feat: complete translator stack

backend/Dockerfile

@@ -0,0 +1,10 @@
+FROM python:3.13-slim
+
+WORKDIR /app
+COPY backend/requirements.txt ./requirements.txt
+RUN pip install --no-cache-dir -r requirements.txt
+
+COPY backend ./backend
+ENV PYTHONPATH=/app
+
+CMD ["uvicorn", "backend.app.main:app", "--host", "0.0.0.0", "--port", "8000"]

backend/app/main.py

@@ -0,0 +1,15 @@
+from fastapi import FastAPI
+from backend.app.routers import health, translation
+
+app = FastAPI()
+
+app.include_router(health.router)
+app.include_router(translation.router, prefix="/api")
+
+@app.on_event("startup")
+async def startup_event():
+    print("Avvio del servizio di traduzione")
+
+@app.on_event("shutdown")
+async def shutdown_event():
+    await translation.ollama_client.close()

backend/app/models/__init__.py

@@ -0,0 +1 @@
+from . import session

backend/app/models/session.py

@@ -0,0 +1,25 @@
+from datetime import datetime
+from typing import List, Optional
+
+from pydantic import BaseModel, Field
+
+
+class SessionState(BaseModel):
+    session_id: str
+    file_title: str
+    text_title: Optional[str]
+    source_lang: str
+    target_lang: str
+    total_chunks: int
+    current_chunk_index: int
+    status: str
+    started_at: Optional[datetime] = None
+    ended_at: Optional[datetime] = None
+    last_chunk_at: Optional[datetime] = None
+    eta_seconds: Optional[int] = None
+    chunks_remaining: Optional[int] = None
+    stop_requested: bool = False
+    error_message: Optional[str] = None
+    translated_text: str = ""
+    translated_chunks: List[str] = Field(default_factory=list)
+    final_document_path: Optional[str] = None

backend/app/processors/__init__.py

@@ -0,0 +1 @@
+"""Collects processor entrypoints for EPUB/PDF/DOCX/TXT."""

backend/app/processors/docx_processor.py

@@ -0,0 +1,14 @@
+from pathlib import Path
+
+from docx import Document
+
+from backend.app.processors.models import DocumentExtraction
+from backend.app.processors.utils import normalize_paragraphs
+
+
+def process_docx(path: Path) -> DocumentExtraction:
+    doc = Document(path)
+    title = doc.core_properties.title or None
+    raw_text = "\n\n".join(p.text for p in doc.paragraphs if p.text.strip())
+    paragraphs = normalize_paragraphs(raw_text)
+    return DocumentExtraction(title=title, paragraphs=paragraphs)

backend/app/processors/epub_processor.py

@@ -0,0 +1,28 @@
+from html import unescape
+from pathlib import Path
+import re
+
+from ebooklib import ITEM_DOCUMENT, epub
+
+from backend.app.processors.models import DocumentExtraction
+from backend.app.processors.utils import normalize_paragraphs
+
+HTML_TAG_RE = re.compile(r"<[^>]+>")
+
+
+def _strip_tags(source: str) -> str:
+    return unescape(HTML_TAG_RE.sub("", source))
+
+
+def process_epub(path: Path) -> DocumentExtraction:
+    book = epub.read_epub(path)
+    title_meta = book.get_metadata("DC", "title")
+    title = title_meta[0][0] if title_meta else None
+
+    paragraphs: list[str] = []
+    for item in book.get_items_of_type(ITEM_DOCUMENT):
+        html = item.get_content().decode("utf-8", errors="ignore")
+        text = _strip_tags(html)
+        paragraphs.extend(normalize_paragraphs(text))
+
+    return DocumentExtraction(title=title, paragraphs=paragraphs)

backend/app/processors/models.py

@@ -0,0 +1,15 @@
+from dataclasses import dataclass
+from typing import List, Optional
+
+@dataclass
+class DocumentExtraction:
+    title: Optional[str]
+    paragraphs: List[str]
+    source_language: Optional[str] = None
+
+@dataclass
+class DocumentChunk:
+    index: int
+    text: str
+    title_hint: Optional[str]
+    paragraph_count: int

