sunholo 0.58.0__tar.gz → 0.58.2__tar.gz

{sunholo-0.58.0 → sunholo-0.58.2}/PKG-INFO

@@ -1,9 +1,9 @@
 Metadata-Version: 2.1
 Name: sunholo
-Version: 0.58.0
+Version: 0.58.2
 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/v0.58.0.tar.gz
+Download-URL: https://github.com/sunholo-data/sunholo-py/archive/refs/tags/v0.58.2.tar.gz
 Author: Holosun ApS
 Author-email: multivac@sunholo.com
 License: Apache License, Version 2.0

{sunholo-0.58.0 → sunholo-0.58.2}/setup.py

@@ -1,7 +1,7 @@
 from setuptools import setup, find_packages
 
 # Define your base version
-version = '0.58.0'
+version = '0.58.2'
 
 setup(
     name='sunholo',

{sunholo-0.58.0 → sunholo-0.58.2}/sunholo/components/retriever.py

@@ -27,7 +27,7 @@ from langchain.retrievers import ContextualCompressionRetriever
 
 
 def load_memories(vector_name):
-    memories = load_config_key("memory", vector_name, filename="config/llm_config.yaml")
+    memories = load_config_key("memory", vector_name, type="vacConfig")
     log.info(f"Found memory settings for {vector_name}: {memories}")
     if len(memories) == 0:
         log.info(f"No memory settings found for {vector_name}")
@@ -49,7 +49,8 @@ def pick_retriever(vector_name, embeddings=None):
                 if embeddings is None:
                     embeddings = get_embeddings(vector_name)
                 vectorstore = pick_vectorstore(vectorstore, vector_name=vector_name, embeddings=embeddings)
-                vs_retriever = vectorstore.as_retriever(search_kwargs=dict(k=3))
+                k_override = value.get('k', 3)
+                vs_retriever = vectorstore.as_retriever(search_kwargs=dict(k=k_override))
                 retriever_list.append(vs_retriever)
             
             if value.get('provider') == "GoogleCloudEnterpriseSearchRetriever":
@@ -68,6 +69,10 @@ def pick_retriever(vector_name, embeddings=None):
     if len(retriever_list) == 0:
         log.info(f"No retrievers were created for {memories}")
         return None
+    
+    k_override = load_config_key("memory_k", vector_name, type="vacConfig")
+    if not k_override:
+        k_override = 3
         
     lotr = MergerRetriever(retrievers=retriever_list)
 
@@ -76,6 +81,6 @@ def pick_retriever(vector_name, embeddings=None):
     pipeline = DocumentCompressorPipeline(transformers=[filter])
     retriever = ContextualCompressionRetriever(
         base_compressor=pipeline, base_retriever=lotr, 
-        k=3)
+        k=k_override)
 
     return retriever

{sunholo-0.58.0 → sunholo-0.58.2}/sunholo/embedder/embed_chunk.py

@@ -115,6 +115,15 @@ def embed_pubsub_chunk(data: dict):
                 embed_llm = value.get('llm')
                 if embed_llm:
                     embeddings = pick_embedding(embed_llm)
+                # check if read only
+                read_only = value.get('readonly')
+                if read_only:
+                    continue
+                # read from a different vector_name
+                vector_name_other = value.get('vector_name')
+                if vector_name_other:
+                    log.warning(f"Using different vector_name for vectorstore: {vector_name_other} overriding {vector_name}")
+                    vector_name = vector_name_other
                 vectorstore_obj = pick_vectorstore(vectorstore, vector_name=vector_name, embeddings=embeddings)
                 vs_retriever = vectorstore_obj.as_retriever(search_kwargs=dict(k=3))
                 vectorstore_list.append(vs_retriever)

{sunholo-0.58.0 → sunholo-0.58.2}/sunholo/llamaindex/import_files.py

@@ -159,12 +159,8 @@ def do_llamaindex(message_data, metadata, vector_name):
         #    description=description,
         #)
 
-
-def llamaindex_chunker_check(message_data, metadata, vector_name):
-    # llamaindex handles its own chunking/embedding
+def check_llamaindex_in_memory(vector_name):
     memories = load_config_key("memory", vector_name=vector_name, type="vacConfig")
-    total_memories = len(memories)
-    llama = None
     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}")
@@ -172,10 +168,19 @@ def llamaindex_chunker_check(message_data, metadata, vector_name):
             if vectorstore:
                 log.info(f"Found vectorstore {vectorstore}")
                 if vectorstore == "llamaindex":
-                    # https://cloud.google.com/vertex-ai/generative-ai/docs/llamaindex-on-vertexai
-                    log.info(f"llamaindex on vertex indexing for {vector_name}")
-                    llama = do_llamaindex(message_data, metadata, vector_name)
-                    log.info(f"Processed llamaindex: {llama}")
+
+                    return True
+    
+    return False
+
+def llamaindex_chunker_check(message_data, metadata, vector_name):
+    # llamaindex handles its own chunking/embedding
+    memories = load_config_key("memory", vector_name=vector_name, type="vacConfig")
+    total_memories = len(memories)
+    llama = None
+    if check_llamaindex_in_memory(vector_name):
+        llama = do_llamaindex(message_data, metadata, vector_name)
+        log.info(f"Processed llamaindex: {llama}")
 
     # If llamaindex is the only entry, return
     if llama and total_memories == 1:

{sunholo-0.58.0 → sunholo-0.58.2}/sunholo.egg-info/PKG-INFO

@@ -1,9 +1,9 @@
 Metadata-Version: 2.1
 Name: sunholo
-Version: 0.58.0
+Version: 0.58.2
 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/v0.58.0.tar.gz
+Download-URL: https://github.com/sunholo-data/sunholo-py/archive/refs/tags/v0.58.2.tar.gz
 Author: Holosun ApS
 Author-email: multivac@sunholo.com
 License: Apache License, Version 2.0