Michele/postgenerator
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

feat(04-01): integrazione Unsplash in pipeline + CSVBuilder + export

Browse Source 
- CSVBuilder.build_csv() e build_csv_content() accettano image_url_map opzionale
- _resolve_image() risolve keyword->URL Unsplash con fallback keyword originale
- _build_rows() chiama _resolve_image per cover, slides e cta image keywords
- JobStatus ha campo image_url_map con persistenza su disco JSON
- GenerationPipeline._resolve_unsplash_keywords() chiamato dopo batch LLM
- Carica unsplash_api_key da settings.json, crea UnsplashService, chiama resolve_keywords
- image_url_map salvato nel job JSON per riuso in export con edits
- Export router recupera image_url_map dal job JSON e passa a build_csv_content
- generate_single NON risolve Unsplash (velocità e riuso map job originale)
This commit is contained in:
Michele
2026-03-09 08:10:06 +01:00
parent afba4c5e9e
commit 9e7205eca2
3 changed files with 157 additions and 9 deletions
Download Patch File Download Diff File Expand all files Collapse all files

47
backend/services/csv_builder.py
View File

@@ -6,6 +6,7 @@ Caratteristiche:
- Mappa GeneratedPost + CalendarSlot -> riga CSV
- Filtra solo PostResult con status="success"
- Scrive su disco in OUTPUTS_PATH/{job_id}.csv
- Supporta image_url_map opzionale: risolve keyword -> URL Unsplash nelle colonne _image_keyword
"""
from __future__ import annotations
@@ -14,7 +15,7 @@ import csv
import io
import logging
from pathlib import Path
from typing import TYPE_CHECKING
from typing import TYPE_CHECKING, Optional
from backend.constants import CANVA_FIELDS
from backend.schemas.generate import PostResult
@@ -35,6 +36,7 @@ class CSVBuilder:
calendar: "CalendarResponse",
job_id: str,
output_dir: Path,
image_url_map: Optional[dict[str, str]] = None,
) -> Path:
"""Genera e scrive il CSV su disco.
@@ -42,11 +44,16 @@ class CSVBuilder:
GeneratedPost + CalendarSlot alle 33 colonne CANVA_FIELDS,
e scrive con encoding utf-8-sig per compatibilità Excel.
Se image_url_map è fornita, le colonne _image_keyword contengono
URL Unsplash reali invece delle keyword testuali originali.
Args:
posts: Lista di PostResult (include success e failed).
calendar: CalendarResponse con i metadati degli slot.
job_id: Identificatore univoco del job (usato come nome file).
output_dir: Directory dove scrivere il file CSV.
image_url_map: Mappa opzionale {keyword: url_unsplash}. Se None,
usa le keyword testuali originali.
Returns:
Path del file CSV scritto su disco.
@@ -54,7 +61,7 @@ class CSVBuilder:
output_dir.mkdir(parents=True, exist_ok=True)
output_path = output_dir / f"{job_id}.csv"
rows = self._build_rows(posts, calendar)
rows = self._build_rows(posts, calendar, image_url_map)
# CRITICO: encoding utf-8-sig (BOM) per Excel + caratteri italiani
with open(output_path, "w", newline="", encoding="utf-8-sig") as f:
@@ -62,10 +69,12 @@ class CSVBuilder:
writer.writeheader()
writer.writerows(rows)
url_count = len(image_url_map) if image_url_map else 0
logger.info(
"CSV scritto | job_id=%s | righe_success=%d | path=%s",
"CSV scritto | job_id=%s | righe_success=%d | url_unsplash=%d | path=%s",
job_id,
len(rows),
url_count,
output_path,
)
return output_path