backend/app/processors/pdf_processor.py

@@ -0,0 +1,40 @@
+from pathlib import Path
+from typing import List
+
+import fitz
+from pdf2image import convert_from_path
+import pytesseract
+
+from backend.app.processors.models import DocumentExtraction
+from backend.app.processors.utils import normalize_paragraphs
+
+
+def process_pdf(path: Path) -> DocumentExtraction:
+    paragraphs = _extract_text_pages(path)
+    if not paragraphs:
+        paragraphs = _extract_via_ocr(path)
+    return DocumentExtraction(title=None, paragraphs=paragraphs)
+
+
+def _extract_text_pages(path: Path) -> List[str]:
+    doc = fitz.open(path)
+    collected = []
+    for page in doc:
+        text = page.get_text("text")
+        if text.strip():
+            collected.extend(normalize_paragraphs(text))
+    return collected
+
+
+def _extract_via_ocr(path: Path) -> List[str]:
+    try:
+        images = convert_from_path(str(path), dpi=200)
+    except Exception:
+        return []
+
+    paragraphs = []
+    for image in images:
+        text = pytesseract.image_to_string(image)
+        if text.strip():
+            paragraphs.extend(normalize_paragraphs(text))
+    return paragraphs

backend/app/processors/registry.py

@@ -0,0 +1,22 @@
+from pathlib import Path
+from typing import Callable, Optional
+
+from backend.app.processors.models import DocumentExtraction
+from backend.app.processors.docx_processor import process_docx
+from backend.app.processors.epub_processor import process_epub
+from backend.app.processors.pdf_processor import process_pdf
+from backend.app.processors.txt_processor import process_txt
+
+ProcessorFn = Callable[[Path], DocumentExtraction]
+
+PROCESSORS: dict[str, ProcessorFn] = {
+    ".txt": process_txt,
+    ".docx": process_docx,
+    ".pdf": process_pdf,
+    ".epub": process_epub,
+}
+
+
+def select_processor(path: Path) -> Optional[ProcessorFn]:
+    suffix = path.suffix.lower()
+    return PROCESSORS.get(suffix)

backend/app/processors/txt_processor.py

@@ -0,0 +1,10 @@
+from pathlib import Path
+
+from backend.app.processors.models import DocumentExtraction
+from backend.app.processors.utils import normalize_paragraphs
+
+
+def process_txt(path: Path) -> DocumentExtraction:
+    text = path.read_text(encoding="utf-8", errors="ignore")
+    paragraphs = normalize_paragraphs(text)
+    return DocumentExtraction(title=path.stem, paragraphs=paragraphs)

backend/app/processors/utils.py

@@ -0,0 +1,18 @@
+import re
+from typing import List
+
+PARAGRAPH_SPLIT_RE = re.compile(r"\n{2,}")
+
+
+def normalize_paragraphs(source: str) -> List[str]:
+    text = source.replace("\r\n", "\n").strip()
+    if not text:
+        return []
+
+    paragraphs = [p.strip() for p in PARAGRAPH_SPLIT_RE.split(text)]
+    normalized = [p for p in paragraphs if p]
+    return normalized
+
+
+def merge_paragraphs(paragraphs: List[str]) -> str:
+    return "\n\n".join(paragraphs)

backend/app/routers/__init__.py

@@ -0,0 +1 @@
+from . import health, translation

backend/app/routers/health.py

@@ -0,0 +1,7 @@
+from fastapi import APIRouter
+
+router = APIRouter()
+
+@router.get("/health")
+async def health_check():
+    return {"status": "ok", "service": "traduttore"}

