sunholo 0.79.5__py3-none-any.whl → 0.80.0__py3-none-any.whl

sunholo/components/retriever.py

@@ -34,7 +34,7 @@ def load_memories(vector_name:str=None, config:ConfigManager=None):
 
     memories = config.vacConfig("memory")
     
-    log.info(f"Found memory settings for {vector_name}: {memories}")
+    log.info(f"Found memory settings for {config.vector_name}: {memories}")
     if not memories or len(memories) == 0:
         log.info(f"No memory settings found for {vector_name}")
         return None
@@ -66,12 +66,13 @@ def pick_retriever(vector_name:str=None, config:ConfigManager=None, embeddings=N
                 embeddings = embeddings or get_embeddings(config=config)
                 read_only = value.get('read_only')
                 try:
-                    vectorstore = pick_vectorstore(vectorstore, 
-                                                vector_name=vector_name, 
-                                                embeddings=embeddings, 
-                                                read_only=read_only)
+                    vectorstore = pick_vectorstore(
+                        vectorstore, 
+                        config=config,
+                        embeddings=embeddings, 
+                        read_only=read_only)
                 except Exception as e:
-                    log.error(f"Failed to pick_vectorstore {vectorstore} for {vector_name} - {str(e)} - skipping")
+                    log.error(f"Failed to pick_vectorstore {vectorstore} for {config.vector_name} - {str(e)} - skipping")
                     continue
                 
                 k_override = value.get('k', 3)

sunholo/llamaindex/llamaindex_class.py

@@ -3,6 +3,7 @@ import tempfile
 
 try:
     from vertexai.preview import rag
+    from google.cloud.aiplatform_v1beta1 import RetrieveContextsResponse
 except ImportError:
     rag = None
 
@@ -37,12 +38,15 @@ class LlamaIndexVertexCorpusManager:
             ImportError: If vertexai is not installed.
         """
         from ..vertex.init import init_vertex
+
         if rag is None:
             raise ImportError("You must install vertexai via `pip install sunholo[gcp]`")
         
         self.config = config
         self.project_id = project_id
         self.location = location
+        self.corpus_display_name = ""
+        self.corpus = ""
 
         if config:
             self.project_id = self.config.vacConfig('project_id') or project_id
@@ -69,10 +73,13 @@ class LlamaIndexVertexCorpusManager:
         
         try:
             # Upload the temporary file
+            log.info(f"Uploading text:{text[:50]}... to {corpus_display_name}")
             uploaded_file = self.upload_file(temp_filename, corpus_display_name, description or text[:50])
         finally:
             # Clean up the temporary file
             os.remove(temp_filename)
+        
+        log.info(f"Successfully uploaded text:{text[:50]}... to {corpus_display_name}")
 
         return uploaded_file
     
@@ -85,7 +92,7 @@ class LlamaIndexVertexCorpusManager:
             display_name=filename,
             description=description or f"Upload for {filename}",
         )
-        log.info("Uploaded file: {rag_file}")
+        log.info(f"Uploaded file: {rag_file}")
 
         return rag_file
     
@@ -104,6 +111,89 @@ class LlamaIndexVertexCorpusManager:
 
