
+      }
+    }
+    """
+    
+    response = requests.post(url, headers=headers, json={'query': query})
+    
+    if response.status_code == 200:
+        data = response.json()
+        if 'errors' in data:
+            print(f"Errore GraphQL: {data['errors']}")
+            return
+
+        organizations = data['data']['account']['organizations']
+        print("\n--- ORGANIZZAZIONI E CANALI BUFFER ---")
+        for org in organizations:
+            print(f"\nOrganizzazione: {org['name']} (ID: {org['id']})")
+            for channel in org['channels']:
+                print(f"  - ID Canale: {channel['id']} | Social: {channel['service']} | Nome: {channel['name']}")
+        print("---------------------------------------\n")
+    else:
+        print(f"Errore API Buffer: {response.status_code} - {response.text}")
+
+if __name__ == "__main__":
+    list_profiles()
@@ -0,0 +1,141 @@
+import os
+import sys
+import json
+import asyncio
+from pathlib import Path
+from dotenv import load_dotenv
+
+# Load environment variables
+load_dotenv()
+
+async def fetch_mcp_details(api_key, api_secret=None):
+    from mcp.client.streamable_http import streamablehttp_client
+    from mcp import ClientSession
+
+    url = "https://mcp.higgsfield.ai/mcp"
+    
+    # MCP server accepts Bearer tokens (unlike REST API which uses "Key" format)
+    # Note: MCP server is designed for OAuth browser auth, API keys provide limited access
+    headers = {
+        "Authorization": f"Bearer {api_key}"
+    }
+    
+    print(f"🔌 Connecting to Higgsfield hosted MCP server ({url})...")
+    
+    try:
+        async with streamablehttp_client(url, headers=headers) as (read, write, _):
+            async with ClientSession(read, write) as session:
+                print("   Initializing MCP session...")
+                await session.initialize()
+                print("✅ Session initialized successfully!")
+                
+                # 1. Query available tools
+                print("\n🔍 Fetching available tools...")
+                tools_res = await session.list_tools()
+                tools = tools_res.tools
+                print(f"   Found {len(tools)} tools:")
+                for t in tools:
+                    desc = t.description[:80] if t.description else "No description"
+                    print(f"   🛠️ Tool: {t.name} - {desc}...")
+                
+                # 2. Query available resources
+                print("\n🔍 Fetching available resources...")
+                resources_res = await session.list_resources()
+                resources = resources_res.resources
+                print(f"   Found {len(resources)} resources:")
+                for r in resources:
+                    print(f"   🔗 Resource: {r.uri} - {r.name}")
+                
+                # 3. Pre-flight check: verify API key works by calling balance
+                print("\n🔑 Verifying API key validity...")
+                try:
+                    preflight = await session.call_tool("balance", {})
+                    preflight_text = preflight.content[0].text if preflight.content else ""
+                    if "something went wrong" in preflight_text.lower() or "error" in preflight_text.lower():
+                        print(f"   ⚠️ API key may be invalid or expired. Server response: {preflight_text}")
+                        print("   💡 Tip: Log in to https://higgsfield.ai and regenerate your API key.")
+                        print("   The MCP connection works, but tool calls are failing server-side.")
+                    else:
+                        print(f"   ✅ API key verified. Account info: {preflight_text}")
+                except Exception as e:
+                    print(f"   ⚠️ Pre-flight check failed: {e}")
+                
+                # 4. Try to list characters via the show_characters tool
+                print("\n📖 Fetching Soul Characters via show_characters tool...")
+                char_tools = [t for t in tools if "character" in t.name.lower() or "soul" in t.name.lower()]
+                
+                if char_tools:
+                    for t in char_tools:
+                        print(f"   Calling tool: {t.name}...")