backend/app/routers/translation.py

@@ -0,0 +1,223 @@
+import asyncio
+import json
+import logging
+import os
+import tempfile
+from datetime import datetime
+from pathlib import Path
+from typing import List
+from uuid import uuid4
+
+from fastapi import APIRouter, BackgroundTasks, File, Form, HTTPException, UploadFile
+from langdetect import DetectorFactory, detect
+from fastapi.responses import StreamingResponse
+
+from backend.app.models.session import SessionState
+from backend.app.processors.registry import select_processor
+from backend.app.services.chunking import chunk_document, chunk_store
+from backend.app.utils.logger import logger
+from backend.app.utils.merge import write_final_document
+from backend.app.utils.ollama_client import OllamaClient
+from backend.app.utils.session_store import session_store
+
+DetectorFactory.seed = 0
+router = APIRouter()
+OLLAMA_BASE_URL = os.getenv("OLLAMA_URL", "https://ollama.solarpi.net")
+OLLAMA_MODEL = os.getenv("OLLAMA_MODEL", "translategemma:latest")
+ollama_client = OllamaClient(OLLAMA_BASE_URL, OLLAMA_MODEL)
+
+
+def _detect_language(paragraphs: List[str]) -> str:
+    snippet = " ".join(paragraphs[:5])
+    if not snippet.strip():
+        return "auto"
+    try:
+        return detect(snippet)
+    except Exception:
+        return "auto"
+
+
+async def _save_upload(upload: UploadFile) -> Path:
+    suffix = Path(upload.filename or "").suffix or ".bin"
+    with tempfile.NamedTemporaryFile(delete=False, suffix=suffix) as tmp:
+        payload = await upload.read()
+        tmp.write(payload)
+        tmp.flush()
+        return Path(tmp.name)
+
+
+def _build_session(session_id: str, file_title: str, extraction_title: str, source_lang: str, target_lang: str, chunk_count: int) -> SessionState:
+    return SessionState(
+        session_id=session_id,
+        file_title=file_title,
+        text_title=extraction_title or file_title,
+        source_lang=source_lang,
+        target_lang=target_lang,
+        total_chunks=chunk_count,
+        current_chunk_index=0,
+        status="pending",
+        chunks_remaining=chunk_count,
+    )
+
+
+async def _run_translation(session_id: str) -> None:
+    session = session_store.get(session_id)
+    if not session:
+        return
+
+    session.status = "running"
+    session.started_at = datetime.utcnow()
+    session_store.update(session)
+    logger.info("Sessione %s: avvio traduzione", session_id)
+
+    chunks = chunk_store.get_chunks(session_id)
+    durations: List[float] = []
+
+    for chunk in chunks:
+        session = session_store.get(session_id)
+        if not session:
+            return
+        if session.stop_requested:
+            session.status = "stopped"
+            session_store.update(session)
+            logger.info("Sessione %s: interrotta dall'utente", session_id)
+            return
+
+        start = datetime.utcnow()
+        try:
+            translation = await ollama_client.translate(chunk.text, session.source_lang, session.target_lang)
+        except Exception as exc:
+            session.status = "error"
+            session.error_message = str(exc)
+            session_store.update(session)
+            logger.error("Sessione %s chunk %s fallito: %s", session_id, chunk.index, exc)
+            return
+
+        now = datetime.utcnow()
+        duration = (now - start).total_seconds()
+        durations.append(duration)
+        if len(durations) > 3:
+            durations.pop(0)
+
+        session.translated_chunks.append(translation)
+        session.translated_text = "\n\n".join(session.translated_chunks)
+        session.current_chunk_index = chunk.index
+        session.chunks_remaining = max(0, session.total_chunks - chunk.index)
+        session.last_chunk_at = now
+        avg = sum(durations) / len(durations) if durations else 0
+        session.eta_seconds = int(avg * session.chunks_remaining) if session.chunks_remaining else 0
+        session_store.update(session)
+        logger.info("Sessione %s: chunk %s tradotto", session_id, chunk.index)
+
+    session.status = "completed"
+    session.ended_at = datetime.utcnow()
