By default, all logs within a project are stored in the default context (represented by the empty string "").

unify.log(x=0, y=1, z=2)

When no context is specified in the interface or the table, then the default context is used. However, you can group your logs into non-overlapping contexts, which can then be displayed in separate tables.

names = ["Zoe", "John", "Jane", "Jim", "Jill"]
qnas = {
    "what is 1 + 1?": 2,
    "what is 2 * 3?": 6,
    "what is 3 - 2?": 1,
    "what is 4 / 2?": 2,
    "what is 5 % 2?": 1
}
regions = ["US", "EU", "UK", "AU", "CA"]

for _ in range(20):
    question, answer = random.choice(
        list(qnas.items())
    )
    unify.log(
        name=random.choice(names),
        question=question,
        answer=answer,
        region=random.choice(regions),
        context="Traffic"
)

predictions = [
    "six", "one", "two", "three", "four", "five"
]

for _ in range(15):
    question, answer = random.choice(
        list(qnas.items())
    )
    unify.log(
        name=random.choice(names),
        question=question,
        prediction=random.choice(
            predictions
        ),
        answer=answer,
        context="Evals"
    )

​
Context Handling

You can use with unify.Context("..."): to create a context, and all logging inside this block will adopt this context.

with unify.Context("Traffic"):
    unify.log(
        name="Zoe",
        question="what is 1 + 1?",
        region="US"
    )
    unify.log(
        name="John",
        question="what is 2 + 2?",
        region="EU"
    )

IMG

These can also be arbitrarily nested, where each context string is joined via / to define the full context path.

with unify.Context("Sciences"):
    with unify.Context("Maths"):
        unify.log(
            name="Zoe",
            question="what is 1 + 1?",
            region="US"
        )
    with unify.Context("Physics"):
        unify.log(
            name="John",
            question="what is the speed of light?",
            region="EU"
        )

IMG

By default, unify.Context("...") also changes the behaviour of get_logs, returning only the logs relevant within the context.

with unify.Context("Sciences"):
    with unify.Context("Maths"):
        unify.log(
            name="Zoe",
            question="what is 1 + 1?",
            region="US"
        )
        assert len(unify.get_logs()) == 1
    with unify.Context("Physics"):
        unify.log(
            name="John",
            question="what is the speed of light?",
            region="EU"
        )
        assert len(unify.get_logs()) == 1
assert len(unify.get_logs()) == 2

This behaviour can be change by setting unify.Context("...", mode="read") which will only set the context for log getting, and unify.Context("...", mode="write") which will only set the context for log setting.

This makes it easy to quickly get/set different logs for different purposes throughout your codebase, without needing to manage contexts via local variables.

​
Shared Logs

Logs can be added to as many contexts as you’d like, without duplication.

exam_results = [
    {"name": "Zoe", "passed": False},
    {"name": "John", "passed": True}
]
with unify.Context("Results"):
    logs = unify.create_logs(entries=exam_results)
with unify.Context("Failed"):
    unify.add_logs_to_context(
        [l.id for l in logs if not l.entries["passed"]]
    )

If any logs are then updated inplace, they will be reflected in all contexts.

[l.update(entries={"passed": True}) for l in logs]

GIF

Note that in this case, Results and Failed are disconnected, and both sets of logs cannot be viewed in a single table.

To organize data in a single table, you can use column contexts, more on those soon!

LogsDatasets