"""Streamlit UI: streamlit run mrag/streamlit_app.py""" from __future__ import annotations import sys import tempfile from pathlib import Path import streamlit as st sys.path.insert(0, str(Path(__file__).resolve().parent.parent)) from mrag.config import Config # noqa: E402 from mrag.pipeline import MultimodalRAG # noqa: E402 st.set_page_config(page_title="Multimodal RAG", layout="wide") @st.cache_resource def get_rag() -> MultimodalRAG: # one instance per server: Qdrant embedded mode allows a single client return MultimodalRAG(Config.load()) rag = get_rag() cfg = rag.cfg with st.sidebar: st.header("Index") files = st.file_uploader("Add documents", accept_multiple_files=True, type=["pdf", "docx", "png", "jpg", "jpeg", "webp", "bmp", "gif", "tif", "tiff", "txt", "md"]) if st.button("Ingest", disabled=not files): upload_dir = Path(cfg.data_dir) / "uploads" upload_dir.mkdir(parents=True, exist_ok=True) paths = [] for f in files: p = upload_dir / Path(f.name).name p.write_bytes(f.getbuffer()) paths.append(p) with st.spinner("Parsing, OCR, embedding..."): s = rag.ingest(paths) st.success(f"{s['files']} file(s): {s['text_chunks']} text chunks, {s['images']} images, " f"{s['ocr_chunks']} OCR chunks") for e in s["errors"]: st.error(e) st.caption(f"Index: {rag.stats()}") st.header("Generator") cfg.llm_backend = st.selectbox("Backend", ["ollama", "openai", "extractive"], index=["ollama", "openai", "extractive"].index(cfg.llm_backend)) if cfg.llm_backend != "extractive": cfg.llm_model = st.text_input("Model", cfg.llm_model) cfg.llm_url = st.text_input("Server URL", cfg.llm_url) cfg.llm_vision = st.checkbox("Send retrieved images to the model", cfg.llm_vision) st.title("Multimodal RAG") q = st.text_input("Ask about your documents") if q: try: res = rag.ask(q, stream=True) st.write_stream(res["answer"]) except Exception as e: st.error(str(e)) texts, images = rag.retrieve(q) res = {"texts": texts, "images": images} if res["images"]: st.subheader("Retrieved images") cols = st.columns(len(res["images"])) for i, (c, im) in enumerate(zip(cols, res["images"]), 1): c.image(im.path, caption=f"[Img {i}] {Path(im.source).name} p.{im.page} {', '.join(im.why)}") with st.expander("Retrieved text"): for i, t in enumerate(res["texts"], 1): st.markdown(f"**[{i}] {Path(t.source).name} p.{t.page}** ({t.kind})") st.write(t.text)