Aissugur Kurganbekov
RU/EN bilingual fixes from adversarial audit: tokenizer, numbers, countries, weights, HS guard, multi-item sets, currencies, OCR langs, xlsx intake, sample-doc isolation
c86642c | # Single-container image: FastAPI serves BOTH the built React app and the /api | |
| # backend on one port. Suited to Hugging Face Spaces and any one-port host. | |
| # docker build -t financeflow . | |
| # docker run -p 7860:7860 financeflow -> http://localhost:7860 | |
| # 1) Build the React/Vite frontend into static files. | |
| FROM node:20-alpine AS frontend | |
| WORKDIR /fe | |
| COPY frontend/package*.json ./ | |
| RUN npm ci | |
| COPY frontend/ ./ | |
| RUN npm run build | |
| # 2) Python backend that also serves those static files. | |
| FROM python:3.12-slim | |
| WORKDIR /app | |
| # System packages: tesseract-ocr powers the optional OCR fallback for scanned / | |
| # image-only PDFs (parsing.py). Without it the app still runs — OCR just no-ops — | |
| # but installing it lets scanned uploads be read. PyMuPDF/pytesseract come via pip. | |
| RUN apt-get update \ | |
| && apt-get install -y --no-install-recommends \ | |
| tesseract-ocr tesseract-ocr-rus tesseract-ocr-kaz tesseract-ocr-chi-sim \ | |
| && rm -rf /var/lib/apt/lists/* | |
| COPY backend/requirements.txt ./requirements.txt | |
| # requirements.txt already pins flashrank (the reranker) and fastembed (the dense | |
| # embeddings for hybrid retrieval). Add the optional LLM SDKs too so "Thinking" | |
| # mode works when a key is provided. | |
| RUN pip install --no-cache-dir -r requirements.txt openai anthropic | |
| COPY backend/app ./app | |
| COPY sample_docs /sample_docs | |
| COPY --from=frontend /fe/dist ./static | |
| # Hugging Face Spaces runs the container as UID 1000; create that user and give | |
| # it ownership of the writable paths (the SQLite dir + sample docs). | |
| RUN useradd -m -u 1000 user \ | |
| && mkdir -p /app/data \ | |
| && chown -R user:user /app /sample_docs | |
| USER user | |
| ENV FRONTEND_DIR=/app/static | |
| # Thinking mode via LLM7.io — a free, keyless OpenAI-compatible community proxy. | |
| # None of these values are secrets: the proxy needs no real key (any non-empty | |
| # string passes), so they can live in the image. For real/sensitive documents, | |
| # override these with Space secrets pointing at a first-party provider. | |
| ENV LLM_PROVIDER=openai \ | |
| OPENAI_BASE_URL=https://api.llm7.io/v1 \ | |
| OPENAI_MODEL=default \ | |
| OPENAI_API_KEY=unused | |
| EXPOSE 7860 | |
| CMD ["uvicorn", "app.server:root", "--host", "0.0.0.0", "--port", "7860"] | |