+                        try:
+                            result = await session.call_tool(t.name, {"action": "list", "status": "ready", "size": 50})
+                            print(f"✅ Tool result:")
+                            print("-" * 50)
+                            if result.content:
+                                text = result.content[0].text
+                                try:
+                                    parsed = json.loads(text)
+                                    if isinstance(parsed, list):
+                                        for char in parsed:
+                                            char_id = char.get("id") or char.get("uuid") or char.get("soul_id")
+                                            name = char.get("name") or "Unnamed Character"
+                                            status = char.get("status", "unknown")
+                                            print(f"🆔 Soul ID: {char_id}")
+                                            print(f"👤 Name   : {name}")
+                                            print(f"📊 Status : {status}")
+                                            print("-" * 50)
+                                    elif isinstance(parsed, dict):
+                                        items = parsed.get("characters") or parsed.get("items") or parsed.get("data") or [parsed]
+                                        if isinstance(items, list):
+                                            for char in items:
+                                                char_id = char.get("id") or char.get("uuid") or char.get("soul_id")
+                                                name = char.get("name") or "Unnamed Character"
+                                                status = char.get("status", "unknown")
+                                                print(f"🆔 Soul ID: {char_id}")
+                                                print(f"👤 Name   : {name}")
+                                                print(f"📊 Status : {status}")
+                                                print("-" * 50)
+                                        else:
+                                            print(json.dumps(parsed, indent=2))
+                                    else:
+                                        print(text)
+                                except json.JSONDecodeError:
+                                    print(text)
+                            else:
+                                print("   (No content returned)")
+                            print("-" * 50)
+                        except Exception as tool_err:
+                                print(f"   ⚠️ Could not invoke tool {t.name}: {tool_err}")
+                else:
+                    print("   ⚠️ No character-related tools found on the server.")
+                    
+                # 4. Also check for any other listing tools
+                list_tools = [t for t in tools if "list" in t.name.lower() and t not in char_tools]
+                if list_tools:
+                    print("\n💡 Other listing tools available:")
+                    for t in list_tools:
+                        desc = t.description[:100] if t.description else "No description"
+                        print(f"   🛠️ {t.name}: {desc}")
+                                
+    except Exception as e:
+        print(f"❌ MCP Connection Error: {e}", file=sys.stderr)
+        import traceback
+        traceback.print_exc()
+        sys.exit(1)
+
+def main():
+    api_key = os.getenv("HIGGSFIELD_API_KEY")
+    api_secret = os.getenv("HIGGSFIELD_SECRET")
+    
+    if not api_key or api_key == "your_higgsfield_api_key_here":
+        print("❌ Error: HIGGSFIELD_API_KEY is not configured in .env.")
+        print("   Please add your real token to .env to authenticate.", file=sys.stderr)
+        sys.exit(1)
+    
+    if not api_secret:
+        print("⚠️ Warning: HIGGSFIELD_SECRET is not set. Some API calls may fail.")
+        
+    asyncio.run(fetch_mcp_details(api_key, api_secret))
+
+if __name__ == "__main__":
+    main()
@@ -0,0 +1,11 @@
+import os
+import google.generativeai as genai
+from dotenv import load_dotenv
+
+load_dotenv()
+genai.configure(api_key=os.getenv("GOOGLE_API_KEY"))
+
+print("Modelli disponibili:")
+for m in genai.list_models():
+    if 'generateContent' in m.supported_generation_methods:
+        print(m.name)
@@ -0,0 +1,49 @@
+import os
+import requests
+from dotenv import load_dotenv
+
+load_dotenv()
+
+def list_profiles():
+    token = os.getenv("BUFFER_ACCESS_TOKEN")
+    url = "https://api.buffer.com/graphql"
+    headers = {
+        "Authorization": f"Bearer {token}",
+        "Content-Type": "application/json"
+    }
+    
+    query = """
+    query {
+      account {
+        id
+        email
+        organizations {
+          id
+          name
+          channels(input: {}) {
+            id
+            name
+            service
+          }
+        }
+      }
+    }
+    """
+    
+    response = requests.post(url, headers=headers, json={'query': query})
+    if response.status_code == 200:
+        data = response.json()
+        acc = data.get('data', {}).get('account')
+        if acc:
+            print(f"\nAccount: {acc['email']} ({acc['id']})")
+            for org in acc['organizations']:
+                print(f"\n--- ORGANIZZAZIONE: {org['name']} ({org['id']}) ---")
+                for c in org['channels']:
+                    print(f"ID: {c['id']} | Servizio: {c['service']} | Nome: {c['name']}")
+        else:
+            print("Account non trovato:", data)
+    else:
+        print(f"Errore {response.status_code}: {response.text}")
+
+if __name__ == "__main__":
+    list_profiles()
@@ -0,0 +1,99 @@
+import os
+import json
+import time
+import asyncio
+from datetime import datetime
+from apscheduler.schedulers.background import BackgroundScheduler
+from telegram import Bot
+from agents import SocialAgents
+from database import Database
+from dotenv import load_dotenv
+
+load_dotenv()
+
+class SchedulerService:
+    def __init__(self):
+        self.db = Database()
+        self.agents = SocialAgents()
+        self.bot_token = os.getenv("TELEGRAM_BOT_TOKEN")
+        self.chat_id = os.getenv("TELEGRAM_CHAT_ID")
+        self.bot = Bot(token=self.bot_token) if self.bot_token else None
+
+    async def send_notification(self, message):
+        if self.bot and self.chat_id:
+            try:
+                await self.bot.send_message(chat_id=self.chat_id, text=message)
+            except Exception as e:
+                print(f"Errore invio Telegram: {e}")
+
+    def run_single_artist_job(self, artist):
+        print(f"[{datetime.now()}] Avvio generazione per {artist['name']}...")