@@ -75,21 +84,26 @@ class CSVBuilder:
posts: list[PostResult],
calendar: "CalendarResponse",
job_id: str,
image_url_map: Optional[dict[str, str]] = None,
) -> str:
"""Genera il CSV come stringa (senza scrivere su disco).
Usato per preview e per la route POST /export/{job_id}/csv
con dati modificati inline dall'utente.
Se image_url_map è fornita, le colonne _image_keyword contengono
URL Unsplash reali invece delle keyword testuali originali.
Args:
posts: Lista di PostResult (include success e failed).
calendar: CalendarResponse con i metadati degli slot.
job_id: Identificatore univoco del job.
image_url_map: Mappa opzionale {keyword: url_unsplash}.
Returns:
Stringa CSV con encoding utf-8-sig (BOM).
"""
rows = self._build_rows(posts, calendar)
rows = self._build_rows(posts, calendar, image_url_map)
output = io.StringIO()
# Aggiungi BOM manualmente per compatibilità Excel
@@ -103,19 +117,38 @@ class CSVBuilder:
# Metodi privati
# ---------------------------------------------------------------------------
def _resolve_image(self, keyword: str, image_url_map: Optional[dict[str, str]]) -> str:
"""Risolve una keyword immagine in URL Unsplash se disponibile.
Args:
keyword: Keyword immagine originale.
image_url_map: Mappa {keyword: url} o None.
Returns:
URL Unsplash se disponibile nella mappa, altrimenti la keyword originale.
"""
if image_url_map and keyword in image_url_map:
return image_url_map[keyword]
return keyword
def _build_rows(
self,
posts: list[PostResult],
calendar: "CalendarResponse",
image_url_map: Optional[dict[str, str]] = None,
) -> list[dict[str, str]]:
"""Costruisce la lista di righe CSV dai risultati.
Filtra solo i post con status="success" e mappa i dati
GeneratedPost + CalendarSlot alle colonne CANVA_FIELDS.
Se image_url_map è fornita, le colonne _image_keyword vengono
risolte in URL Unsplash quando disponibili.
Args:
posts: Lista completa di PostResult.
calendar: CalendarResponse con i metadati degli slot.
image_url_map: Mappa opzionale {keyword: url_unsplash}.
Returns:
Lista di dict con chiavi = CANVA_FIELDS.
@@ -152,7 +185,7 @@ class CSVBuilder:
# --- Cover slide (3 colonne) ---
row["cover_title"] = post.cover_title
row["cover_subtitle"] = post.cover_subtitle
row["cover_image_keyword"] = post.cover_image_keyword
row["cover_image_keyword"] = self._resolve_image(post.cover_image_keyword, image_url_map)
# --- Slide centrali s2-s7 (6 slide x 3 colonne = 18 colonne) ---
slide_labels = ["s2", "s3", "s4", "s5", "s6", "s7"]
@@ -161,7 +194,7 @@ class CSVBuilder:
slide = post.slides[idx]
row[f"{label}_headline"] = slide.headline
row[f"{label}_body"] = slide.body
row[f"{label}_image_keyword"] = slide.image_keyword
row[f"{label}_image_keyword"] = self._resolve_image(slide.image_keyword, image_url_map)
else:
# Fallback se slides ha meno di 6 elementi (non dovrebbe accadere)
row[f"{label}_headline"] = ""
@@ -171,7 +204,7 @@ class CSVBuilder:
# --- CTA slide (3 colonne) ---
row["cta_text"] = post.cta_text
row["cta_subtext"] = post.cta_subtext
row["cta_image_keyword"] = post.cta_image_keyword
row["cta_image_keyword"] = self._resolve_image(post.cta_image_keyword, image_url_map)
# --- Caption Instagram (1 colonna) ---
row["caption_instagram"] = post.caption_instagram

