sunholo 0.57.2__tar.gz → 0.58.2__tar.gz

{sunholo-0.57.2/sunholo.egg-info → sunholo-0.58.2}/PKG-INFO

@@ -1,9 +1,9 @@
 Metadata-Version: 2.1
 Name: sunholo
-Version: 0.57.2
+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/v0.57.2.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
@@ -24,13 +24,19 @@ Requires-Dist: langchain-community
 Provides-Extra: all
 Requires-Dist: asyncpg; extra == "all"
 Requires-Dist: flask; extra == "all"
+Requires-Dist: google-auth; extra == "all"
+Requires-Dist: google-auth-httplib2; extra == "all"
+Requires-Dist: google-auth-oauthlib; extra == "all"
 Requires-Dist: google-cloud-aiplatform; extra == "all"
 Requires-Dist: google-api-python-client; extra == "all"
 Requires-Dist: google-cloud-alloydb-connector[pg8000]; extra == "all"
+Requires-Dist: google-cloud-bigquery; extra == "all"
 Requires-Dist: google-cloud-build; extra == "all"
 Requires-Dist: google-cloud-logging; extra == "all"
 Requires-Dist: google-cloud-storage; extra == "all"
 Requires-Dist: google-cloud-pubsub; extra == "all"
+Requires-Dist: google-cloud-discoveryengine; extra == "all"
+Requires-Dist: google-generativeai; extra == "all"
 Requires-Dist: gunicorn; extra == "all"
 Requires-Dist: httpcore; extra == "all"
 Requires-Dist: httpx; extra == "all"
@@ -44,25 +50,45 @@ Requires-Dist: langchain_google_alloydb_pg; extra == "all"
 Requires-Dist: langchain-anthropic; extra == "all"
 Requires-Dist: langfuse; extra == "all"
 Requires-Dist: pg8000; extra == "all"
+Requires-Dist: pgvector; extra == "all"
+Requires-Dist: psycopg2-binary; extra == "all"
+Requires-Dist: pypdf; extra == "all"
 Requires-Dist: fastapi; extra == "all"
+Requires-Dist: supabase; extra == "all"
+Requires-Dist: tiktoken; extra == "all"
 Requires-Dist: python-socketio; extra == "all"
 Provides-Extra: database
 Requires-Dist: asyncpg; extra == "database"
+Requires-Dist: supabase; extra == "database"
 Requires-Dist: sqlalchemy; extra == "database"
 Requires-Dist: pg8000; extra == "database"
+Requires-Dist: pgvector; extra == "database"
+Requires-Dist: psycopg2-binary; extra == "database"
 Requires-Dist: lancedb; extra == "database"
+Provides-Extra: pipeline
+Requires-Dist: GitPython; extra == "pipeline"
+Requires-Dist: lark; extra == "pipeline"
+Requires-Dist: pypdf; extra == "pipeline"
+Requires-Dist: tabulate; extra == "pipeline"
 Provides-Extra: gcp
+Requires-Dist: google-auth; extra == "gcp"
+Requires-Dist: google-auth-httplib2; extra == "gcp"
+Requires-Dist: google-auth-oauthlib; extra == "gcp"
 Requires-Dist: google-cloud-aiplatform; extra == "gcp"
+Requires-Dist: google-cloud-bigquery; extra == "gcp"
 Requires-Dist: google-cloud-build; extra == "gcp"
 Requires-Dist: google-cloud-storage; extra == "gcp"
 Requires-Dist: google-cloud-logging; extra == "gcp"
 Requires-Dist: google-cloud-pubsub; extra == "gcp"
+Requires-Dist: google-cloud-discoveryengine; extra == "gcp"
+Requires-Dist: google-generativeai; extra == "gcp"
 Requires-Dist: langchain-google-genai; extra == "gcp"
 Requires-Dist: langchain_google_alloydb_pg; extra == "gcp"
 Requires-Dist: google-api-python-client; extra == "gcp"
 Requires-Dist: google-cloud-alloydb-connector[pg8000]; extra == "gcp"
 Provides-Extra: openai
 Requires-Dist: langchain-openai; extra == "openai"
+Requires-Dist: tiktoken; extra == "openai"
 Provides-Extra: anthropic
 Requires-Dist: langchain-anthropic; extra == "anthropic"
 Provides-Extra: http
@@ -73,6 +99,7 @@ Requires-Dist: httpcore; extra == "http"
 Requires-Dist: httpx; extra == "http"
 Requires-Dist: langfuse; extra == "http"
 Requires-Dist: python-socketio; extra == "http"
+Requires-Dist: requests; extra == "http"
 