+        try:
+            # Sincronizzazione automatica da Drive e locale
+            os.system("python src/cloud_sync.py")
+            os.system("python src/sync_assets.py")
+
+            # Validazione materiale prima della generazione
+            audio = self.db.get_audio_for_artist(artist['id'])
+            image = self.db.get_unused_image(artist['id'])
+            
+            if not audio or not image:
+                msg = f"⚠️ Materiale mancante per {artist['name']}. Caricare MP3 e Foto!"
+                print(msg)
+                asyncio.run(self.send_notification(msg))
+                return
+
+            results = self.agents.run_for_artist(artist)
+            for result in results:
+                if 'error' not in result:
+                    self.db.save_draft(
+                        artist_id=artist['id'],
+                        title=result['title'],
+                        caption=result['caption'],
+                        hashtags=result['hashtags'],
+                        image_path=result['image_path'],
+                        image_url=result.get('image_url'),
+                        audio_analysis=result['audio_analysis'],
+                        video_url=result.get('video_url')
+                    )
+            asyncio.run(self.send_notification(f"✅ Post pronti per {artist['name']} (Foto + Video)!"))
+        except Exception as e:
+            err = f"❌ Errore per {artist['name']}: {str(e)}"
+            print(err)
+            asyncio.run(self.send_notification(err))
+
+    def start(self):
+        scheduler = BackgroundScheduler()
+        
+        with open('config.json', 'r') as f:
+            config = json.load(f)
+            artists = config['artists']
+
+        for artist in artists:
+            s_time = artist.get('schedule_time', '09:00')
+            try:
+                hour, minute = map(int, s_time.split(':'))
+                scheduler.add_job(
+                    self.run_single_artist_job, 
+                    'cron', 
+                    hour=hour, 
+                    minute=minute, 
+                    args=[artist],
+                    id=f"job_{artist['id'].replace(' ', '_')}"
+                )
+                print(f"Pianificato {artist['name']} alle {s_time}")
+            except Exception as e:
+                print(f"Errore pianificazione per {artist['name']}: {e}")
+
+        scheduler.start()
+        print("Scheduler avviato.")
+        
+        try:
+            while True:
+                time.sleep(60)
+        except (KeyboardInterrupt, SystemExit):
+            scheduler.shutdown()
+
+if __name__ == "__main__":
+    service = SchedulerService()
+    service.start()
@@ -0,0 +1,35 @@
+import os
+import json
+from database import Database
+
+def sync():
+    db = Database()
+    
+    # Load config to get artists
+    with open('config.json', 'r') as f:
+        config = json.load(f)
+        artists = [a['id'] for a in config['artists']]
+
+    base_path = "data"
+    
+    for artist_id in artists:
+        # Sync Images
+        img_dir = os.path.join(base_path, "images", artist_id)
+        if os.path.exists(img_dir):
+            for file in os.listdir(img_dir):
+                if file.lower().endswith(('.png', '.jpg', '.jpeg', '.webp')):
+                    file_path = os.path.join(img_dir, file)
+                    db.add_media(file_path, 'image', artist_id)
+        
+        # Sync Audio
+        audio_dir = os.path.join(base_path, "audio", artist_id)
+        if os.path.exists(audio_dir):
+            for file in os.listdir(audio_dir):
+                if file.lower().endswith(('.mp3', '.wav', '.m4a')):
+                    file_path = os.path.join(audio_dir, file)
+                    db.add_media(file_path, 'audio', artist_id)
+
+    print("Sincronizzazione completata.")
