sunholo 0.56.7__tar.gz → 0.57.1__tar.gz

{sunholo-0.56.7 → sunholo-0.57.1}/PKG-INFO

@@ -1,9 +1,9 @@
 Metadata-Version: 2.1
 Name: sunholo
-Version: 0.56.7
+Version: 0.57.1
 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/v0.56.7.tar.gz
+Download-URL: https://github.com/sunholo-data/sunholo-py/archive/refs/tags/v0.57.1.tar.gz
 Author: Holosun ApS
 Author-email: multivac@sunholo.com
 License: Apache License, Version 2.0

{sunholo-0.56.7 → sunholo-0.57.1}/setup.py

@@ -1,7 +1,7 @@
 from setuptools import setup, find_packages
 
 # Define your base version
-version = '0.56.7'
+version = '0.57.1'
 
 setup(
     name='sunholo',

{sunholo-0.56.7 → sunholo-0.57.1}/sunholo/agents/flask/qna_routes.py

@@ -29,9 +29,16 @@ try:
 except ImportError:
     print("No flask installed for agents.flask.register_qna_routes, install via `pip install sunholo[http]`")
 
+try:
+    from langfuse.decorators import langfuse_context, observe
+except ImportError as err:
+    print(f"No langfuse installed for agents.flask.register_qna_routes, install via `pip install sunholo[http]` - {str(err)}")
+    
+
 def register_qna_routes(app, stream_interpreter, vac_interpreter):
 
     @app.route('/vac/streaming/<vector_name>', methods=['POST'])
+    @observe()
     def stream_qa(vector_name):
         prep = prep_vac(request, vector_name)
         log.debug(f"Processing prep: {prep}")
@@ -54,6 +61,7 @@ def register_qna_routes(app, stream_interpreter, vac_interpreter):
                 model=vac_config.get("model") or vac_config.get("llm")
             )
 
+        @observe()
         def generate_response_content():
 
             chunks = ""
@@ -77,9 +85,9 @@ def register_qna_routes(app, stream_interpreter, vac_interpreter):
                         generation.end(output=json.dumps(chunk))
                         span.end(output=json.dumps(chunk))
                         trace.update(output=json.dumps(chunk))
-                    yield f"###JSON_START###{json.dumps(chunk)}###JSON_END###"
   
-                    return
+                    return json.dumps(chunk)
+                
                 else:
                     # Otherwise, we yield the plain text chunks as they come in.
                     if chunk:
@@ -100,6 +108,7 @@ def register_qna_routes(app, stream_interpreter, vac_interpreter):
         return response
 
     @app.route('/vac/<vector_name>', methods=['POST'])
+    @observe()
     def process_qna(vector_name):
         prep = prep_vac(request, vector_name)
         log.debug(f"Processing prep: {prep}")
@@ -177,7 +186,9 @@ def create_langfuse_trace(request, vector_name):
     )
 
 def prep_vac(request, vector_name):
-    trace = create_langfuse_trace(request, vector_name)
+    #trace = create_langfuse_trace(request, vector_name)
+    trace = None
+    span = None
     data = request.get_json()
     log.info(f"vac/{vector_name} got data: {data}")
     config, _ = load_config("config/llm_config.yaml")

{sunholo-0.56.7 → sunholo-0.57.1}/sunholo/chunker/splitter.py

@@ -32,12 +32,13 @@ def chunk_doc_to_docs(documents: list, extension: str = ".md", min_size: int = 8
     docstore_doc_id, documents = send_doc_to_docstore(documents, vector_name=vector_name)
 
     doc_summaries = summarise_docs(documents, vector_name=vector_name)
-    for doc in documents:
-        # Assuming each doc has a unique identifier in its metadata under 'objectId'
-        objectId = doc.metadata.get("objectId")
-        if objectId and objectId in doc_summaries:
-            # If the objectId is found in doc_summaries, add the summary location to the document's metadata
-            doc.metadata["summary_location"] = doc_summaries[objectId]
+    if doc_summaries:
+        for doc in documents:
+            # Assuming each doc has a unique identifier in its metadata under 'objectId'
+            objectId = doc.metadata.get("objectId")
+            if objectId and objectId in doc_summaries:
+                # If the objectId is found in doc_summaries, add the summary location to the document's metadata
+                doc.metadata["summary_location"] = doc_summaries[objectId]
 
