How to visualise or access manual metrics

[satwikkansal]

Question

I'm recording some values using logfire.metric_histogram method, where exactly can I access these values in Logfire UI (or data tables). And does logfire have capability to plot them on logfire dashboard? If so, how do I go about doing that. Any help will be appreciated!

[alexmojaki]

You can use them by querying the metrics table in the explore view, dashboards, and alerts. We still need to work on making this easier. Here's an example:

import logfire

logfire.configure()

hist = logfire.metric_histogram("my_little_histogram2")

for i in range(10):
    for j in range(100):
        hist.record(i * 100 + j)
    logfire.force_flush()  # export a batch of metrics

There's some summary statistics associated with histograms which are easy to work with:

select histogram_min, histogram_max, histogram_count, histogram_sum
from metrics
where metric_name = 'my_little_histogram2'
order by histogram_min

select min(histogram_min), max(histogram_max), sum(histogram_count), sum(histogram_sum)
from metrics
where metric_name = 'my_little_histogram2'
order by min(histogram_min)

The raw data in histogram buckets is stored in a way that is quite difficult to extract, we definitely need to expose this as a user-friendly view:

with indices as (
    select *, unnest(generate_series(1, array_length(exp_histogram_positive_bucket_counts, 1)::integer)) AS bucket_index
    from metrics
),
counts as (
    select *, exp_histogram_positive_bucket_counts[bucket_index] as bucket_count
    from indices
),
bounds as (
    select
        *,
        power(2, (exp_histogram_positive_bucket_counts_offset+bucket_index-1)/power(2.0, exp_histogram_scale)) as lower_bound,
        power(2, (exp_histogram_positive_bucket_counts_offset+bucket_index  )/power(2.0, exp_histogram_scale)) as upper_bound
    from counts
    where bucket_count > 0
)
select lower_bound, upper_bound, bucket_count from bounds
where metric_name = 'my_little_histogram2'
order by lower_bound

Each row here says that there are bucket_count values between lower_bound and upper_bound inclusive.

Unfortunately we also don't currently have a way to plot these histogram buckets in a dashboard.

[satwikkansal]

I see, thanks! Is it possible to query them in some way via Python?

[alexmojaki]

Yes, with https://logfire.pydantic.dev/docs/guides/advanced/query-api/