 ## Introduction
 This is the Sunholo Python project, a comprehensive toolkit for working with language models and vector stores on Google Cloud Platform. It provides a wide range of functionalities and utilities to facilitate the development and deployment of language model applications.

{sunholo-0.57.2 → sunholo-0.58.2}/setup.py

@@ -1,7 +1,7 @@
 from setuptools import setup, find_packages
 
 # Define your base version
-version = '0.57.2'
+version = '0.58.2'
 
 setup(
     name='sunholo',
@@ -37,13 +37,19 @@ setup(
         'all': [
             "asyncpg",
             "flask",
+            "google-auth",
+            "google-auth-httplib2",
+            "google-auth-oauthlib",
             "google-cloud-aiplatform",
             "google-api-python-client",
             "google-cloud-alloydb-connector[pg8000]",
+            "google-cloud-bigquery",
             "google-cloud-build",
             "google-cloud-logging",
             "google-cloud-storage",
             "google-cloud-pubsub",
+            "google-cloud-discoveryengine",
+            "google-generativeai",
             "gunicorn",
             "httpcore",
             "httpx",
@@ -57,21 +63,41 @@ setup(
             "langchain-anthropic",
             "langfuse",
             "pg8000",
+            "pgvector",
+            "psycopg2-binary",
+            "pypdf",
             "fastapi",
+            "supabase",
+            "tiktoken",
             "python-socketio"
         ],
         'database': [
             "asyncpg",
+            "supabase",
             "sqlalchemy",
             "pg8000",
+            "pgvector",
+            "psycopg2-binary",
             "lancedb",
         ],
+        'pipeline': [
+            "GitPython",
+            "lark",
+            "pypdf",
+            "tabulate",
+        ],
         'gcp': [
+            "google-auth",
+            "google-auth-httplib2",
+            "google-auth-oauthlib",
             "google-cloud-aiplatform",
+            "google-cloud-bigquery",
             "google-cloud-build",
             "google-cloud-storage",
             "google-cloud-logging",
             "google-cloud-pubsub",
+            "google-cloud-discoveryengine",
+            "google-generativeai",
             "langchain-google-genai",
             "langchain_google_alloydb_pg",
             "google-api-python-client",
@@ -79,6 +105,7 @@ setup(
         ],
         'openai': [
             "langchain-openai",
+            "tiktoken"
         ],
         'anthropic': [
             "langchain-anthropic",        
@@ -91,6 +118,7 @@ setup(
             "httpx",
             "langfuse",
             "python-socketio",
+            "requests"
         ]
     },
     classifiers=[

{sunholo-0.57.2 → sunholo-0.58.2}/sunholo/__init__.py

@@ -16,6 +16,7 @@ from . import pubsub
 from . import qna
 from . import streaming
 from . import utils
+from . import vertex
 import logging
 
 __all__ = ['agents', 

{sunholo-0.57.2 → sunholo-0.58.2}/sunholo/agents/flask/qna_routes.py

@@ -61,7 +61,6 @@ def register_qna_routes(app, stream_interpreter, vac_interpreter):
                 model=vac_config.get("model") or vac_config.get("llm")
             )
 
-        @observe()
         def generate_response_content():
 
             for chunk in start_streaming_chat(question=all_input["user_input"],

{sunholo-0.57.2 → sunholo-0.58.2}/sunholo/chunker/data_to_embed_pubsub.py

@@ -63,14 +63,14 @@ def data_to_embed_pubsub(data: dict):
     if metadata:
         metadata["vector_name"] = vector_name
 
