Excluding BigQuery Traces in LangChain with Phoenix Monitoring

Harikrishna D. · 2024-10-27T23:38:42.355Z

Hey I'm currently using Phoenix to monitor my LangGraph Applcation. As it is a sql_agent, I keep getting a bunch of the BigQuery traces. I want to exclude these traces. The connection to BigQuery is a LangChain Toolkit. How do I exclude the BigQuery traces while using the LangChain Instrumentor.

Excluding BigQuery Traces in LangChain with Phoenix Monitoring | Arize AI Community

7 comments

· Sorted by Oldest

RunLLM
APP
·
To exclude specific traces, such as those from BigQuery, while using the LangChain Instrumentor in Phoenix, you can use the suppress_tracing context manager. This allows you to temporarily pause tracing for specific sections of your code. Here's a basic example: LBKSfrom phoenix.trace import suppress_tracingLBKSwith suppress_tracing(): # Code that connects to BigQuery or any other operations you want to exclude # For example, running BigQuery queries here won't generate traces ...LBKSThis method will prevent traces from being generated for the code block within the suppress_tracing context manager. Would you like more details on this process?Sources: - Tracing Core Concepts
💡 Hint: Mention RunLLM in the thread for followups.
Xander S.
·
Hey Harikrishna D., do you know where the Big Query traces are coming from? Do you have some other OTel instrumentation in addition to OpenInference?

Harikrishna D.

{"db": {"name": "ai-projects-431207", "system": "BigQuery"}, "path": "/projects/ai-projects-431207/queries"}

I've created a SQLDataBaseToolkit in my class. As I initiate the class, I'm guessing the traces are logging from here.

class DBAgent:
    def __init__(self,creds,dataset_id):
        self.dataset_id = dataset_id
        self.credentials = service_account.Credentials.from_service_account_file(
            creds
        )
        self.service_account_file = creds
        self.project_id = self.credentials.project_id
        self.client = bigquery.Client(
            project=self.project_id, credentials=self.credentials
        )
        self.sqlalchemy_url = f'bigquery://{self.project_id}/{self.dataset_id}?credentials_path={self.service_account_file}'
        self.db_manager = SQLDatabase.from_uri(self.sqlalchemy_url)
        self.llm_manager = ChatGoogleGenerativeAI(
            model="gemini-1.5-pro-002",
            temperature=0,
            max_tokens=None,
            timeout=None,
            max_retries=2,
            )
        self.toolkit = SQLDatabaseToolkit(db=self.db_manager,llm=self.llm_manager)

Xander S.
·
hey Harikrishna D., are you able to provide the full code snippet including imports
Xander S.
·
I don't believe the OTel instrumentation is configured inside LangChain. Do you import and configure opentelemetry inside your application code?
Xander S.
·
It looks like google-cloud-bigquery automatically creates OTel spans if it finds opentelemetry installed in the runtime. You can try disabling this with
import google.cloud.bigquery.opentelemetry_tracing google.cloud.bigquery.opentelemetry_tracing.HAS_OPENTELEMETRY = False
Alternatively, you can explore defining a custom OTel span processor to filter out the unwanted traces. Hope this helps!
Harikrishna D.
·
This fixed my error. Thanks for the clarity in opentelemetry
👍1

RunLLM
APP
·
To exclude specific traces, such as those from BigQuery, while using the LangChain Instrumentor in Phoenix, you can use the suppress_tracing context manager. This allows you to temporarily pause tracing for specific sections of your code. Here's a basic example: LBKSfrom phoenix.trace import suppress_tracingLBKSwith suppress_tracing(): # Code that connects to BigQuery or any other operations you want to exclude # For example, running BigQuery queries here won't generate traces ...LBKSThis method will prevent traces from being generated for the code block within the suppress_tracing context manager. Would you like more details on this process?Sources: - Tracing Core Concepts
💡 Hint: Mention RunLLM in the thread for followups.
Xander S.
·
Hey Harikrishna D., do you know where the Big Query traces are coming from? Do you have some other OTel instrumentation in addition to OpenInference?

Harikrishna D.

{"db": {"name": "ai-projects-431207", "system": "BigQuery"}, "path": "/projects/ai-projects-431207/queries"}

I've created a SQLDataBaseToolkit in my class. As I initiate the class, I'm guessing the traces are logging from here.

class DBAgent:
    def __init__(self,creds,dataset_id):
        self.dataset_id = dataset_id
        self.credentials = service_account.Credentials.from_service_account_file(
            creds
        )
        self.service_account_file = creds
        self.project_id = self.credentials.project_id
        self.client = bigquery.Client(
            project=self.project_id, credentials=self.credentials
        )
        self.sqlalchemy_url = f'bigquery://{self.project_id}/{self.dataset_id}?credentials_path={self.service_account_file}'
        self.db_manager = SQLDatabase.from_uri(self.sqlalchemy_url)
        self.llm_manager = ChatGoogleGenerativeAI(
            model="gemini-1.5-pro-002",
            temperature=0,
            max_tokens=None,
            timeout=None,
            max_retries=2,
            )
        self.toolkit = SQLDatabaseToolkit(db=self.db_manager,llm=self.llm_manager)

Xander S.
·
hey Harikrishna D., are you able to provide the full code snippet including imports
Xander S.
·
I don't believe the OTel instrumentation is configured inside LangChain. Do you import and configure opentelemetry inside your application code?
Xander S.
·
It looks like google-cloud-bigquery automatically creates OTel spans if it finds opentelemetry installed in the runtime. You can try disabling this with
import google.cloud.bigquery.opentelemetry_tracing google.cloud.bigquery.opentelemetry_tracing.HAS_OPENTELEMETRY = False
Alternatively, you can explore defining a custom OTel span processor to filter out the unwanted traces. Hope this helps!
Harikrishna D.
·
This fixed my error. Thanks for the clarity in opentelemetry
👍1