+    session.chunks_remaining = 0
+    session.eta_seconds = 0
+    session.final_document_path = str(write_final_document(session.session_id, session.translated_text))
+    logger.info("Sessione %s completata. Documento finale: %s", session_id, session.final_document_path)
+    session_store.update(session)
+
+
+@router.post("/upload")
+async def upload_document(target_lang: str = Form(...), file: UploadFile = File(...)):
+    processor = select_processor(Path(file.filename or ""))
+    if not processor:
+        raise HTTPException(status_code=400, detail="Formato file non supportato")
+
+    temp_path = await _save_upload(file)
+    try:
+        extraction = processor(temp_path)
+    finally:
+        temp_path.unlink(missing_ok=True)
+
+    language = extraction.source_language or _detect_language(extraction.paragraphs)
+    session_id = uuid4().hex
+    chunks = chunk_document(session_id, extraction)
+    logger.info("Sessione %s: %s chunk generati", session_id, len(chunks))
+
+    session = _build_session(
+        session_id,
+        file.filename or "sconosciuto",
+        extraction.title,
+        language,
+        target_lang,
+        len(chunks),
+    )
+    session_store.create(session)
+
+    return {
+        "session_id": session.session_id,
+        "total_chunks": session.total_chunks,
+        "source_lang": session.source_lang,
+    }
+
+
+@router.post("/start")
+async def start_translation(session_id: str, background: BackgroundTasks):
+    session = session_store.get(session_id)
+    if not session:
+        raise HTTPException(status_code=404, detail="Sessione non trovata")
+
+    if session.status == "running":
+        return {"message": "Traduzione già in corso", "session_id": session_id}
+
+    session.status = "running"
+    session.stop_requested = False
+    session_store.update(session)
+    logger.info("Sessione %s: start richiesta", session_id)
+    background.add_task(_run_translation, session_id)
+
+    return {"message": "Traduzione avviata", "session_id": session_id}
+
+
+@router.post("/stop")
+async def stop_translation(session_id: str):
+    session = session_store.get(session_id)
+    if not session:
+        raise HTTPException(status_code=404, detail="Sessione non trovata")
+
+    session.stop_requested = True
+    session.status = "stopping"
+    session_store.update(session)
+    logger.info("Sessione %s: stop richiesto", session_id)
+    return {"message": "Stop richiesto", "session_id": session_id}
+
+
+@router.get("/status/{session_id}")
+async def translation_status(session_id: str):
+    session = session_store.get(session_id)
+    if not session:
+        raise HTTPException(status_code=404, detail="Sessione non trovata")
+    return session
+
+
+async def _sse_generator(session_id: str):
+    while True:
+        session = session_store.get(session_id)
+        if not session:
+            break
+        payload = {
+            "status": session.status,
+            "current_chunk": session.current_chunk_index,
+            "total_chunks": session.total_chunks,
+            "chunks_remaining": session.chunks_remaining,
+            "eta_seconds": session.eta_seconds,
+            "translated_text": session.translated_text,
+            "error_message": session.error_message,
+            "file_title": session.file_title,
+            "text_title": session.text_title,
+            "started_at": session.started_at.isoformat() if session.started_at else None,
+            "ended_at": session.ended_at.isoformat() if session.ended_at else None,
+        }
+        yield f"data: {json.dumps(payload)}\n\n"
+        if session.status in {"completed", "error", "stopped"}:
+            break
+        await asyncio.sleep(1)
+
+
+@router.get("/stream/{session_id}")
+async def translation_stream(session_id: str):
+    if not session_store.get(session_id):
+        raise HTTPException(status_code=404, detail="Sessione non trovata")
+    headers = {"Cache-Control": "no-cache", "X-Accel-Buffering": "no"}
+    return StreamingResponse(_sse_generator(session_id), media_type="text/event-stream", headers=headers)

