Update app.py

app.py

@@ -15,19 +15,24 @@ from streamlit_lottie import st_lottie
 st.set_page_config(page_title="QA-project", page_icon="📇")
 os.environ['TOKENIZERS_PARALLELISM'] = "false"
 DATA_DIR = './dataset'
+NAMES_DICT_PATH = 'mod_names_dict.pkl'
 DOCS_PATH = os.path.join(DATA_DIR, 'all_docs_36838.pkl')
 LOTTIE_PATH = './img/108423-search-for-documents.json'
-PROG_TITLE = "QA project Demo"
+PROG_TITLE = "Научные кейсы"
+PROG_SUBTITLE = "Рекомендации по существующим в компании компонентам цифровых продуктов для решения новых бизнес-задач"
+
+
+
 # Adjust to a question that you would like users to see in the search bar when they load the UI:
 DEFAULT_QUESTION_AT_STARTUP = os.getenv("DEFAULT_QUESTION_AT_STARTUP", "Что делает Домашняя бухгалтерия?")
 DEFAULT_ANSWER_AT_STARTUP = os.getenv("DEFAULT_ANSWER_AT_STARTUP", "Домашняя бухгалтерия позволяет вести счета в разных валютах")
 
-def place_header_center(text, lottie_data):
-    img, title= st.columns([1,3])
-    with img:
-        st_lottie(lottie_data, height=150)
-    with title:
-        st.title(text)
+
+@st.experimental_memo
+def load_dict(path):
+    with open(path, "rb") as f:
+        loaded = dill.load(f)
+    return loaded
 
 
 @st.experimental_memo
@@ -41,49 +46,23 @@ def load_and_write_data(document_store):
 
     with open(DOCS_PATH, "rb") as f:
         docs = dill.load(f)
-
+    
     document_store.write_documents(docs)
 
 
-def get_backlink(result):
-    if result.get("document", None):
-        doc = result["document"]
-        if isinstance(doc, dict):
-            if doc.get("meta", None):
-                if isinstance(doc["meta"], dict):
-                    if doc["meta"].get("url", None):
-                        return doc["meta"]["url"]
-    return None
-
-
-def get_doc_name(result):
-    if result.get("document", None):
-        doc = result["document"]
-        if isinstance(doc, dict):
-            if doc.get("meta", None):
-                if isinstance(doc["meta"], dict):
-                    if doc["meta"].get("name", None):
-                        return doc["meta"]["name"]
-    return None
-
 def get_doc_reg_id(result):
-    if result.get("document", None):
-        doc = result["document"]
-        if isinstance(doc, dict):
-            if doc.get("meta", None):
-                if isinstance(doc["meta"], dict):
-                    if doc["meta"].get("reg_id", None):
-                        return doc["meta"]["reg_id"]
+    if result.get("reg_id", None):
+        reg_id = result["reg_id"]
+        return reg_id 
     return None
+
+   
 # Haystack Components
-# @st.cache(allow_output_mutation=True)
-# def start_haystack():
 document_store = InMemoryDocumentStore() # use_bm25=True
 load_and_write_data(document_store)
 retriever = TfidfRetriever(document_store=document_store)
 reader = FARMReader(model_name_or_path="DeepPavlov/rubert-base-cased-sentence",
-                    use_gpu=False,
-                    num_processes=1)
+                    use_gpu=False)
 pipeline = ExtractiveQAPipeline(reader, retriever)
 
 
@@ -101,16 +80,23 @@ def reset_results(*args):
 
 # Streamlit App
 lottie_data = get_lottie(LOTTIE_PATH)
-place_header_center(PROG_TITLE, lottie_data)
+img, title= st.columns([2,3])
+with img:
+    st_lottie(lottie_data) #, height=350
+with title:        
+    st.title(PROG_TITLE)
+    st.subheader(PROG_SUBTITLE)
+
 