-    if metadata.get("return_chunks"):
-        log.info("attributes.return_chunks=True detected, skipping process chunks queue")
-        output_list = []
-        if chunks:
-            for chunk in chunks:
-                output_list.append({"page_content": chunk.page_content, "metadata": chunk.metadata})
-                
-        return output_list
+        if metadata.get("return_chunks"):
+            log.info("attributes.return_chunks=True detected, skipping process chunks queue")
+            output_list = []
+            if chunks:
+                for chunk in chunks:
+                    output_list.append({"page_content": chunk.page_content, "metadata": chunk.metadata})
+                    
+            return output_list
 
     process_docs_chunks_vector_name(chunks, vector_name, metadata)
 

sunholo-0.58.2/sunholo/cli/cli.py

@@ -0,0 +1,39 @@
+import argparse
+
+from .configs import setup_list_configs_subparser
+from .deploy import setup_deploy_subparser
+from .cli_init import setup_init_subparser
+
+def main(args=None):
+    """
+    Entry point for the sunholo console script. This function parses command line arguments
+    and invokes the appropriate functionality based on the user input.
+
+    Example commands:
+    ```bash
+    sunholo deploy --config_path . --gcs_bucket your-gcs-bucket --lancedb_bucket your-lancedb-bucket
+    ```
+    """
+    parser = argparse.ArgumentParser(description="sunholo CLI tool for deploying GenAI VACs")
+    subparsers = parser.add_subparsers(title='commands', 
+                                       description='Valid commands', 
+                                       help='Commands', 
+                                       dest='command', 
+                                       required=True)
+
+    # Setup deploy command
+    setup_deploy_subparser(subparsers)
+    # Setup list-configs command
+    setup_list_configs_subparser(subparsers)
+    # init
+    setup_init_subparser(subparsers)
+
+    args = parser.parse_args(args)
+
+    if hasattr(args, 'func'):
+        args.func(args)
+    else:
+        parser.print_help()
+
+if __name__ == "__main__":
+    main()

sunholo-0.58.2/sunholo/cli/cli_init.py

@@ -0,0 +1,80 @@
+import os
+import shutil
+from ..utils.config import get_module_filepath
+
+def init_project(args):
+    """
+    Initializes a new sunholo project with a basic configuration file and directory structure.
+
+**Explanation:**
+
+1. **Import Necessary Modules:**
+   - `os` for file system operations.
+   - `shutil` for copying files and directories.
+   - `log` from `sunholo.logging` for logging messages.
+   - `get_module_filepath` from `sunholo.utils.config` to get the absolute path of template files.
+
+2. **`init_project` Function:**
+   - Takes an `args` object from argparse, containing the `project_name`.
+   - Creates the project directory using `os.makedirs`.
+   - Copies template files from the `templates/project` directory to the new project directory using `shutil.copy` and `shutil.copytree`.
+   - Logs informative messages about the initialization process.
+
+3. **`setup_init_subparser` Function:**
+   - Sets up the `init` subcommand for the `sunholo` CLI.
+   - Adds an argument `project_name` to specify the name of the new project.
+   - Sets the `func` attribute to `init_project`, so the parser knows which function to call when the `init` command is used.
+
+**Template Files (`templates/project`):**
+
+You'll need to create a `templates/project` directory within your `sunholo` package and place the following template files in it:
+
+* **`config/llm_config.yaml`:** A basic configuration file with placeholders for LLM settings, vector stores, etc.
+* **`config/cloud_run_urls.json`:** A template for Cloud Run URLs.
+* **`app.py`:** A basic Flask app that can be customized for the project.
+* **`.gitignore`:** A gitignore file to exclude unnecessary files from version control.
+* **`README.md`:** A README file with instructions for setting up and running the project.
+
+**Usage:**
+
+After adding this code to your `cli.py` and creating the template files, users can initialize a new project using the following command:
+
+```bash
+sunholo init my_genai_project
+```
+
+This will create a new directory named `my_genai_project` with the template files, allowing users to start building their GenAI application.
+
+    """
+    project_name = args.project_name
+    project_dir = os.path.join(os.getcwd(), project_name)
+
+    print(f"Initializing project: {project_name} in directory: {project_dir}")
+
+    # Create project directory
+    if os.path.exists(project_dir):
+        print(f"Directory {project_dir} already exists. Please choose a different project name.")
+        return
+
+    os.makedirs(project_dir)
+
+    # Copy template files
+    template_dir = get_module_filepath("templates/project")
+    for filename in os.listdir(template_dir):
+        src_path = os.path.join(template_dir, filename)
+        dest_path = os.path.join(project_dir, filename)
+        if os.path.isfile(src_path):
+            shutil.copy(src_path, dest_path)
+        elif os.path.isdir(src_path):
+            shutil.copytree(src_path, dest_path)
+
+    print(f"Project {project_name} initialized successfully.")
+    print(f"Navigate to {project_dir} and customize the configuration files in the 'config' directory.")
+
+def setup_init_subparser(subparsers):
+    """
+    Sets up an argparse subparser for the 'init' command.
+    """
+    init_parser = subparsers.add_parser('init', help='Initializes a new sunholo project.')
+    init_parser.add_argument('project_name', help='The name of the new project.')
+    init_parser.set_defaults(func=init_project)