backend/app/services/__init__.py

@@ -0,0 +1 @@
+from .chunking import chunk_document, chunk_store

backend/app/services/chunking.py

@@ -0,0 +1,87 @@
+from dataclasses import dataclass
+from datetime import datetime
+from typing import Dict, List, Optional
+
+from backend.app.processors.models import DocumentChunk, DocumentExtraction
+
+
+@dataclass
+class SessionChunks:
+    extraction: DocumentExtraction
+    chunks: List[DocumentChunk]
+    created_at: datetime
+
+
+class ChunkStore:
+    def __init__(self):
+        self._sessions: Dict[str, SessionChunks] = {}
+
+    def register(self, session_id: str, extraction: DocumentExtraction, chunks: List[DocumentChunk]) -> None:
+        self._sessions[session_id] = SessionChunks(extraction, chunks, datetime.utcnow())
+
+    def get_chunks(self, session_id: str) -> List[DocumentChunk]:
+        session = self._sessions.get(session_id)
+        return session.chunks if session else []
+
+    def get_session(self, session_id: str) -> Optional[SessionChunks]:
+        return self._sessions.get(session_id)
+
+    def remove(self, session_id: str) -> None:
+        self._sessions.pop(session_id, None)
+
+
+chunk_store = ChunkStore()
+
+
+def _looks_like_title(paragraph: str) -> bool:
+    stripped = paragraph.strip()
+    if not stripped:
+        return False
+    if stripped.startswith("#"):
+        return True
+    if stripped.endswith(":") and len(stripped) < 60:
+        return True
+    if stripped.isupper() and len(stripped) < 120:
+        return True
+    return False
+
+
+def chunk_document(session_id: str, extraction: DocumentExtraction, chunk_size: int = 2_500) -> List[DocumentChunk]:
+    paragraphs = [p.strip() for p in extraction.paragraphs if p.strip()]
+    if not paragraphs:
+        chunk_store.register(session_id, extraction, [])
+        return []
+
+    chunks: List[DocumentChunk] = []
+    buffer: List[str] = []
+    buffer_len = 0
+    current_title = extraction.title
+
+    def flush_chunk() -> None:
+        nonlocal buffer, buffer_len, current_title
+        if not buffer:
+            return
+        chunk_text = "\n\n".join(buffer)
+        chunks.append(
+            DocumentChunk(
+                index=len(chunks) + 1,
+                text=chunk_text,
+                title_hint=current_title,
+                paragraph_count=len(buffer),
+            )
+        )
+        buffer = []
+        buffer_len = 0
+
+    for paragraph in paragraphs:
+        if _looks_like_title(paragraph):
+            current_title = paragraph
+        addition_len = len(paragraph)
+        if buffer and buffer_len + addition_len + 2 > chunk_size:
+            flush_chunk()
+        buffer.append(paragraph)
+        buffer_len += addition_len + 2
+
+    flush_chunk()
+    chunk_store.register(session_id, extraction, chunks)
+    return chunks

backend/app/utils/__init__.py

@@ -0,0 +1 @@
+"""Helpers for ETA, config, and state tracking will live here."""

backend/app/utils/logger.py

@@ -0,0 +1,15 @@
+import logging
+from pathlib import Path
+
+LOG_FILE = Path("backend.log")
+
+logging.basicConfig(
+    level=logging.INFO,
+    format="%(asctime)s %(levelname)s %(name)s %(message)s",
+    handlers=[
+        logging.StreamHandler(),
+        logging.FileHandler(LOG_FILE, encoding="utf-8"),
+    ],
+)
+
+logger = logging.getLogger("traduttore")

backend/app/utils/merge.py

@@ -0,0 +1,10 @@
+from pathlib import Path
+
+OUTPUT_BASE = Path("_bmad-output").resolve()
+OUTPUT_BASE.mkdir(exist_ok=True)
+
+
+def write_final_document(session_id: str, content: str) -> Path:
+    target = OUTPUT_BASE / f"translated-{session_id}.txt"
+    target.write_text(content, encoding="utf-8")
+    return target

backend/app/utils/ollama_client.py

@@ -0,0 +1,37 @@
+from typing import Optional
+
+import httpx
+
+
+def _build_prompt(source_lang: str, target_lang: str, text: str) -> str:
+    return (
+        f"Traduci dal {source_lang} al {target_lang}:\\n" f"{text}"
+    )
+
+
+class OllamaClient:
+    def __init__(self, base_url: str, model: str):
+        self._base_url = base_url.rstrip("/")
+        self._model = model
+        self._client = httpx.AsyncClient(base_url=self._base_url, timeout=60.0)
+
+    async def translate(self, text: str, source_lang: str, target_lang: str) -> str:
+        payload = {
+            "model": self._model,
+            "prompt": _build_prompt(source_lang or "auto", target_lang, text),
+            "stream": False,
+        }
+        try:
+            resp = await self._client.post("/api/chat", json=payload)
+            resp.raise_for_status()
+            body = resp.json()
+        except httpx.HTTPError as exc:
+            raise RuntimeError(f"Errore Ollama: {exc}") from exc
+
+        if isinstance(body, dict):
+            message = body.get("response") or ""
+            return message.strip()
+        return ""
+
+    async def close(self) -> None:
+        await self._client.aclose()

