add langfuse credentials pulled from secret

Author: strickvl

llm-complete-guide/deployment_hf.py

@@ -8,8 +8,6 @@
 from utils.llm_utils import process_input_with_retrieval
 from zenml.client import Client
 
-langfuse = Langfuse()
-
 logging.basicConfig(level=logging.INFO)
 logger = logging.getLogger(__name__)
 
@@ -27,6 +25,23 @@
     raise RuntimeError(f"Application startup failed: {e}")
 
 
+LANGFUSE_PUBLIC_KEY = os.getenv(
+    "LANGFUSE_PUBLIC_KEY", secret.secret_values["LANGFUSE_PUBLIC_KEY"]
+)
+LANGFUSE_SECRET_KEY = os.getenv(
+    "LANGFUSE_SECRET_KEY", secret.secret_values["LANGFUSE_SECRET_KEY"]
+)
+LANGFUSE_HOST = os.getenv(
+    "LANGFUSE_HOST", secret.secret_values["LANGFUSE_HOST"]
+)
+
+langfuse = Langfuse(
+    public_key=LANGFUSE_PUBLIC_KEY,
+    secret_key=LANGFUSE_SECRET_KEY,
+    host=LANGFUSE_HOST,
+)
+
+
 def get_langfuse_trace_id() -> str | None:
     """Get the trace from Langfuse.
 

llm-complete-guide/steps/rag_deployment.py

@@ -13,16 +13,28 @@
 ZENML_API_TOKEN = os.environ.get("ZENML_API_TOKEN")
 ZENML_STORE_URL = os.environ.get("ZENML_STORE_URL")
 
+secret = Client().get_secret(SECRET_NAME)
+
 if not ZENML_API_TOKEN or not ZENML_STORE_URL:
     # Get ZenML server URL and API token from the secret store
-    secret = Client().get_secret(SECRET_NAME)
     ZENML_API_TOKEN = ZENML_API_TOKEN or secret.secret_values.get(
         "zenml_api_token"
     )
     ZENML_STORE_URL = ZENML_STORE_URL or secret.secret_values.get(
         "zenml_store_url"
     )
 
+
+LANGFUSE_PUBLIC_KEY = os.environ.get(
+    "LANGFUSE_PUBLIC_KEY", secret.secret_values.get("LANGFUSE_PUBLIC_KEY")
+)
+LANGFUSE_SECRET_KEY = os.environ.get(
+    "LANGFUSE_SECRET_KEY", secret.secret_values.get("LANGFUSE_SECRET_KEY")
+)
+LANGFUSE_HOST = os.environ.get(
+    "LANGFUSE_HOST", secret.secret_values.get("LANGFUSE_HOST")
+)
+
 SPACE_USERNAME = os.environ.get("ZENML_HF_USERNAME", "zenml")
 SPACE_NAME = os.environ.get("ZENML_HF_SPACE_NAME", "llm-complete-guide-rag")
 SECRET_NAME = os.environ.get("ZENML_PROJECT_SECRET_NAME", "llm-complete")
@@ -51,6 +63,7 @@
 huggingface-hub
 elasticsearch
 tenacity
+pinecone
 {chr(10).join(gcp_reqs)}
 """
 
@@ -128,6 +141,27 @@ def gradio_rag_deployment() -> None:
             value=str(SECRET_NAME),
         )
 
+    if LANGFUSE_PUBLIC_KEY is not None:
+        api.add_space_secret(
+            repo_id=hf_repo_id,
+            key="LANGFUSE_PUBLIC_KEY",
+            value=str(LANGFUSE_PUBLIC_KEY),
+        )
+
+    if LANGFUSE_SECRET_KEY is not None:
+        api.add_space_secret(
+            repo_id=hf_repo_id,
+            key="LANGFUSE_SECRET_KEY",
+            value=str(LANGFUSE_SECRET_KEY),
+        )
+
+    if LANGFUSE_HOST is not None:
+        api.add_space_secret(
+            repo_id=hf_repo_id,
+            key="LANGFUSE_HOST",
+            value=str(LANGFUSE_HOST),
+        )
+
     files_to_upload = {
         "deployment_hf.py": "app.py",
         "utils/llm_utils.py": "utils/llm_utils.py",

llm-complete-guide/utils/llm_utils.py

@@ -21,7 +21,6 @@
 
 import logging
 import os
-import uuid
 
 import pinecone
 from elasticsearch import Elasticsearch
@@ -285,7 +284,9 @@ def get_db_conn() -> connection:
         raise
 
 
-def get_pinecone_client(model_version_name_or_id: str = "dev") -> pinecone.Index:
+def get_pinecone_client(
+    model_version_name_or_id: str = "dev",
+) -> pinecone.Index:
     """Get a Pinecone index client.
 
     Returns:
@@ -306,7 +307,9 @@ def get_pinecone_client(model_version_name_or_id: str = "dev") -> pinecone.Index
         model_version_name_or_number_or_id=model_version_name_or_id,
     )
 
-    index_name_from_secret = client.get_secret(SECRET_NAME_PINECONE).secret_values.get("pinecone_index", "zenml-docs")
+    index_name_from_secret = client.get_secret(
+        SECRET_NAME_PINECONE
+    ).secret_values.get("pinecone_index", "zenml-docs")
 
     if model_version_name_or_id == "production":
         index_name = f"{index_name_from_secret}-prod"
@@ -322,7 +325,7 @@ def get_pinecone_client(model_version_name_or_id: str = "dev") -> pinecone.Index
             name=index_name,
             dimension=EMBEDDING_DIMENSIONALITY,
             metric="cosine",
-            spec=ServerlessSpec(cloud="aws", region="us-east-1")
+            spec=ServerlessSpec(cloud="aws", region="us-east-1"),
         )
     else:
         try:
@@ -331,7 +334,9 @@ def get_pinecone_client(model_version_name_or_id: str = "dev") -> pinecone.Index
             ]
         except KeyError:
             index_name = index_name_from_secret
-            model_version.run_metadata["vector_store"]["index_name"] = index_name
+            model_version.run_metadata["vector_store"]["index_name"] = (
+                index_name
+            )
 
         # Create index if it doesn't exist
         if index_name not in pc.list_indexes().names():
@@ -341,7 +346,7 @@ def get_pinecone_client(model_version_name_or_id: str = "dev") -> pinecone.Index
                 metric="cosine",
                 spec=ServerlessSpec(cloud="aws", region="us-east-1"),
             )
-        
+
     return pc.Index(index_name)
 
 
@@ -672,7 +677,9 @@ def process_input_with_retrieval(
             include_metadata=True,
         )
     elif vector_store == "pinecone":
-        pinecone_index = get_pinecone_client(model_version_name_or_id=model_version_stage)
+        pinecone_index = get_pinecone_client(
+            model_version_name_or_id=model_version_stage
+        )
         similar_docs = get_topn_similar_docs(
             query_embedding=query_embedding,
             pinecone_index=pinecone_index,