sunholo-0.58.2/sunholo/cli/configs.py

@@ -0,0 +1,29 @@
+from ..utils.config import load_all_configs
+
+from pprint import pprint
+
+def list_configs(args):
+    """
+    Lists configuration files, filtered by kind if specified.
+    """
+    print("Listing configuration files")
+    configs = load_all_configs()
+
+    if args.kind:
+        if args.kind in configs:
+            print(f"## Config kind: {args.kind}")
+            pprint(configs[args.kind])
+        else:
+            print(f"No configurations found for kind: {args.kind}")
+    else:
+        for kind, config in configs.items():
+            pprint(f"## Config kind: {kind}") 
+            pprint(config)
+
+def setup_list_configs_subparser(subparsers):
+    """
+    Sets up an argparse subparser for the 'list-configs' command.
+    """
+    list_configs_parser = subparsers.add_parser('list-configs', help='Lists all configuration files and their details.')
+    list_configs_parser.add_argument('--kind', help='Filter configurations by kind.')
+    list_configs_parser.set_defaults(func=list_configs)

sunholo-0.58.2/sunholo/cli/deploy.py

@@ -0,0 +1,43 @@
+import os
+from subprocess import Popen
+from ..utils.config import load_all_configs
+
+def deploy_vac(args):
+    """
+    Deploys the VAC by running a Flask app locally.
+    """
+    print(f"Deploying VAC: {args.vac_name} locally")
+
+    # Load the vacConfig
+    configs_by_kind = load_all_configs()
+    vac_config = configs_by_kind.get('vacConfig', {}).get('vac', {}).get(args.vac_name)
+
+    if not vac_config:
+        raise ValueError(f"No configuration found for VAC: {args.vac_name}")
+
+    # Assuming the Flask app is in 'app.py' within the config path
+    app_path = os.path.join(args.config_path, 'app.py')
+    if not os.path.exists(app_path):
+        raise ValueError(f"app.py not found in {args.config_path}")
+
+    print(f"Running Flask app from {app_path}")
+
+    # Run the Flask app
+    command = ["python", app_path]
+    print(f"Running Flask app with command: {' '.join(command)}")
+    process = Popen(command)
+    process.communicate()
+
+def setup_deploy_subparser(subparsers):
+    """
+    Sets up an argparse subparser for the 'deploy' command.
+
+    Example command:
+    ```bash
+    sunholo deploy "vac_name" --config_path .
+    ```
+    """
+    deploy_parser = subparsers.add_parser('deploy', help='Triggers a deployment of a VAC.')
+    deploy_parser.add_argument('vac_name', help='The name of the VAC to deploy.')
+    deploy_parser.add_argument('--config_path', default='.', help='Path to the directory containing the config folder `config/` and Flask app `app.py`, defaults to current directory.  Set _CONFIG_FOLDER env var to change config location.')
+    deploy_parser.set_defaults(func=deploy_vac)

{sunholo-0.57.2 → 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-0.57.2 → 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-0.57.2 → sunholo-0.58.2}/sunholo/llamaindex/import_files.py

@@ -1,54 +1,14 @@
-
 try:
     from vertexai.preview import rag
-    from vertexai.preview.generative_models import GenerativeModel, Tool
-    import vertexai
 except ImportError:
     rag = None
 
 from ..logging import log
 from ..utils.config import load_config_key
+from ..vertex import init_vertex
 