         return response
     
+    def list_files(self, corpus_display_name:str):
+        corpus = self.find_corpus_from_list(corpus_display_name)
+        files = rag.list_files(corpus_name=corpus.name)
+
+        log.info(f"--Files in {corpus.name}:\n{files}")
+        
+        return files
+
+    def find_file_from_list(self, display_name: str, corpus_display_name:str):
+        """
+        Finds a file from the list of files by its display name.
+
+        Args:
+            display_name (str): The display name of the file.
+            corpus_display_name (str): The display name of the corpus to look within
+
+        Returns:
+            The found file object if it exists, otherwise None.
+        """
+        files = self.list_files(corpus_display_name)
+        for file in files:
+            if display_name == file.display_name:
+                log.info(f"Found existing file with display name: {display_name}")
+
+                return file
+        
+        return None
+    
+    def get_file(self, file_display_name:str=None, file_name:str=None, corpus_display_name:str=None):
+        
+        if file_display_name:
+            rag_file = self.find_file_from_list(file_display_name, corpus_display_name)
+            log.info(f"Found {rag_file} via display name: {file_display_name}")
+            
+            return rag_file
+        
+        if not file_name:
+            raise ValueError("Need to supply one of file_display_name or file_name")
+
+        corpus = self.find_corpus_from_list(corpus_display_name)
+        if file_name.startswith("projects/"):
+            rag_file = rag.get_file(name=file_name)
+        else:
+            if not corpus_display_name:
+                raise ValueError("Must supply corpus_display_name if not a full file_name")
+            rag_file = rag.get_file(name=file_name, corpus_name=corpus.name)
+        
+        log.info(f"Found {rag_file}")
+
+        return rag_file
+    
+    def delete_file(self, file_name, corpus_display_name:str):
+
+        corpus = self.find_corpus_from_list(corpus_display_name)
+        if file_name.startswith("projects/"):
+            rag.delete_file(name=file_name)
+        else:
+            if not corpus_display_name:
+                raise ValueError("Must supply corpus_display_name if not a full file_name")
+            rag.delete_file(name=file_name, corpus_name=corpus.name)
+        
+        log.info(f"File {file_name} deleted.")
+
+        return True
+    
+    def query_corpus(self, query:str, corpus_disply_name:str):
+        corpus = self.find_corpus_from_list(corpus_disply_name)
+        
+        response:RetrieveContextsResponse = rag.retrieval_query(
+            rag_resources=[
+                rag.RagResource(
+                    rag_corpus=corpus.name,
+                    # Supply IDs from `rag.list_files()`.
+                    # rag_file_ids=["rag-file-1", "rag-file-2", ...],
+                )
+            ],
+            text=query,
+            similarity_top_k=10,  # Optional
+            vector_distance_threshold=0.5,  # Optional
+        )
+
+        return response
+
     def list_corpora(self):
         """
         List all VertexAI Corpus for the project/location
@@ -120,10 +210,13 @@ class LlamaIndexVertexCorpusManager:
         Returns:
             The found corpus object if it exists, otherwise None.
         """
+        if display_name == self.corpus_display_name:
+            return self.corpus
         corpora = self.list_corpora()
         for corp in corpora:
             if display_name == corp.display_name:
                 log.info(f"Found existing corpus with display name: {display_name}")
+                self.corpus = corp
                 return corp
         return None
 
@@ -211,6 +304,29 @@ def llamaindex_command(args):
         manager.upload_file(filename=args.filename, corpus_display_name=args.display_name, description=args.description)
     elif args.action == "upload_text":
         manager.upload_text(text=args.text, corpus_display_name=args.display_name, description=args.description)
+    elif args.action == "list_files":
+        files = manager.list_files(corpus_display_name=args.display_name)
+        if files:
+            console.print(files)
+        else:
+            console.print("No files found for {args.display_name}")
+    elif args.action == "get_file":
+        file = manager.get_file(file_display_name=args.file_name, corpus_display_name=args.display_name)
+        console.print(file)
+        return file
+    elif args.action == "delete_file":
+        deleted = manager.delete_file(args.file_name, corpus_display_name=args.display_name)
+        if deleted:
+            console.print(f"Deleted {args.file_name}")
+        else:
+            console.print(f"ERROR: Could not delete {args.file_name}")
+
+    elif args.action == "query":
+        answer = manager.query_corpus(args.query, corpus_disply_name=args.display_name)
+        if answer:
+            console.print(answer)
+        else:
+            console.print(f"No answer found for {args.query} in {args.display_name}")
     else:
         console.print(f"Unknown action: {args.action}")
 
@@ -270,5 +386,27 @@ def setup_llamaindex_subparser(subparsers):
     upload_text_parser.add_argument('text', help='The text content to upload')
     upload_text_parser.add_argument('--description', help='Description of the text upload', default=None)
 
