sunholo 0.80.4__tar.gz → 0.80.6__tar.gz

{sunholo-0.80.4 → sunholo-0.80.6}/PKG-INFO

@@ -1,9 +1,9 @@
 Metadata-Version: 2.1
 Name: sunholo
-Version: 0.80.4
+Version: 0.80.6
 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/v0.80.4.tar.gz
+Download-URL: https://github.com/sunholo-data/sunholo-py/archive/refs/tags/v0.80.6.tar.gz
 Author: Holosun ApS
 Author-email: multivac@sunholo.com
 License: Apache License, Version 2.0

{sunholo-0.80.4 → sunholo-0.80.6}/setup.py

@@ -1,7 +1,7 @@
 from setuptools import setup, find_packages
 
 # Define your base version
-version = '0.80.4'
+version = '0.80.6'
 
 setup(
     name='sunholo',

{sunholo-0.80.4 → sunholo-0.80.6}/sunholo/chunker/process_chunker_data.py

@@ -15,6 +15,7 @@ from ..discovery_engine.chunker_handler import discovery_engine_chunker_check
 from .publish import process_docs_chunks_vector_name
 from .splitter import chunk_doc_to_docs
 from ..azure.blobs import is_azure_blob
+from ..utils import ConfigManager
 
 from ..custom_logging import log
 
@@ -29,13 +30,15 @@ def process_chunker_data(message_data, metadata, vector_name):
 
     log.debug(f"Found metadata in pubsub: {metadata=}")
 
+    config=ConfigManager(vector_name)
+
     # checks if only a llamaindex chunking/embedder, return early as no other processing needed
     llamacheck = llamaindex_chunker_check(message_data, metadata, vector_name)
     if llamacheck:
         return llamacheck
     
     # if only a discovery engine memory, return early as no other processing needed
-    discovery_check = discovery_engine_chunker_check(message_data, metadata, vector_name)
+    discovery_check = discovery_engine_chunker_check(message_data, metadata, config=config)
     if discovery_check:
         return discovery_check
 

{sunholo-0.80.4 → sunholo-0.80.6}/sunholo/components/retriever.py

@@ -60,6 +60,9 @@ def pick_retriever(vector_name:str=None, config:ConfigManager=None, embeddings=N
 
                 if vectorstore == "vertex_ai_search" or vectorstore == "discovery_engine":
                     # use direct retriever
+                    if value.get('chunks'):
+                        log.warning(f"{config.vector_name} will not be using GoogleVertexAISearchRetriever with chunks vertex AI search as not supported yet")
+                        continue
                     from langchain.retrievers import GoogleVertexAISearchRetriever
                     gcp_config = config.vacConfig('gcp_config')
                     try:
@@ -76,6 +79,7 @@ def pick_retriever(vector_name:str=None, config:ConfigManager=None, embeddings=N
                         continue
                     
                     retriever_list.append(gcp_retriever)
+                    continue
 
                 from_metadata_id = value.get('from_metadata_id')
                 if from_metadata_id:

{sunholo-0.80.4 → sunholo-0.80.6}/sunholo/discovery_engine/__init__.py

@@ -1 +1,2 @@
 from .discovery_engine_client import DiscoveryEngineClient
+from .get_ai_search_chunks import get_all_chunks

{sunholo-0.80.4 → sunholo-0.80.6}/sunholo/discovery_engine/chunker_handler.py

@@ -1,5 +1,5 @@
 from ..custom_logging import log
-from ..utils.config import load_config_key
+from ..utils import ConfigManager
 from ..utils.gcp_project import get_gcp_project
 from ..components import load_memories
 
@@ -7,7 +7,7 @@ from .discovery_engine_client import DiscoveryEngineClient
 from .create_new import create_new_discovery_engine
     
 
-def do_discovery_engine(message_data, metadata, vector_name):
+def do_discovery_engine(message_data:str, metadata:dict, config:ConfigManager=None):
     """
 
     Example:
@@ -15,13 +15,13 @@ def do_discovery_engine(message_data, metadata, vector_name):
     message_data = "gs://bucket_name/path_to_file.txt"
     metadata = {"user": "admin"}
     vector_name = "example_vector"
-    response = do_discovery_engine(message_data, metadata, vector_name)
+    response = do_discovery_engine(message_data, metadata, config=config)
     print(response)
     # Imported file to corpus: {'status': 'success'}
     ```
     """
 
