Skip to content

Define the component first in the README #35

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
kwisatzh merged 1 commit into from
Oct 10, 2026
Merged

Define the component first in the README #35

Changes from all commits
Commits
Show all changes
1 commit
Select commit
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
27 changes: 15 additions & 12 deletions README.md
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -2,26 +2,29 @@

[![OpenSSF Scorecard](https://api.scorecard.dev/projects/github.com/llm-measurement/fleetdiff/badge)](https://scorecard.dev/viewer/?uri=github.com/llm-measurement/fleetdiff)

**See what changed in your agent application, using summaries instead of raw traces.**
fleetdiff is a command-line tool that compares summary files from LLM and agent
traffic and reports what changed between two time windows: whether token use rose
from more requests or larger ones, which users or sessions drove the change, and
whether usage data is complete. It reads summaries exported by the
[otelcol-genai-sketches](https://github.com/llm-measurement/otelcol-genai-sketches)
collector or any [llm-sketchkit](https://github.com/llm-measurement/llm-sketchkit)
producer. It runs locally and only reads files: no account, upload or model API key.

| | |
| --- | --- |
| Commands | `investigate` (one application, two windows), `scan` (unusual windows over history), `compare` (several systems), `inspect` (check an OTLP capture), `diagnose` (review a collector config) |
| Install | Signed release binaries for Linux and macOS (amd64 and arm64), or `go install github.com/llm-measurement/fleetdiff/cmd/fleetdiff@latest` |
| Status | Pre-1.0; see the [changelog](CHANGELOG.md) |
| License | Apache-2.0 |

![Sessions demo: one of eight tracked sessions is flagged for review, accounting for 90.91% of attributed tokens.](docs/media/sessions.gif)

[Run this demo](examples/sessions/README.md) | [Read the transcript](docs/media/sessions-transcript.txt)

Did token usage rise because you made more requests, or because each request
used more tokens? Which tracked contributors changed? Is the comparison missing
usage data? fleetdiff reads small summary files and answers locally. Start with
one application; combine compatible exports when you add workers or separately
operated systems. Keep your existing trace backend. No account, upload, or model
API key is needed.
## Try It In A Minute

Comparison inputs are summary exports from
[otelcol-genai-sketches](https://github.com/llm-measurement/otelcol-genai-sketches)
or [llm-sketchkit](https://github.com/llm-measurement/llm-sketchkit).
The `inspect` command checks a local OTLP capture before you set up summary export.

## Try It In A Minute

```sh
git clone https://github.com/llm-measurement/fleetdiff.git
cd fleetdiff
Expand Down
Loading