+    # LlamaIndex list_files command
+    list_files_parser = llamaindex_subparsers.add_parser('list_files', help='List all files in a corpus')
+    list_files_parser.add_argument('display_name', help='The name of the corpus')
+    list_files_parser.add_argument('vac', nargs='?', default="global", help='The VAC config to set it up for')
+
+    # LlamaIndex get_file command
+    get_file_parser = llamaindex_subparsers.add_parser('get_file', help='Get a file from a corpus')
+    get_file_parser.add_argument('display_name', help='The name of the corpus')
+    get_file_parser.add_argument('file_name', help='The name of the file to get')
+    get_file_parser.add_argument('vac', nargs='?', default="global", help='The VAC config to set it up for')
+
+    # LlamaIndex delete_file command
+    delete_file_parser = llamaindex_subparsers.add_parser('delete_file', help='Delete a file from a corpus')
+    delete_file_parser.add_argument('display_name', help='The name of the corpus')
+    delete_file_parser.add_argument('file_name', help='The name of the file to delete')
+    delete_file_parser.add_argument('vac', nargs='?', default="global", help='The VAC config to set it up for')
+
+    # LlamaIndex query command
+    query_parser = llamaindex_subparsers.add_parser('query', help='Query a corpus')
+    query_parser.add_argument('display_name', help='The name of the corpus')
+    query_parser.add_argument('query', help='The query string')
+    query_parser.add_argument('vac', nargs='?', default="global", help='The VAC config to set it up for')
+
     llamaindex_parser.set_defaults(func=llamaindex_command)
-    

sunholo/llamaindex/user_history.py

@@ -0,0 +1,61 @@
+from ..utils import ConfigManager
+from ..custom_logging import log
+from .llamaindex_class import LlamaIndexVertexCorpusManager
+
+import datetime
+
+def add_user_history_rag(
+        user_id:str, 
+        config:ConfigManager, 
+        question:str, 
+        answer:str, 
+        metadata:dict={},
+        user_history_template:str=None):
+    # add user history to its own RAG store
+
+    log.info(f"Adding user history to RAG store: {question} and {answer}")
+
+    manager = LlamaIndexVertexCorpusManager(config)
+
+    corpus = manager.create_corpus(user_id, description=f"Personal user history for {user_id}")
+
+    current_datetime = datetime.datetime.now()
+
+    # Convert to string with desired format
+    current_datetime_str = current_datetime.strftime('%Y-%m-%d %H:%M:%S')
+
+    if user_history_template is None:
+        user_history_template="""Question from {user_id} at {the_date}: {the_question}\nAnswer: {the_answer}\nMetadata:{the_metadata}"""
+
+    log.info(f"Found corpus for {user_id}: {corpus}")
+    user_history = user_history_template.format(
+        user_id=user_id, 
+        the_date=current_datetime_str,
+        the_question=question, 
+        the_answer=answer,
+        the_metadata=metadata
+        )
+
+    try:
+        manager.upload_text(
+            text=user_history, 
+            corpus_display_name=user_id, 
+            description=f"{user_id} chat history for {current_datetime}"
+            )
+    except Exception as err:
+        log.error(f"Could not upload LlamaIndex QNA RAG history: {str(err)}")
+    
+    return user_history
+
+def get_user_history_chunks(user_id:str, config:ConfigManager, query):
+
+    manager = LlamaIndexVertexCorpusManager(config)
+
+    manager.create_corpus(user_id)
+
+    response = manager.query_corpus(query, user_id)
+    user_history_memory = []
+    for chunk in response.contexts.contexts:
+        user_history_memory.append(chunk.text)
+    
+    return "\n".join(user_history_memory)

sunholo/utils/config_class.py

@@ -159,6 +159,7 @@ class ConfigManager:
             if isinstance(value, dict) and key in dict1 and isinstance(dict1[key], dict):
                 dict1[key] = self._merge_dicts(dict1[key], value)
             else:
+                print(f"Merging '{key}' to new '{value}'")
                 dict1[key] = value
         return dict1
 