+
+if __name__ == "__main__":
+    sync()
@@ -0,0 +1,49 @@
+import os
+import requests
+from dotenv import load_dotenv
+
+load_dotenv()
+
+def test_publish():
+    token = os.getenv("BUFFER_ACCESS_TOKEN")
+    url = "https://api.buffer.com/graphql"
+    headers = {
+        "Authorization": f"Bearer {token}",
+        "Content-Type": "application/json"
+    }
+    
+    mutation = """
+    mutation ($input: CreatePostInput!) {
+      createPost(input: $input) {
+        ... on PostActionSuccess {
+          post {
+            id
+          }
+        }
+        ... on MutationError {
+          message
+        }
+      }
+    }
+    """
+    
+    variables = {
+        "input": {
+            "channelId": "69f9ea855c4c051afa117baa",
+            "text": "Test di pubblicazione via API (In Coda) 🎼🚀",
+            "schedulingType": "automatic",
+            "mode": "addToQueue"
+        }
+    }
+
+    print("Inviando TEST REALE (In Coda) a Buffer...")
+    response = requests.post(url, headers=headers, json={
+        'query': mutation,
+        'variables': variables
+    })
+    
+    print(f"Status Code: {response.status_code}")
+    print(f"Risposta: {response.text}")
+
+if __name__ == "__main__":
+    test_publish()
@@ -0,0 +1,51 @@
+import os
+import requests
+from dotenv import load_dotenv
+
+load_dotenv()
+
+def test_draft():
+    token = os.getenv("BUFFER_ACCESS_TOKEN")
+    url = "https://api.buffer.com/graphql"
+    headers = {
+        "Authorization": f"Bearer {token}",
+        "Content-Type": "application/json"
+    }
+    
+    # Canale Instagram
+    channel_id = "69f9ea855c4c051afa117baa"
+    
+    mutation = """
+    mutation ($input: CreatePostInput!) {
+      createPost(input: $input) {
+        ... on PostActionSuccess {
+          post {
+            id
+            text
+          }
+        }
+        ... on MutationError {
+          message
+        }
+      }
+    }
+    """
+    
+    variables = {
+        "input": {
+            "channelId": channel_id,
+            "text": "TEST BOZZA DAL BOT - " + os.popen("date").read().strip(),
+            "schedulingType": "automatic",
+            "mode": "addToQueue",
+            "saveToDraft": True
+        }
+    }
+    
+    response = requests.post(url, headers=headers, json={'query': mutation, 'variables': variables})
+    if response.status_code == 200:
+        print("Risposta Buffer:", response.json())
+    else:
+        print(f"Errore {response.status_code}: {response.text}")
+
+if __name__ == "__main__":
+    test_draft()
@@ -0,0 +1,28 @@
+import json
+from agents import SocialAgents
+from database import Database
+
+def test():
+    db = Database()
+    agents = SocialAgents()
+    
+    with open('config.json', 'r') as f:
+        config = json.load(f)
+        # Cerchiamo l'artista 2
+        artist = next((a for a in config['artists'] if a['id'] == 'artista_2'), None)
+        
+    if artist:
+        print(f"--- Inizio generazione per {artist['name']} ---")
+        result = agents.run_for_artist(artist)
+        
+        print("\n--- RISULTATO GENERAZIONE ---")
+        print(f"IMMAGINE: {result['image_path']}")
+        print(f"\nANALISI AUDIO:\n{result['audio_analysis']}")
+        print(f"\nCAPTION:\n{result['caption']}")
+        print(f"\nHASHTAGS:\n{result['hashtags']}")
+        print("-----------------------------")
+    else:
+        print("Artista 2 non trovato in config.json")
+
+if __name__ == "__main__":
+    test()
@@ -0,0 +1,184 @@
+import os
+import subprocess
+import requests
+import json
+import time
+
+class VideoGenerator:
+    def __init__(self, cache_dir="data/cache"):
+        self.cache_dir = cache_dir
+        os.makedirs(self.cache_dir, exist_ok=True)
+
+    def generate_video(self, media_paths, audio_path, title, start_time=0, output_path=None, bpm=120, focus_points=None):
+        if isinstance(media_paths, str):
+            media_paths = [media_paths]
+            
+        if not output_path:
+            output_path = os.path.join(self.cache_dir, "final_video.mp4")
+            
+        # Assicuriamoci che la cartella di destinazione esista
+        os.makedirs(os.path.dirname(output_path), exist_ok=True)
+        
+        if os.path.exists(output_path):