     # Combine entire documents that are smaller than min_size
     combined_documents_content = ""

{sunholo-0.56.7 → sunholo-0.57.1}/sunholo/cli/cli.py

@@ -1,8 +1,8 @@
 import argparse
 try:
-    from google.cloud import build_v1
+    from google.cloud.devtools import cloudbuild_v1
 except ImportError:
-    build_v1 = None
+    cloudbuild_v1 = None
 
 from ..logging import log
 
@@ -16,21 +16,21 @@ def trigger_build(args):
     Example:
         trigger_build(args) where args contains project_id, trigger_id, repo_name, and branch_name.
     """
-    if not build_v1:
+    if not cloudbuild_v1:
         log.warning("Can't deploy - google-cloud-build not installed, enable via `pip install sunholo[gcp]")
 
         return None
 
-    client = build_v1.services.cloud_build.CloudBuildClient()
+    client = cloudbuild_v1.CloudBuildClient()
     # Assuming the build source uses the path to the cloudbuild.yaml if specified.
-    source = build_v1.RepoSource(
+    source = cloudbuild_v1.RepoSource(
         project_id=args.project_id, 
         repo_name=args.repo_name, 
         branch_name=args.branch_name,
         substitutions=args.substitutions,
         dir=args.config_path  # Path to directory containing cloudbuild.yaml
     )
-    request = build_v1.RunBuildTriggerRequest(
+    request = cloudbuild_v1.RunBuildTriggerRequest(
         project_id=args.project_id, 
         trigger_id=args.trigger_id, 
         source=source
@@ -62,7 +62,7 @@ def main(args=None):
 
     Example commands:
     ```bash
-    sunholo deploy --project_id "my-gcp-project" --trigger_id "my-trigger-id" --repo_name "my-repo" --branch_name "dev" --config_path "app/vac/my_vac/"
+    sunholo deploy --vac edmonbrain --config=llm_config.yaml
     ```
     """
     parser = argparse.ArgumentParser(description="sunholo CLI tool for deploying applications using Google Cloud Build.")