backend/app/utils/session_store.py

@@ -0,0 +1,23 @@
+from typing import Dict, Optional
+
+from backend.app.models.session import SessionState
+
+
+class SessionStore:
+    def __init__(self):
+        self._sessions: Dict[str, SessionState] = {}
+
+    def create(self, state: SessionState) -> None:
+        self._sessions[state.session_id] = state
+
+    def get(self, session_id: str) -> Optional[SessionState]:
+        return self._sessions.get(session_id)
+
+    def update(self, state: SessionState) -> None:
+        self._sessions[state.session_id] = state
+
+    def remove(self, session_id: str) -> None:
+        self._sessions.pop(session_id, None)
+
+
+session_store = SessionStore()

backend/requirements.txt

@@ -0,0 +1,11 @@
+fastapi
+uvicorn[standard]
+pydantic
+httpx
+python-multipart
+langdetect
+python-docx
+PyMuPDF
+pdf2image
+pytesseract
+ebooklib

backend/tests/test_health.py

@@ -0,0 +1,9 @@
+from fastapi.testclient import TestClient
+from backend.app.main import app
+
+client = TestClient(app)
+
+def test_health_check():
+    resp = client.get("/health")
+    assert resp.status_code == 200
+    assert resp.json().get("status") == "ok"

backend/tests/test_processors.py

@@ -0,0 +1,69 @@
+import json
+from pathlib import Path
+from typing import Callable
+
+import fitz
+from ebooklib import epub
+from docx import Document
+
+from backend.app.processors.registry import select_processor
+from backend.app.services.chunking import chunk_document
+
+
+def _create_docx(path: Path) -> Path:
+    doc = Document()
+    doc.add_heading("Titolo principale", level=1)
+    doc.add_paragraph("Primo paragrafo del documento.")
+    doc.add_paragraph("Secondo paragrafo con alcune frasi.")
+    doc.save(path)
+    return path
+
+
+def _create_pdf(path: Path) -> Path:
+    doc = fitz.open()
+    page = doc.new_page()
+    page.insert_text((72, 72), "Titolo PDF\n\nParagrafo PDF di prova.")
+    doc.save(path)
+    doc.close()
+    return path
+
+
+def _create_epub(path: Path) -> Path:
+    book = epub.EpubBook()
+    book.set_identifier("test-123")
+    book.set_title("EPUB di prova")
+    book.set_language("it")
+
+    chapter = epub.EpubHtml(title="Capitolo", file_name="chap.xhtml")
+    chapter.content = "<h1>Capitolo</h1><p>Paragrafo EPUB.</p>"
+    book.add_item(chapter)
+    book.toc = (epub.Link("chap.xhtml", "Capitolo", "chap"),)
+    book.add_item(epub.EpubNcx())
+    book.add_item(epub.EpubNav())
+    epub.write_epub(str(path), book)
+    return path
+
+
+def _create_txt(path: Path) -> Path:
+    path.write_text("Titolo TXT\n\nParagrafo 1\n\nParagrafo 2", encoding="utf-8")
+    return path
+
+
+generator_map: dict[str, Callable[[Path], Path]] = {
+    ".docx": _create_docx,
+    ".pdf": _create_pdf,
+    ".epub": _create_epub,
+    ".txt": _create_txt,
+}
+
+
+def test_processor_chain(tmp_path: Path):
+    for suffix, generator in generator_map.items():
+        target = tmp_path / f"sample{suffix}"
+        path = generator(target)
+        processor = select_processor(path)
+        assert processor is not None, f"Processor missing for {suffix}"
+        extraction = processor(path)
+        chunks = chunk_document(f"session-{suffix}", extraction)
+        assert chunks, f"Nessun chunk per {suffix}"
+        assert all(chunk.text.strip() for chunk in chunks)