+            os.remove(output_path)
+
+        # Assicuriamoci che focus_points sia una lista della stessa lunghezza di media_paths
+        if not focus_points:
+            focus_points = [None] * len(media_paths)
+            
+        safe_title = title.replace(":", "\\:").replace("'", "").replace("%", "")
+        
+        # Calcolo tempi basato su BPM (Sincronizzazione)
+        beat_duration = 60.0 / bpm
+        num_assets = len(media_paths)
+        
+        # Costruiamo il comando FFmpeg dinamico
+        inputs = []
+        for asset in media_paths:
+            inputs.extend(['-i', asset])
+        
+        filter_complex = ""
+        
+        import random
+        for i, asset in enumerate(media_paths):
+            is_video = asset.lower().endswith(('.mp4', '.mov', '.avi', '.mkv'))
+            focus = focus_points[i] if i < len(focus_points) else None
+            
+            # Ritmo basato su multipli del beat
+            beats_per_clip = random.choice([2, 4])
+            if random.random() > 0.8: beats_per_clip = 8
+            
+            d_img_curr = beat_duration * beats_per_clip
+            d_frames = int(d_img_curr * 25)
+            
+            # Effetti comuni
+            clean_effects = [
+                "eq=contrast=1.2:saturation=1.5",
+                "colorbalance=rs=0.2:gs=0.1:bs=0.2,eq=contrast=1.3:saturation=1.4",
+                "eq=brightness=0.05:contrast=1.4:saturation=1.3",
+                "unsharp=5:5:1.0:5:5:0.0"
+            ]
+            effect = random.choice(clean_effects)
+            if random.random() > 0.7:
+                effect += ",rgbashift=rh=2:rv=2:gh=-2:gv=-2"
+            
+            flash = f"fade=t=in:st=0:d=0.2:color=white,"
+
+            if not is_video:
+                # LOGICA IMMAGINE: Zoompan
+                zoom_speed = 0.005 if beats_per_clip < 4 else 0.002
+                
+                if focus:
+                    # Se abbiamo il focus dall'AI, puntiamo lì (coordinate 0-100)
+                    fx = focus.get('x', 50) / 100.0
+                    fy = focus.get('y', 40) / 100.0 # Leggermente più su del centro se non specificato
+                    
+                    # Centriamo il pan sul punto di interesse
+                    target_x = f"iw*{fx}-(iw/zoom/2)"
+                    target_y = f"ih*{fy}-(ih/zoom/2)"
+                else:
+                    # Altrimenti movimento casuale classico
+                    dir_x = random.uniform(-0.5, 0.5)
+                    dir_y = random.uniform(-0.3, 0.3)
+                    target_x = f"iw/2-(iw/zoom/2)+({dir_x}*on)"
+                    target_y = f"ih/2-(ih/zoom/2)+({dir_y}*on)"
+                
+                filter_complex += f"[{i}:v]scale=1080:1920:force_original_aspect_ratio=increase,crop=1080:1920," \
+                                  f"zoompan=z='min(max(zoom,1.0)+{zoom_speed},1.5)':" \
+                                  f"x='{target_x}':" \
+                                  f"y='{target_y}':" \
+                                  f"d={d_frames}:s=1080x1920," \
+                                  f"{flash}{effect},setsar=1[v{i}];"
+            else:
+                # LOGICA VIDEO: Trim e Scale
+                # Prendiamo uno spezzone casuale (assumiamo video > 5s per sicurezza)
+                start_trim = random.uniform(0, 2.0)
+                filter_complex += f"[{i}:v]scale=1080:1920:force_original_aspect_ratio=increase,crop=1080:1920," \
+                                  f"trim=start={start_trim}:duration={d_img_curr},setpts=PTS-STARTPTS," \
+                                  f"fps=25,{flash}{effect},setsar=1[v{i}];"
+        
+        concat_inputs = ""
+        for i in range(num_assets):
+            concat_inputs += f"[v{i}]"
+            
+        filter_complex += f"{concat_inputs}concat=n={num_assets}:v=1:a=0[outv]"
+
+        temp_block_path = os.path.join(self.cache_dir, "temp_block.mp4")
+        cmd_block = ['ffmpeg', '-y'] + inputs + [
+            '-filter_complex', filter_complex,
+            '-map', '[outv]',
+            '-c:v', 'libx264', '-crf', '23', '-preset', 'fast', '-pix_fmt', 'yuv420p',
+            temp_block_path
+        ]
+
+        print(f"🛠️ FFmpeg: Generazione blocco base (Filtro complesso)...")