 # Create a RAG Corpus, Import Files
 
-def init_vertex(gcp_config):
-    """
-    Initializes the Vertex AI environment using the provided Google Cloud Platform configuration.
-
-    This function configures the Vertex AI API session with specified project and location details
-    from the gcp_config dictionary. It is essential to call this function at the beginning of a session
-    before performing any operations related to Vertex AI.
-
-    Parameters:
-        gcp_config (dict): A dictionary containing the Google Cloud Platform configuration with keys:
-            - 'project_id': The Google Cloud project ID to configure for Vertex AI.
-            - 'location': The Google Cloud region to configure for Vertex AI.
-
-    Raises:
-        KeyError: If the necessary keys ('project_id' or 'location') are missing in the gcp_config dictionary.
-        ModuleNotFoundError: If the Vertex AI module is not installed and needs to be installed via pip.
-
-    Example:
-    ```python
-    gcp_config = {
-         'project_id': 'your-project-id',
-         'location': 'us-central1'
-    }
-    init_vertex(gcp_config)
-    # This will initialize the Vertex AI session with the provided project ID and location.
-
-    Note:
-        Ensure that the 'vertexai' module is installed and correctly configured before calling this function.
-        The function assumes that the required 'vertexai' library is available and that the logging setup is already in place.
-    """
-    if not rag:
-        log.error("Need to install vertexai module via `pip install google-cloud-aiplatform`")
-    
-    # Initialize Vertex AI API once per session
-    project_id = gcp_config.get('project_id')
-    location = gcp_config.get('location')
-    vertexai.init(project=project_id, location=location)
-
 def get_corpus(gcp_config):
     """
     Retrieves a LlamaIndex corpus from Vertex AI based on the provided Google Cloud configuration.
@@ -86,6 +46,9 @@ def get_corpus(gcp_config):
         print("Error fetching corpus:", str(e))
     ```
     """
+    if not rag:
+        raise ValueError("Need to install vertexai module via `pip install sunholo[gcp]`")
+    
     project_id = gcp_config.get('project_id')
     location = gcp_config.get('location')
     rag_id = gcp_config.get('rag_id')
@@ -136,7 +99,10 @@ def do_llamaindex(message_data, metadata, vector_name):
     # Imported file to corpus: {'status': 'success'}
     ```
     """
-    gcp_config = load_config_key("gcp_config", vector_name=vector_name, filename = "config/llm_config.yaml")
+    if not rag:
+        raise ValueError("Need to install vertexai module via `pip install sunholo[gcp]`")
+
+    gcp_config = load_config_key("gcp_config", vector_name=vector_name, type="vacConfig")
     if not gcp_config:
         raise ValueError(f"Need config.{vector_name}.gcp_config to configure llamaindex on VertexAI")
 
@@ -154,7 +120,7 @@ def do_llamaindex(message_data, metadata, vector_name):
     log.info(f"Found llamaindex corpus: {corpus}")
 
     # native support for cloud storage and drive links
-    chunker_config = load_config_key("chunker", vector_name=vector_name, filename="config/llm_config.yaml")
+    chunker_config = load_config_key("chunker", vector_name=vector_name, type="vacConfig")
 
     if message_data.startswith("gs://") or message_data.startswith("https://drive.google.com"):
         log.info(f"rag.import_files for {message_data}")
@@ -193,12 +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
-    memories = load_config_key("memory", vector_name=vector_name, filename = "config/llm_config.yaml")
-    total_memories = len(memories)
-    llama = None
+def check_llamaindex_in_memory(vector_name):
+    memories = load_config_key("memory", vector_name=vector_name, type="vacConfig")
     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}")
@@ -206,10 +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-0.57.2 → sunholo-0.58.2}/sunholo/logging.py

@@ -249,4 +249,12 @@ def log_folder_location(folder_name):
     else:
         logging.warning(f"The folder '{folder_name}' does not exist in the current working directory: {current_working_directory}")
 
-log = setup_logging("sunholo")
+# lazy eval
+_logger = None
+def get_logger():
+    global _logger
+    if _logger is None:
+        _logger = setup_logging("sunholo")
+    return _logger
+
+log = get_logger()