Improving prompts - Summarization and Internet - 2

src/main/askai/core/component/cache_service.py

@@ -34,6 +34,11 @@
 if not REC_DIR.exists():
     REC_DIR.mkdir(parents=True, exist_ok=True)
 
+# Transcribed audio directory.
+PERSIST_DIR: Path = Path(str(CACHE_DIR) + "/chroma")
+if not PERSIST_DIR.exists():
+    PERSIST_DIR.mkdir(parents=True, exist_ok=True)
+
 
 class CacheService(metaclass=Singleton):
     """Provide a cache service for previously used queries, audio generation, etc..."""

src/main/askai/core/component/summarizer.py

@@ -30,9 +30,11 @@
 
 from askai.core.askai_events import AskAiEvents
 from askai.core.askai_messages import msg
+from askai.core.component.cache_service import PERSIST_DIR
 from askai.core.model.summary_result import SummaryResult
 from askai.core.support.langchain_support import lc_llm
 from askai.core.support.shared_instances import shared
+from askai.exception.exceptions import DocumentsNotFound
 
 
 class Summarizer(metaclass=Singleton):
@@ -74,10 +76,13 @@ def generate(self, folder: str | Path, glob: str = None) -> None:
         log.info("Summarizing documents from '%s'", self.path)
         embeddings = lc_llm.create_embeddings()
         documents: List[Document] = DirectoryLoader(self.folder, glob=self.glob).load()
-        texts: List[Document] = self._text_splitter.split_documents(documents)
-        v_store = Chroma.from_documents(texts, embeddings)
-        self._retriever = RetrievalQA.from_chain_type(
-            llm=lc_llm.create_model(), chain_type="stuff", retriever=v_store.as_retriever())
+        if len(documents) > 0:
+            texts: List[Document] = self._text_splitter.split_documents(documents)
+            v_store = Chroma.from_documents(texts, embeddings, persist_directory=str(PERSIST_DIR))
+            self._retriever = RetrievalQA.from_chain_type(
+                llm=lc_llm.create_model(), chain_type="stuff", retriever=v_store.as_retriever())
+        else:
+            raise DocumentsNotFound(f"Unable to find any document to summarize at: '{self.path}'")
 
     def query(self, *queries: str) -> Optional[List[SummaryResult]]:
         """Answer questions about the summarized content."""
@@ -103,9 +108,10 @@ def query_one(self, query_string: str) -> Optional[SummaryResult]:
 
 assert (summarizer := Summarizer().INSTANCE) is not None
 
+
 if __name__ == '__main__':
     shared.create_engine('openai', 'gpt-3.5-turbo')
-    summarizer.generate('/Users/hjunior/HomeSetup/docs', '**/*.*')
+    summarizer.generate('/Users/hugo/.config/hhs/log', '**/*.*')
     print(summarizer.query(
         "What is HomeSetup?",
         "How can I install HomeSetup?",

src/main/askai/core/processor/processor_proxy.py

@@ -61,7 +61,7 @@ def process(self, question: str) -> Tuple[bool, QueryResponse]:
             output = object_mapper.of_json(response.message, QueryResponse)
             if not isinstance(output, QueryResponse):
                 log.error(msg.invalid_response(output))
-                output = QueryResponse(question=question, terminating=True, response=response.message)
+                output = response.message
             else:
                 status = True
         else:

src/main/askai/core/processor/summary_processor.py

@@ -27,6 +27,7 @@
 from askai.core.processor.ai_processor import AIProcessor
 from askai.core.support.object_mapper import object_mapper
 from askai.core.support.shared_instances import shared
+from askai.exception.exceptions import DocumentsNotFound
 
 
 class SummaryProcessor(AIProcessor):
@@ -57,8 +58,9 @@ def process(self, query_response: QueryResponse) -> Tuple[bool, Optional[str]]:
                     status = True
             else:
                 output = msg.llm_error(response.message)
-        except (FileNotFoundError, ValueError) as err:
+        except (FileNotFoundError, ValueError, DocumentsNotFound) as err:
             output = msg.llm_error(err)
+            status = True
 
         return status, output
 

src/main/askai/core/support/utilities.py

@@ -31,13 +31,13 @@
 from askai.language.language import Language
 
 ASKAI_CHAT_ICONS = {
-    "": "%RED%",
-    "": "%BLUE%",
-    "": "%BLUE%",
-    "": "%BLUE%",
-    "": "%YELLOW%",
-    "": "%YELLOW%",
-    "": "%ORANGE%",
+    "": "%RED%\n\n",
+    "": "%BLUE%\n\n",
+    "": "%BLUE%\n\n",
+    "": "%BLUE%\n\n",
+    "": "%YELLOW%\n\n",
+    "": "%YELLOW%\n\n",
+    "": "%ORANGE%\n\n",
 }
 
