add search title

src/mcp/server.py

@@ -9,6 +9,13 @@ from dotenv import load_dotenv
 import os
 
 
+from sentence_transformers import SentenceTransformer
+import faiss
+import numpy as np
+from src.configs.config import TITLE_FAISS_INDEX_FILE, TITLE_EMBEDDINGS_FILE, EMBEDDING_MODEL, METADATA_FILE
+import pandas as pd
+
+
 LOG_FILE = os.path.join(LOG_DIR, "Agents.log")
 logging.basicConfig(
     filename=LOG_FILE,
@@ -123,6 +130,39 @@ async def summarize_video_by_link(link: str):
     })
     return prompt
 
+@mcp.prompt(
+    name="summarize_doc_by_title",
+    description="Summarize a document using its title"
+)
+async def summarize_doc_by_title(title_query: str):
+    result = search_title(title_query)
+    link = result["link"]
+    return await summarize_doc_by_link(link)
+
+@mcp.prompt(
+    name="summarize_document",
+    description="Summarize a document by link or title. If the query contains 'https', it is treated as a link."
+)
+async def summarize_document(query: str):
+    if "https" in query:
+        return await summarize_doc_by_link(query)
+    else:
+        return await summarize_doc_by_title(query)
+# search_title 
+def search_title(query):
+    # Load index and metadata
+    index = faiss.read_index(str(TITLE_FAISS_INDEX_FILE))
+    model = SentenceTransformer(EMBEDDING_MODEL)
+    metadata = pd.read_csv(METADATA_FILE)
+    titles = metadata["Nom du document"].tolist()
+    links = metadata["Lien"].tolist()
+    # Embed query
+    query_vec = model.encode([query], convert_to_tensor=False)
+    D, I = index.search(np.array(query_vec), k=1)
+    best_idx = I[0][0]
+    return {"title": titles[best_idx], "link": links[best_idx]}
+
+
 def sample_chunks(chunks, n=3):
     """Pick N evenly spaced chunks across the whole video."""