{sunholo-0.79.5.dist-info → sunholo-0.80.0.dist-info}/METADATA

@@ -1,9 +1,9 @@
 Metadata-Version: 2.1
 Name: sunholo
-Version: 0.79.5
+Version: 0.80.0
 Summary: Large Language Model DevOps - a package to help deploy LLMs to the Cloud.
 Home-page: https://github.com/sunholo-data/sunholo-py
-Download-URL: https://github.com/sunholo-data/sunholo-py/archive/refs/tags/v0.79.5.tar.gz
+Download-URL: https://github.com/sunholo-data/sunholo-py/archive/refs/tags/v0.80.0.tar.gz
 Author: Holosun ApS
 Author-email: multivac@sunholo.com
 License: Apache License, Version 2.0

{sunholo-0.79.5.dist-info → sunholo-0.80.0.dist-info}/RECORD

@@ -55,7 +55,7 @@ sunholo/cli/swagger.py,sha256=absYKAU-7Yd2eiVNUY-g_WLl2zJfeRUNdWQ0oH8M_HM,1564
 sunholo/cli/vertex.py,sha256=8130YCarxHL1UC3aqblNmUwGZTXbkdL4Y_FOnZJsWiI,2056
 sunholo/components/__init__.py,sha256=IDoylb74zFKo6NIS3RQqUl0PDFBGVxM1dfUmO7OJ44U,176
 sunholo/components/llm.py,sha256=5wRVf7lIb7q1vRADNcdQp26L9l4vGHFIvjtUDurZN_s,11488
-sunholo/components/retriever.py,sha256=KK8Df3NH8jNpEymPBfR2913rrQZN-PRMUGedtYNIV6k,6905
+sunholo/components/retriever.py,sha256=ECTXmKiFYT7B9oFaXSEbmQ9m2OZOOpqzcPccFi9KDpc,6861
 sunholo/components/vectorstore.py,sha256=xKk7micTRwZckaI7U6PxvFz_ZSjCH48xPTDYiDcv2tc,5913
 sunholo/database/__init__.py,sha256=Zz0Shcq-CtStf9rJGIYB_Ybzb8rY_Q9mfSj-nviM490,241
 sunholo/database/alloydb.py,sha256=c1PEmK9fJCxYaVmKv4emvOoXrajV7KqaVK5mqpeksvM,11527
@@ -90,7 +90,8 @@ sunholo/langfuse/prompts.py,sha256=27BsVfihM6-h1jscbkGSO4HsATl-d4ZN6tcNCVztWoY,1
 sunholo/llamaindex/__init__.py,sha256=DlY_cHWCsVEV1C5WBgDdHRgOMlJc8pDoCRukUJ8PT9w,88
 sunholo/llamaindex/get_files.py,sha256=6rhXCDqQ_lrIapISQ_OYQDjiSATXvS_9m3qq53-oIl0,781
 sunholo/llamaindex/import_files.py,sha256=Bnic5wz8c61af9Kwq8KSrNBbc4imYnzMtBCb2jzSImI,6224
-sunholo/llamaindex/llamaindex_class.py,sha256=45N_xU0gDdgUPNv5m36DopRmG9JHFErwmgssKZ3wOqA,11058
+sunholo/llamaindex/llamaindex_class.py,sha256=MYCgmJ2ebN0yrJbkshGIoaoX0pKi5yONLKbEPULDXRY,17049
+sunholo/llamaindex/user_history.py,sha256=SwBNDOU2hsAzOMVwDjXQsR3d8RPUZofKTlZV_pAraO8,1934
 sunholo/lookup/__init__.py,sha256=47DEQpj8HBSa-_TImW-5JCeuQeRkm5NMpJWZG3hSuFU,0
 sunholo/lookup/model_lookup.yaml,sha256=O7o-jP53MLA06C8pI-ILwERShO-xf6z_258wtpZBv6A,739