{sunholo-0.56.7 → sunholo-0.57.1}/sunholo/langfuse/prompts.py

@@ -3,6 +3,20 @@ from ..utils import load_config_key
 
 # Load the YAML file
 def load_prompt_from_yaml(key, prefix="sunholo", file_path=None):
+    """
+    Returns a string you can use with Langfuse PromptTemplate.from_template() 
+
+    Will first try to load from the Langfuse prompt library, if unavailable will look in promptConfig type file.
+
+    Langfuse prompts have {{ two braces }}, Langchain prompts have { one brace }.
+
+    Example:
+
+    ```python
+    from sunholo.langfuse.prompts import load_prompt_from_yaml
+    from langchain_core.prompts import PromptTemplate
+    
+    """
     from langfuse import Langfuse
 
     # Initialize Langfuse client

{sunholo-0.56.7 → sunholo-0.57.1}/sunholo/qna/parsers.py

@@ -21,18 +21,35 @@ def document_to_dict(document):
     }
 
 def parse_output(bot_output):
+    """
+    Parses VAC output assuming it has an 'answer' and an optional 'source_documents' key
+    
+    """
     if isinstance(bot_output, str):
+
         return {"answer": bot_output}
+    
     if isinstance(bot_output, dict) and 'source_documents' in bot_output:
-        if 'source_documents' in bot_output:
-            bot_output['source_documents'] = [document_to_dict(doc) for doc in bot_output['source_documents']]
-        if bot_output.get("answer", None) is None or bot_output.get("answer") == "":
+        bot_output['source_documents'] = [document_to_dict(doc) for doc in bot_output['source_documents']]
+        if not bot_output.get("answer") or bot_output.get("answer") == "":
+            bot_output['answer'] = "(No text was returned)"
+
+        return bot_output
+    
+    elif isinstance(bot_output, dict) and 'metadata' in bot_output and isinstance(bot_output.get('metadata')) and 'source_documents' in bot_output.get('metadata'):
+        metadata = bot_output.get('metadata')
+        bot_output['source_documents'] = [document_to_dict(doc) for doc in metadata['source_documents']]
+        if not bot_output.get("answer") or bot_output.get("answer") == "":
             bot_output['answer'] = "(No text was returned)"
+
         return bot_output
+    
     elif isinstance(bot_output, dict):
         if not bot_output.get("answer"):
             raise ValueError(f"VAC output was not a string or a dict with the key 'answer' - got: {bot_output}")
         else:
+
             return bot_output
+        
     else:
         log.error(f"Couldn't parse output for:\n {bot_output}")

{sunholo-0.56.7 → sunholo-0.57.1}/sunholo/streaming/content_buffer.py

@@ -51,7 +51,6 @@ class ContentBuffer:
         Adds the given text to the existing content of the buffer.
         """
         self.content += text
-        log.debug(f"Written {text} to buffer")
     
     def read(self) -> str:
         """
@@ -61,8 +60,7 @@ class ContentBuffer:
             str: The content of the buffer.
 
         Provides the entire content stored in the buffer.
-        """
-        log.debug(f"Read content from buffer")    
+        """   
         return self.content
 
     def clear(self):
@@ -71,7 +69,6 @@ class ContentBuffer:
 
         Empties the buffer content, resetting it to an empty string.
         """
-        log.debug(f"Clearing content buffer")
         self.content = ""
 
     

{sunholo-0.56.7 → sunholo-0.57.1}/sunholo/streaming/streaming.py

@@ -39,8 +39,6 @@ def start_streaming_chat(question,
     if not check_kwargs_support(qna_func):
         yield "No **kwargs in qna_func - please add it"
 
-    # Immediately yield to indicate the process has started.
-    yield "Thinking...\n"
     log.info(f"Streaming chat with wait time {wait_time} seconds and timeout {timeout} seconds and kwargs {kwargs}")
     # Initialize the chat
     content_buffer = ContentBuffer()
@@ -82,12 +80,6 @@ def start_streaming_chat(question,
             yield content_to_send
             content_buffer.clear()
             start = time.time() # reset timeout
-        else:
-            if time.time() - first_start < wait_time:
-                # If the initial wait period hasn't passed yet, keep sending "..."
-                yield "..."
-            else:
-                log.info("No content to send")
 
         elapsed_time = time.time() - start
         if elapsed_time > timeout: # If the elapsed time exceeds the timeout
@@ -162,9 +154,10 @@ async def start_streaming_chat_async(question, vector_name, qna_func, chat_histo
     The `start_streaming_chat_async` function is used in a coroutine that prints messages as they are generated.
     """
     
-    yield "Thinking...\n"
     log.info(f"Streaming chat with wait time {wait_time} seconds and timeout {timeout} seconds and kwargs {kwargs}")
-
+    if not check_kwargs_support(qna_func):
+        yield "No **kwargs in qna_func - please add it"
+        
     content_buffer = ContentBuffer()
     chat_callback_handler = BufferStreamingStdOutCallbackHandler(content_buffer=content_buffer, tokens=".!?\n")
 

{sunholo-0.56.7 → sunholo-0.57.1}/sunholo.egg-info/PKG-INFO

@@ -1,9 +1,9 @@
 Metadata-Version: 2.1
 Name: sunholo
-Version: 0.56.7
+Version: 0.57.1
 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/v0.56.7.tar.gz
+Download-URL: https://github.com/sunholo-data/sunholo-py/archive/refs/tags/v0.57.1.tar.gz
 Author: Holosun ApS
 Author-email: multivac@sunholo.com
 License: Apache License, Version 2.0