106
backend/services/generation_pipeline.py
View File

@@ -18,6 +18,7 @@ from dataclasses import dataclass, field
from pathlib import Path
from typing import Literal, Optional
from backend.config import DATA_PATH
from backend.schemas.calendar import CalendarRequest, CalendarResponse, CalendarSlot
from backend.schemas.generate import (
GenerateResponse,
@@ -30,6 +31,7 @@ from backend.services.csv_builder import CSVBuilder
from backend.services.format_selector import FormatSelector
from backend.services.llm_service import LLMService
from backend.services.prompt_service import PromptService
from backend.services.unsplash_service import UnsplashService
logger = logging.getLogger(__name__)
@@ -69,6 +71,8 @@ class JobStatus:
calendar: Optional[CalendarResponse] = None
error: Optional[str] = None
campagna: str = ""
image_url_map: Optional[dict[str, str]] = None
"""Mappa keyword->URL Unsplash risolta dopo la generazione batch. None se Unsplash non configurato."""
# ---------------------------------------------------------------------------
@@ -372,6 +376,18 @@ class GenerationPipeline:
job.results.append(post_result)
job.completed += 1
# Risolvi keyword immagine via Unsplash (se API key configurata)
image_url_map: Optional[dict[str, str]] = None
try:
image_url_map = await self._resolve_unsplash_keywords(job)
job.image_url_map = image_url_map
except Exception as e:
logger.warning(
"Unsplash resolution fallita | job_id=%s | errore=%s — continuo con keyword testuali",
job_id,
str(e),
)
# Genera CSV con i risultati
success_results = [r for r in job.results if r.status == "success"]
if success_results:
@@ -381,12 +397,14 @@ class GenerationPipeline:
calendar=calendar,
job_id=job_id,
output_dir=self._outputs_path,
image_url_map=image_url_map,
)
logger.info(
"CSV generato | job_id=%s | success=%d/%d",
"CSV generato | job_id=%s | success=%d/%d | url_unsplash=%d",
job_id,
len(success_results),
job.total,
len(image_url_map) if image_url_map else 0,
)
# Salva metadata job su disco per persistenza
@@ -499,6 +517,90 @@ class GenerationPipeline:
# Altrimenti usa la mappa formato -> prompt
return _FORMAT_TO_PROMPT.get(formato, _DEFAULT_PROMPT)
# ---------------------------------------------------------------------------
# Integrazione Unsplash
# ---------------------------------------------------------------------------
async def _resolve_unsplash_keywords(
self,
job: JobStatus,
) -> Optional[dict[str, str]]:
"""Risolve le keyword immagine dei post in URL Unsplash.
Carica la settings da disco per verificare se unsplash_api_key e' configurata.
Se non e' configurata, ritorna None (nessuna risoluzione, usa keyword testuali).
Estrae tutte le keyword uniche dai PostResult success:
- cover_image_keyword
- slide.image_keyword per ogni slide
- cta_image_keyword
Args:
job: JobStatus con i risultati generati.
Returns:
Dizionario {keyword: url} per le keyword risolte, o None se Unsplash non configurato.
"""
import json as _json
# Carica settings per controllare unsplash_api_key
settings_path = DATA_PATH / "config" / "settings.json"
unsplash_api_key: Optional[str] = None
if settings_path.exists():
try:
data = _json.loads(settings_path.read_text(encoding="utf-8"))
unsplash_api_key = data.get("unsplash_api_key")
except Exception as e:
logger.warning("Errore lettura settings per Unsplash: %s", str(e))
if not unsplash_api_key:
logger.debug("unsplash_api_key non configurata — skip risoluzione keyword")
return None
# Estrai tutte le keyword uniche dai post success
keywords: list[str] = []
for post_result in job.results:
if post_result.status != "success" or post_result.post is None:
continue
post = post_result.post
if post.cover_image_keyword:
keywords.append(post.cover_image_keyword)
for slide in post.slides:
if slide.image_keyword:
keywords.append(slide.image_keyword)
if post.cta_image_keyword:
keywords.append(post.cta_image_keyword)
if not keywords:
logger.debug("Nessuna keyword immagine trovata nei post")
return None
logger.info(
"Avvio risoluzione Unsplash | job_id=%s | keyword_totali=%d",
job.job_id,
len(keywords),
)
# Crea e usa UnsplashService
unsplash_cache_path = DATA_PATH / "unsplash_cache.json"
unsplash = UnsplashService(
api_key=unsplash_api_key,
cache_path=unsplash_cache_path,
)
try:
image_url_map = await unsplash.resolve_keywords(keywords)
logger.info(
"Risoluzione Unsplash completata | job_id=%s | risolte=%d/%d",
job.job_id,
len(image_url_map),
len(set(keywords)),
)
return image_url_map if image_url_map else None
finally:
await unsplash.close()
# ---------------------------------------------------------------------------
# Persistenza su disco
# ---------------------------------------------------------------------------
@@ -523,6 +625,7 @@ class GenerationPipeline:
"error": job.error,
"results": [r.model_dump() for r in job.results],
"calendar": job.calendar.model_dump() if job.calendar else None,
"image_url_map": job.image_url_map,
}
with open(job_path, "w", encoding="utf-8") as f:
@@ -566,6 +669,7 @@ class GenerationPipeline:
calendar=calendar,
error=data.get("error"),
campagna=data.get("campagna", ""),
image_url_map=data.get("image_url_map"),
)
# Metti in memoria per accesso futuro