-    memories = load_memories(vector_name)
+    memories = load_memories(config=config)
     tools = []
 
     if not memories:
@@ -38,7 +38,7 @@ def do_discovery_engine(message_data, metadata, vector_name):
                     continue
                 #location = gcp_config.get('location') 
                 corpus = DiscoveryEngineClient(
-                    data_store_id=vector_name, 
+                    data_store_id=config.vector_name, 
                     project_id=get_gcp_project(),
                     # location needs to be 'eu' or 'us' which doesn't work with other configurations
                     #location=location or global_location
@@ -65,14 +65,14 @@ def do_discovery_engine(message_data, metadata, vector_name):
                 log.error(f"Error importing {message_data} - {corp=} - {str(err)}")
 
                 if str(err).startswith("404"):
-                    log.info(f"Attempting to create a new DiscoveryEngine corpus: {vector_name}")
+                    log.info(f"Attempting to create a new DiscoveryEngine corpus: {config.vector_name}")
                     try:
-                        new_corp = create_new_discovery_engine(vector_name)
+                        new_corp = create_new_discovery_engine(config)
                     except Exception as err:
-                        log.error(f"Failed to create new DiscoveryEngine {vector_name} - {str(err)}")
+                        log.error(f"Failed to create new DiscoveryEngine {config.vector_name} - {str(err)}")
                         continue
                     if new_corp:
-                        log.info(f"Found new DiscoveryEngine {vector_name=} - {new_corp=}")
+                        log.info(f"Found new DiscoveryEngine {config.vector_name=} - {new_corp=}")
                         response = corp.import_documents(
                             gcs_uri=message_data
                         )
@@ -86,8 +86,9 @@ def do_discovery_engine(message_data, metadata, vector_name):
         log.warning("Only gs:// data is supported for Discovery Engine")
 
 
-def check_discovery_engine_in_memory(vector_name):
-    memories = load_config_key("memory", vector_name=vector_name, kind="vacConfig")
+def check_discovery_engine_in_memory(config:ConfigManager):
+    memories = config.vacConfig("memory")
+
     for memory in memories:  # Iterate over the list
         for key, value in memory.items():  # Now iterate over the dictionary
             log.info(f"Found memory {key}")
@@ -99,16 +100,33 @@ def check_discovery_engine_in_memory(vector_name):
     
     return False
 
-def discovery_engine_chunker_check(message_data, metadata, vector_name):
+def check_write_memories(config:ConfigManager):
+    write_mem = []
+    memories = config.vacConfig("memory")
+    for memory in memories:
+        for key, value in memory.items():
+            if value.get('read_only'):
+                continue
+            write_mem.append(memory)
+    
+    return write_mem
+
+def discovery_engine_chunker_check(message_data, metadata, vector_name:str=None, config:ConfigManager=None):
+
+    if config is None:
+        if vector_name is None:
+            raise ValueError("Must provide config or vector_name")
+        config = ConfigManager(vector_name=vector_name)
+
     # discovery engine handles its own chunking/embedding
-    memories = load_config_key("memory", vector_name=vector_name, kind="vacConfig")
+    memories = config.vacConfig("memory")
     if not memories:
         return None
     
-    total_memories = len(memories)
+    total_memories = len(check_write_memories(config))
     llama = None
-    if check_discovery_engine_in_memory(vector_name):
-        llama = do_discovery_engine(message_data, metadata, vector_name)
+    if check_discovery_engine_in_memory(config):
+        llama = do_discovery_engine(message_data, metadata, config=config)
         log.info(f"Processed discovery engine: {llama}")
 
     # If discovery engine is the only entry, return

{sunholo-0.80.4 → sunholo-0.80.6}/sunholo/discovery_engine/create_new.py

@@ -1,10 +1,10 @@
 from .discovery_engine_client import DiscoveryEngineClient
-from ..utils.config import load_config_key
+from ..utils import ConfigManager
 from ..utils.gcp_project import get_gcp_project
 
-def create_new_discovery_engine(vector_name):
+def create_new_discovery_engine(config:ConfigManager):
 
