5 sources (#26)

* changed back to 5 sources

* removed number of chunks

* ran black

app.py

@@ -75,20 +75,20 @@ def format_sources(matched_documents: pd.DataFrame) -> str:
         return ""
 
     documents_answer_template: str = "📝 Here are the sources I used to answer your question:\n\n{documents}\n\n{footnote}"
-    document_template: str = "[🔗 {document.source}: {document.title}]({document.url}), highest relevance: {document.similarity_to_answer:2.1f} % | # total chunks matched: {document.repetition:d}"
+    document_template: str = "[🔗 {document.source}: {document.title}]({document.url}), highest relevance: {document.similarity_to_answer:2.1f} %"  # | # total chunks matched: {document.repetition:d}"
 
     matched_documents.similarity_to_answer = (
         matched_documents.similarity_to_answer * 100
     )
 
-    matched_documents["repetition"] = matched_documents.groupby("title")[
-        "title"
-    ].transform("size")
+    # matched_documents["repetition"] = matched_documents.groupby("title")[
+    #     "title"
+    # ].transform("size")
 
     # drop duplicates, keep highest ranking ones
     matched_documents = matched_documents.sort_values(
         "similarity_to_answer", ascending=False
-    ).drop_duplicates("title", keep="first")[:5]
+    ).drop_duplicates("title", keep="first")
 
     # Revert back to correct display
     display_source_to_ui = {

cfg.py

@@ -73,7 +73,7 @@ Q:
     },
     retriever_cfg={
         "path": f"{DEEPLAKE_DATASET_PATH}",
-        "top_k": 10,
+        "top_k": 5,
         "thresh": 0.55,
         "max_tokens": 13000,
         "embedding_model": "text-embedding-ada-002",