+        res1 = subprocess.run(cmd_block, capture_output=True, text=True)
+        if res1.returncode != 0:
+            print(f"❌ ERRORE FFMPEG BLOCK: {res1.stderr}")
+            raise Exception("FFmpeg failed to create base block")
+            
+        print(f"🛠️ FFmpeg: Assemblaggio finale (Loop + Audio)...")
+        cmd_final = [
+            'ffmpeg', '-y',
+            '-stream_loop', '-1', '-i', temp_block_path,
+            '-ss', str(start_time), '-i', audio_path,
+            '-c:v', 'copy',
+            '-c:a', 'aac', '-b:a', '192k',
+            '-t', '30', '-shortest',
+            output_path
+        ]
+        
+        res2 = subprocess.run(cmd_final, capture_output=True, text=True)
+        if os.path.exists(temp_block_path):
+            os.remove(temp_block_path)
+            
+        if res2.returncode != 0:
+            print(f"ERRORE FFMPEG FINAL: {res2.stderr}")
+            raise Exception("FFmpeg failed to assemble final video")
+            
+        return output_path
+
+    def generate_thumbnail(self, video_path, output_path=None):
+        """Genera una miniatura locale dal video generato"""
+        if not output_path:
+            output_path = video_path.replace(".mp4", ".jpg")
+            
+        if os.path.exists(output_path): return output_path
+        
+        print(f"🖼️ Generazione miniatura locale: {output_path}...")
+        cmd = [
+            'ffmpeg', '-y', '-ss', '00:00:01', 
+            '-i', video_path, 
+            '-vframes', '1', '-q:v', '2', 
+            output_path
+        ]
+        res = subprocess.run(cmd, capture_output=True)
+        if res.returncode == 0:
+            return output_path
+        return None
+
+    def upload_to_ephemeral(self, file_path):
+        """Carica su host temporanei per ottenere un URL pubblico per Buffer"""
+        # Tentativo 1: BashUpload (Molto semplice e diretto)
+        print(f"Tentativo upload su BashUpload...")
+        try:
+            filename = os.path.basename(file_path)
+            with open(file_path, 'rb') as f:
+                response = requests.put(f'https://bashupload.com/{filename}', data=f, timeout=30)
+                if response.status_code == 200:
+                    # BashUpload restituisce l'URL nel corpo della risposta
+                    for line in response.text.split('\n'):
+                        if 'https://bashupload.com/' in line:
+                            return line.strip().split(' ')[-1]
+        except Exception as e: print(f"BashUpload errore: {e}")
+
+        # Tentativo 2: Uguu.se
+        print(f"Tentativo upload su Uguu.se...")
+        try:
+            with open(file_path, 'rb') as f:
+                response = requests.post('https://uguu.se/upload.php', files={'files[]': f}, timeout=30)
+                if response.status_code == 200:
+                    return response.json()['files'][0]['url']
+        except Exception as e: print(f"Uguu.se errore: {e}")
+        
+        return None