3 files changed, 199 insertions, 0 deletions
diff --git a/src/umapyai/__init__.py b/src/umapyai/__init__.py
new file mode 100644
index 0000000..4898488
--- /dev/null
+++ b/src/umapyai/__init__.py
@@ -0,0 +1,187 @@
+import os
+import sys
+import time
+import chromadb
+from sentence_transformers import SentenceTransformer
+import requests
+import subprocess
+import psutil
+from .constants import ARTICLES_DIRECTORY, CHROMA_DIRECTORY, CHROMA_COLLECTION, CHUNK_SIZE, EMBEDDING_MODEL, OLLAMA_MODEL, TOP_K, OLLAMA_URL
+
+
+def is_ollama_live():
+  try:
+    response = requests.get(f"{OLLAMA_URL}/api/tags", timeout=2)
+
+    return response.status_code == 200
+  except Exception:
+    return False
+
+
+def start_ollama_server():
+  print("Starting Ollama server with OLLAMA_ORIGINS='*' ...")
+
+  environment = os.environ.copy()
+  environment["OLLAMA_ORIGINS"] = "*"
+  process = subprocess.Popen(["ollama", "serve"],
+                             env=environment,
+                             stdout=subprocess.PIPE,
+                             stderr=subprocess.STDOUT,
+                             text=True)
+
+  for _ in range(30):
+    if is_ollama_live():
+      print("Ollama is now live.")
+
+      return process
+
+    time.sleep(1)
+
+  print("ERROR: Ollama server did not start after 30 seconds.")
+  process.terminate()
+  sys.exit(1)
+
+
+def kill_ollama(process):
+  print("Killing Ollama ...")
+
+  try:
+    parent_process = psutil.Process(process.pid)
+
+    for child_process in parent_process.children(recursive=True):
+      child_process.terminate()
+
+    parent_process.terminate()
+  except Exception as error:
+    print("Error killing Ollama:", error)
+
+
+def ensure_model_pulled(model):
+  try:
+    tags = requests.get(f"{OLLAMA_URL}/api/tags").json().get("models", [])
+
+    if not any(model in m.get("name", "") for m in tags):
+      print(f"Pulling model '{model}' ...")
+      subprocess.run(["ollama", "pull", model], check=True)
+    else:
+      print(f"Model '{model}' already pulled.")
+  except Exception as e:
+    print("Couldn't check/pull Ollama model:", e)
+    print("Proceeding anyway ...")
+
+
+def main():
+  ollama_process = None
+  started_ollama = False
+
+  try:
+    if not is_ollama_live():
+      ollama_process = start_ollama_server()
+      started_ollama = True
+    else:
+      print("Ollama is already running.")
+
+    ensure_model_pulled(OLLAMA_MODEL)
+
+    print("Chunking articles ...")
+
+    chunks = []
+
+    for file_name in os.listdir(ARTICLES_DIRECTORY):
+      if not file_name.endswith(".txt"):
+        continue
+
+      with open(
+          os.path.join(ARTICLES_DIRECTORY, file_name),
+          encoding="utf-8") as file:
+        words = file.read().split()
+
+      for i in range(0, len(words), CHUNK_SIZE):
+        chunk = " ".join(words[i:i + CHUNK_SIZE])
+
+        if chunk.strip():
+          chunks.append({"source": file_name, "chunk": chunk})
+
+    print(f"Total chunks: {len(chunks)}")
+    print("Generating embeddings ...")
+
+    model = SentenceTransformer(EMBEDDING_MODEL)
+
+    for chunk in chunks:
+      chunk["embedding"] = model.encode(chunk["chunk"])
+
+    print("Storing embeddings in ChromaDB ...")
+
+    chroma_client = chromadb.PersistentClient(path=CHROMA_DIRECTORY)
+
+    if CHROMA_COLLECTION in [
+        collection.name for collection in chroma_client.list_collections()
+    ]:
+      print("Collection exists, deleting and recreating for fresh import ...")
+
+      chroma_client.delete_collection(CHROMA_COLLECTION)
+
+    collection = chroma_client.get_or_create_collection(CHROMA_COLLECTION)
+
+    for i, chunk in enumerate(chunks):
+      collection.add(
+          ids=[str(i)],
+          documents=[chunk["chunk"]],
+          embeddings=[chunk["embedding"].tolist()],
+          metadatas=[{
+              "source": chunk["source"]
+          }])
+
+    def find_relevant_chunks(query, top_k=TOP_K):
+      q_embed = model.encode(query)
+      results = collection.query(
+          query_embeddings=[q_embed.tolist()], n_results=top_k)
+      documents = results['documents'][0]
+      metadatas = results['metadatas'][0]
+
+      return [(document, metadata)
+              for document, metadata in zip(documents, metadatas)]
+
+    def query_ollama(prompt):
+      url = f"{OLLAMA_URL}/api/generate"
+      payload = {
+          "model": OLLAMA_MODEL,
+          "prompt": prompt,
+          "stream": False,
+      }
+
+      try:
+        response = requests.post(url, json=payload)
+
+        response.raise_for_status()
+
+        return response.json().get('response', '').strip()
+      except Exception as error:
+        return f"Error communicating with Ollama: {error}"
+
+    print("\nReady! Ask your Uma Musume build questions (type 'exit' to quit):")
+
+    while True:
+      user_query = input("\n> ")
+
+      if user_query.strip().lower() == "exit":
+        break
+
+      top_chunks = find_relevant_chunks(user_query)
+      context = "\n\n".join([c[0] for c in top_chunks])
+      full_prompt = (
+          "You are an expert Uma Musume: Pretty Derby build guide advisor.\n"
+          "Answer the user's question using ONLY the following context. "
+          "If the answer isn't in the context, say you don't know.\n\n"
+          f"Context:\n{context}\n\n"
+          f"Question: {user_query}\nAnswer:")
+      answer = query_ollama(full_prompt)
+
+      print("\n", answer)
+      print(
+          "\nSources:", ", ".join(
+              sorted(set(metadata['source'] for _, metadata in top_chunks))))
+  finally:
+    if started_ollama and ollama_process is not None:
+      kill_ollama(ollama_process)
+      print("Ollama server stopped.")
diff --git a/src/umapyai/__main__.py b/src/umapyai/__main__.py
new file mode 100644
index 0000000..851db54
--- /dev/null
+++ b/src/umapyai/__main__.py
@@ -0,0 +1,4 @@
+import umapyai
+import sys
+
+sys.exit(umapyai.main())
diff --git a/src/umapyai/constants.py b/src/umapyai/constants.py
new file mode 100644
index 0000000..9add8ce
--- /dev/null
+++ b/src/umapyai/constants.py
@@ -0,0 +1,8 @@
+ARTICLES_DIRECTORY = "uma_articles_clean"
+CHROMA_DIRECTORY = "./chromadb"
+CHROMA_COLLECTION = "uma_guides"
+CHUNK_SIZE = 350  # words
+EMBEDDING_MODEL = "all-MiniLM-L6-v2"
+OLLAMA_MODEL = "llama3.2"
+TOP_K = 4
+OLLAMA_URL = "http://localhost:11434"