Initial Ulimi AI space with extenarnal pdf dataset

app.py

@@ -0,0 +1,28 @@
+import os
+import streamlit as st
+from huggingface_hub import snapshot_download
+from langchain.document_loaders import PyPDFLoader
+from langchain.text_splitter import RecursiveCharacterTextSplitter
+
+st.set_page_config(page_title="Ulimi AI", layout="wide")
+st.title("🌽 Ulimi AI – Agricultural Intelligence for Malawi")
+
+DOCS_DIR = "docs"
+
+if not os.path.exists(DOCS_DIR):
+    with st.spinner("Downloading knowledge base..."):
+        snapshot_download(
+            repo_id="MicohEscobar/Ulimi-AI-Docs",
+            repo_type="dataset",
+            local_dir=DOCS_DIR
+        )
+
+pdfs = []
+for file in os.listdir(DOCS_DIR):
+    if file.endswith(".pdf"):
+        loader = PyPDFLoader(os.path.join(DOCS_DIR, file))
+        pdfs.extend(loader.load())
+
+st.success(f"Loaded {len(pdfs)} document pages")
+
+st.write("RAG pipeline ready. Embeddings + QA next.")

requirements.txt

@@ -0,0 +1,7 @@
+streamlit
+langchain
+langchain-community
+pypdf
+sentence-transformers
+faiss-cpu
+huggingface_hub