 sunholo/patches/__init__.py,sha256=47DEQpj8HBSa-_TImW-5JCeuQeRkm5NMpJWZG3hSuFU,0
@@ -116,7 +117,7 @@ sunholo/utils/__init__.py,sha256=Hv02T5L2zYWvCso5hzzwm8FQogwBq0OgtUbN_7Quzqc,89
 sunholo/utils/api_key.py,sha256=Ct4bIAQZxzPEw14hP586LpVxBAVi_W9Serpy0BK-7KI,244
 sunholo/utils/big_context.py,sha256=gJIP7_ZL-YSLhOMq8jmFTMqH1wq8eB1NK7oKPeZAq2s,5578
 sunholo/utils/config.py,sha256=aG29MXcL5qzQMtCMqcdy-2ysDCYf9Zn_ZLk5NNOQNSE,8982
-sunholo/utils/config_class.py,sha256=DEY2w7ks5fXmtJ4SwqELhUF4NJFVOuRwxeVFZJlYPf0,8963
+sunholo/utils/config_class.py,sha256=k2wKuvEQlbqfUzujGcrDH_ZuPEhtYrlksJr-sBy-elg,9022
 sunholo/utils/config_schema.py,sha256=Wv-ncitzljOhgbDaq9qnFqH5LCuxNv59dTGDWgd1qdk,4189
 sunholo/utils/gcp.py,sha256=uueODEpA-P6O15-t0hmcGC9dONLO_hLfzSsSoQnkUss,4854
 sunholo/utils/gcp_project.py,sha256=0ozs6tzI4qEvEeXb8MxLnCdEVoWKxlM6OH05htj7_tc,1325
@@ -132,9 +133,9 @@ sunholo/vertex/init.py,sha256=1OQwcPBKZYBTDPdyU7IM4X4OmiXLdsNV30C-fee2scQ,2875
 sunholo/vertex/memory_tools.py,sha256=pgSahVDh7GPEulu3nl-w0jb5lTClb4TCnVxPnMokNZY,7533
 sunholo/vertex/safety.py,sha256=S9PgQT1O_BQAkcqauWncRJaydiP8Q_Jzmu9gxYfy1VA,2482
 sunholo/vertex/type_dict_to_json.py,sha256=uTzL4o9tJRao4u-gJOFcACgWGkBOtqACmb6ihvCErL8,4694
-sunholo-0.79.5.dist-info/LICENSE.txt,sha256=SdE3QjnD3GEmqqg9EX3TM9f7WmtOzqS1KJve8rhbYmU,11345
-sunholo-0.79.5.dist-info/METADATA,sha256=B4e56x_LDOAPMTDXey40Sa4P0P9j69ZqL2pZUVjPaj8,7348
-sunholo-0.79.5.dist-info/WHEEL,sha256=R0nc6qTxuoLk7ShA2_Y-UWkN8ZdfDBG2B6Eqpz2WXbs,91
-sunholo-0.79.5.dist-info/entry_points.txt,sha256=bZuN5AIHingMPt4Ro1b_T-FnQvZ3teBes-3OyO0asl4,49
-sunholo-0.79.5.dist-info/top_level.txt,sha256=wt5tadn5--5JrZsjJz2LceoUvcrIvxjHJe-RxuudxAk,8
-sunholo-0.79.5.dist-info/RECORD,,
+sunholo-0.80.0.dist-info/LICENSE.txt,sha256=SdE3QjnD3GEmqqg9EX3TM9f7WmtOzqS1KJve8rhbYmU,11345
+sunholo-0.80.0.dist-info/METADATA,sha256=dL9qKmYHLd8M7rmztP84WKfSQ_aoVfLmwA6pU0E3jM0,7348
+sunholo-0.80.0.dist-info/WHEEL,sha256=R0nc6qTxuoLk7ShA2_Y-UWkN8ZdfDBG2B6Eqpz2WXbs,91
+sunholo-0.80.0.dist-info/entry_points.txt,sha256=bZuN5AIHingMPt4Ro1b_T-FnQvZ3teBes-3OyO0asl4,49
+sunholo-0.80.0.dist-info/top_level.txt,sha256=wt5tadn5--5JrZsjJz2LceoUvcrIvxjHJe-RxuudxAk,8
+sunholo-0.80.0.dist-info/RECORD,,