 
@@ -51,18 +51,19 @@ def beautify(text: Any) -> str:
         r'www\.[a-zA-Z0-9][a-zA-Z0-9-]+[a-zA-Z0-9]\.[^\s]{2,}|https?:\/\/(?:www\.|(?!www))'
         r'[a-zA-Z0-9]+\.[^\s]{2,}|www\.[a-zA-Z0-9]+\.[^\s]{2,})')
     text = str(text)
-    text = re.sub(r"\n{2,}", '\n\n', text)
     text = re.sub(r"[Hh]ints?( and tips)?[-:\s][ \n\t]*", f"{ASKAI_CHAT_ICONS['']}{''}  Tips: ", text)
     text = re.sub(r"[Aa]nalysis[-:\s][ \n\t]*", f"{ASKAI_CHAT_ICONS['']}{''}  Analysis: ", text)
     text = re.sub(r"[Ss]ummary[-:\s][ \n\t]*", f"{ASKAI_CHAT_ICONS['']}{''}  Summary:", text)
     text = re.sub(r"([Jj]oke( [Tt]ime)?)[-:\s][ \n\t]*", f"{ASKAI_CHAT_ICONS['']}{''}  Joke: ", text)
     text = re.sub(r"[Ff]un [Ff]acts?[-:\s][ \n\t]*", f"{ASKAI_CHAT_ICONS['']}{''}  Fun Fact: ", text)
     text = re.sub(r"[Aa]dvice[-:\s][ \n\t]*", f"{ASKAI_CHAT_ICONS['']}{''}  Advice: ", text)
-    text = re.sub(r"Errors?[-:\s][ \n\t]*", f"{ASKAI_CHAT_ICONS['']}{''}  Error: ", text)
+    text = re.sub(r"Errors?[-:\s][ \n\t]*", f"%EL0%{ASKAI_CHAT_ICONS['']}{''}  Error: ", text)
+    text = re.sub(r"^\n+", '', text, re.MULTILINE)
+    text = re.sub(r"\n{2,}", '\n', text, re.MULTILINE)
     text = re.sub(re_url, r'%CYAN% \1%GREEN%', text)
     # fmt: on
 
-    return text
+    return text.rstrip()
 
 
 def read_resource(base_dir: str, filename: str, file_ext: str = ".txt") -> str:
@@ -194,16 +195,20 @@ def stream_text(text: Any, tempo: int = 1, language: Language = Language.EN_US)
 
 if __name__ == '__main__':
     display_text("""
-    Este text tem ln
 
 
 
 
-     aqui
 
 
 
-     este
+
+
+According to the summarized content, the log files at ~/.config/hhs/log contain various information that can be helpful for tracking activities or troubleshooting issues. I recommend reviewing them regularly to stay on top of what's happening in that directory.
+
+
+
+Fun Fact: Log files are like diaries for your system, documenting everything that happens behind the scenes!
 
      """
-         )
+                 )

src/main/askai/exception/exceptions.py

@@ -50,3 +50,7 @@ class InvalidJsonMapping(HSBaseException):
 
 class InvalidInputDevice(HSBaseException):
     """Raised when an invalid recording input device is used."""
+
+
+class DocumentsNotFound(HSBaseException):
+    """Raised when no documents are found for summarization."""

src/main/askai/resources/assets/prompts/proxy-prompt.txt

@@ -4,8 +4,6 @@ ${persona}
 
 Before responding to the user, it is imperative that you follow the step-by-step instructions provided below in sequential order:
 
-- Prior to making any decisions, it's crucial to review the entire chat history. The context provided previously may influence the outcome, so it's important to ensure that the answer isn't already within that context before proceeding.
-
 - Determine whether the query clear and intelligible.
 
 - Determine whether the query suggests the user intends to end the conversation.
@@ -21,3 +19,5 @@ Before responding to the user, it is imperative that you follow the step-by-step
 - The final response 'JSON' must contain the boolean fields: 'intelligible', 'terminating', 'require_internet' and 'require_summarization'.
 
 - The final response 'JSON' must contain the string fields: fields: 'query_type' and 'question'.
+
+- Prior to making any decisions, it's crucial to review the entire chat history. The context provided previously may influence the outcome, so it's important to ensure that the answer isn't already within that context before proceeding.