 st.markdown("""
-This QA demo uses a [Haystack Extractive QA Pipeline](https://haystack.deepset.ai/components/ready-made-pipelines#extractiveqapipeline) with 
-an [InMemoryDocumentStore](https://haystack.deepset.ai/components/document-store) which contains documents about different program modules
-Go ahead and ask questions about the program modules functionality!
+Это демонстрационная версия сервиса поисковой системы программных продуктов с использованием технологии
+ [Haystack Extractive QA Pipeline](https://haystack.deepset.ai/components/ready-made-pipelines#extractiveqapipeline) 
+и [InMemoryDocumentStore](https://haystack.deepset.ai/components/document-store)
+Чтобы испытать сервис можно задавать вопросы в свободной форме по функционалу программных продуктов.
 """, unsafe_allow_html=True)
 
 question = st.text_input("", value=st.session_state.question, max_chars=100, on_change=reset_results)
-
+mod_names_dict = load_dict(NAMES_DICT_PATH)
 
 def ask_question(question):
     prediction = pipeline.run(query=question, params={"Retriever": {"top_k": 10}, "Reader": {"top_k": 5}})
@@ -119,13 +105,16 @@ def ask_question(question):
     for answer in answers:
         answer = answer.to_dict()
         if answer.get("answer", None):
+            document = [doc for doc in prediction["documents"] if (doc.to_dict()["id"] == answer["document_id"])][0]
             results.append(
                 {
                     "context": "..." + answer["context"] + "...",
                     "answer": answer.get("answer", None),
-                    "source": answer["meta"]["name"],
+                    "source": answer["meta"]["name"] if answer["meta"].get("name", None) else answer["meta"]['url'],
                     "relevance": round(answer["score"] * 100, 2),
-                    "document": [doc for doc in response["documents"] if doc["id"] == answer["document_id"]][0],
+                    "document": document.content,
+                    "doc_score": document.score,
+                    "reg_id": document.meta["reg_id"],
                     "offset_start_in_doc": answer["offsets_in_document"][0]["start"],
                     "_raw": answer,
                 }
@@ -144,7 +133,7 @@ def ask_question(question):
 
 
 if question:
-    with st.spinner("🕰️    Performing semantic search on program modules..."):
+    with st.spinner("🕰️    Производится семантический поиск по информационной базе ..."):
         try:
             msg = 'Asked ' + question
             logging.info(msg)
@@ -154,34 +143,31 @@ if question:
 
 
 if st.session_state.results:
-    st.write('## Top Results')
+    st.write('## Результаты')
     for count, result in enumerate(st.session_state.results):
         if result["answer"]:
-            answer, context = result["answer"], result["context"]
-            start_idx = context.find(answer)
-            end_idx = start_idx + len(answer)
+            answer, context = result["answer"], result["document"]
+            start_idx = context.find(result["context"])
+            end_idx = start_idx + len(result["context"])
+            reg_id = get_doc_reg_id(result)  
+            module_info = ''         
+            if reg_id:                
+                module_name = mod_names_dict.get(reg_id, None)
+            if module_name:
+                module_info = f"**Наименование модуля/программы: :orange[{module_name}]**"
+            else:
+                module_info = f"Наименование модуля/программы отсутствует!"
+            
+            st.markdown(f"{module_info} - **Релевантность:** {result['relevance']}")
             st.write(
-                markdown(context[:start_idx] + str(annotation(body=answer, label="ANSWER", background="#ff700f", color='#ffffff')) + context[end_idx:]),
+                markdown(context[:start_idx] + str(annotation(body=result["context"], label="ANSWER", background="#ff700f", color='#ffffff')) + context[end_idx:]),
                 unsafe_allow_html=True,
             )
-            source = ""
-            url = get_backlink(result)
-            name = get_doc_name(result)
-            reg_id = get_doc_reg_id(result)
-            if name:
-                source += f"[{result['document']['meta']['name']}]"
-            
-            if url:
-                source += f"({result['document']['meta']['url']})"
 
-            if reg_id:
-                source += f"({result['document']['meta']['reg_id']})"
-            if source:
-                st.markdown(f"**Relevance:** {result['relevance']} -  **Source:** {source}")
-            else:
-                st.markdown(f"**Relevance:** {result['relevance']}")
+            st.markdown(f"**Источник:** {result['source']}")
+
             
         else:
             st.info(
-                "🤔    Haystack is unsure whether any of the documents contain an answer to your question. Try to reformulate it!"
+                "🤔    Поисковая система не справилась с Вашим запросом. Попробуйте его переформулировать!"
             )