Context Propagation

mollog supports context-local structured fields using Python contextvars.

This is useful for:

• request IDs in async services
• trace IDs in worker pools
• tenant or batch IDs in data pipelines
• any field that should follow the current execution flow without being passed manually every time

All context operations live under the Context namespace class.

Bind fields globally for the current context

from mollog import Context, get_logger

logger = get_logger("api")
token = Context.bind(request_id="req-7", trace_id="trace-9")

try:
    logger.info("accepted")
finally:
    Context.reset(token)

Use a scoped context

from mollog import Context, get_logger

logger = get_logger("worker")

with Context.scope(batch="nightly", request_id="req-42"):
    logger.info("started")

Named scopes and logfire spans

When you pass a name to Context.scope(...), it records a scope field in the context and opens a matching logfire span if logfire is configured:

from mollog import Context, get_logger

logger = get_logger("api")

with Context.scope("request", user_id=42):
    logger.info("handling")         # mollog record carries scope="request"
    logger.fire("handled")          # logfire event attached to the span

If logfire isn't configured, the span is a no-op and the scope still works as a plain bind/reset block.

Merge behavior

Field precedence is:

1. context-local fields (Context.bind / Context.scope)
2. logger-bound fields (Logger.bind)
3. per-call extra

Later sources overwrite earlier ones. That means explicit per-call values always win.