BigQueryLoggerConfig

When True, skip CHAIN_START/CHAIN_END/CHAIN_ERROR events for chains that are not LangGraph nodes, root invocations, or recognizable user-defined chains. Reduces noise in multi-agent LangGraph deployments where many framework-internal chains (ChannelWrite, RunnableLambda, Branch, ...) fire.

Static key-value pairs attached to every event under attributes.custom_tags. Useful for tagging deployments / cohorts (e.g. {"env": "staging", "agent_role": "sales"}).

When True, dump the user-supplied RunnableConfig metadata (minus keys already projected onto first-class columns) into attributes.session_metadata. Mirrors ADK's session_metadata enrichment so dashboards can correlate by deployment-specific keys (gchat thread-id, customer_id, ...).

Optional hook (raw_content, event_type) -> formatted invoked before content parsing. Use to redact PII or coerce custom payloads to dict/str before they hit the parser.

When True, additively ALTER TABLE ADD COLUMN any new fields that future versions of this handler add to the schema. Gated by a langchain_bq_schema_version table label so the diff runs at most once per schema version. Never drops or renames columns.

When True, automatically CREATE OR REPLACE per-event-type analytics views beside the events table. Each view unnests the JSON columns into typed top-level columns (v_llm_response.usage_total_tokens instead of JSON_VALUE(attributes, '$.usage.total_tokens')).

Prefix for auto-created view names. "v" produces views like v_llm_request, v_tool_completed. Set per-table when several handler instances share one dataset to avoid collisions.