-    chunker_config = load_config_key("chunker", vector_name=vector_name, kind="vacConfig")
+    chunker_config = config.vacConfig("chunker")
 
     chunk_size = 500
     if chunker_config:
@@ -18,7 +18,7 @@ def create_new_discovery_engine(vector_name):
     #location = gcp_config.get('location')
 
     de = DiscoveryEngineClient(
-                    data_store_id=vector_name, 
+                    data_store_id=config.vector_name, 
                     project_id=project_id,
                     # location needs to be 'eu' or 'us' which doesn't work with other configurations
                     #location=location

sunholo-0.80.6/sunholo/discovery_engine/get_ai_search_chunks.py

@@ -0,0 +1,53 @@
+from ..utils import ConfigManager
+from ..utils.gcp_project  import get_gcp_project
+from ..custom_logging import log
+from .discovery_engine_client import DiscoveryEngineClient
+from ..components import load_memories
+
+def get_all_chunks(question:str, config:ConfigManager):
+    """
+    Look through a config memory key and find all Vertex AI Search retrievers, call them and return a joined string of chunks
+
+        args: question - question to search similarity for
+        config: A ConfigManager object
+
+        returns: a big string of chunks
+    """
+    memories = load_memories(config=config)
+    chunks = []
+
+    if not memories:
+        return None
+
+    vector_name = config.vector_name
+    for memory in memories:
+        for key, value in memory.items():  # Now iterate over the dictionary
+            log.info(f"Found memory {key}")
+            vectorstore = value.get('vectorstore')
+            if vectorstore == "discovery_engine" or vectorstore == "vertex_ai_search":
+                if value.get('read_only'):
+                    new_vector_name = value.get('vector_name')
+                    if not new_vector_name:
+                        log.warning("read_only specified but no new vector_name to read from")
+                    vector_name = new_vector_name
+                
+                num_chunks = value.get('num_chunks') or 3
+
+                chunk = get_chunks(question, vector_name, num_chunks)
+                if chunk:
+                    chunks.append(chunk)
+    if chunks:
+        return "\n".join(chunks)
+
+    log.warning(f"No chunks found for {vector_name}")
+    return None
+
+def get_chunks(question, vector_name, num_chunks):
+    de = DiscoveryEngineClient(vector_name, project_id=get_gcp_project())
+    try:
+        return de.get_chunks(question, num_previous_chunks=num_chunks, num_next_chunks=num_chunks)
+    except Exception as err:
+        log.error(f"No discovery engine chunks found: {str(err)}")
+    
+
+

{sunholo-0.80.4 → sunholo-0.80.6}/sunholo/gcs/add_file.py

@@ -118,12 +118,13 @@ def add_file_to_gcs(filename: str,
     if os.getenv('EXTENSIONS_BUCKET'):
         bucket_filepath = os.path.basename(filename)
 
-    if vector_name is None:
-            vector_name = "global"
+    if not vector_name:
+        vector_name = "global"
     
     if not bucket_filepath:
         
         bucket_filepath = f"{vector_name}/{year}/{month}/{day}/{hour}/{os.path.basename(filename)}"
+    
     bucket_filepath_prev = f"{vector_name}/{year}/{month}/{day}/{hour_prev}/{os.path.basename(filename)}"
 
     blob = bucket.blob(bucket_filepath)

{sunholo-0.80.4 → sunholo-0.80.6}/sunholo.egg-info/PKG-INFO

@@ -1,9 +1,9 @@
 Metadata-Version: 2.1
 Name: sunholo
-Version: 0.80.4
+Version: 0.80.6
 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/v0.80.4.tar.gz
+Download-URL: https://github.com/sunholo-data/sunholo-py/archive/refs/tags/v0.80.6.tar.gz
 Author: Holosun ApS
 Author-email: multivac@sunholo.com
 License: Apache License, Version 2.0

{sunholo-0.80.4 → sunholo-0.80.6}/sunholo.egg-info/SOURCES.txt

@@ -85,6 +85,7 @@ sunholo/discovery_engine/__init__.py
 sunholo/discovery_engine/chunker_handler.py
 sunholo/discovery_engine/create_new.py
 sunholo/discovery_engine/discovery_engine_client.py
+sunholo/discovery_engine/get_ai_search_chunks.py
 sunholo/embedder/__init__.py
 sunholo/embedder/embed_chunk.py
 sunholo/gcs/__init__.py