Community

Connect your traces.Langfuse.LangSmith.Weave.
Get a PR that saves you money.

Jetty proactively analyzes your traces, finds savings, and delivers pull requests you can trust.

Get your free analysisRead-only access · Data encrypted · Revoke anytime

Recommendations delivered as PRs.
Ready to merge.

No dashboards to check. No reports to read. Jetty delivers pull requests with the reasoning, evidence, and the code change.

Route intent_classification to cheaper gpt-4o-mini modelOpenacme/chatbot #37 · jetty-agent[bot]
Description
Summary
What:Route intent_classification to gpt-4o-mini
Why:Classification doesn't need expensive models
Impact:~$320/mo savings (10-20x reduction)
▶📊 Detailed Analysis & Rationale
▶⚠️ Risk Assessment · LOW
▶✅ Validation Plan
Files Changed (1)
docker/config.example.yaml+4 -1
@@ -12,7 +12,22 @@
models:
default:
alias: litellm_llm.default
model: gpt-4.1
+ simple:
+ alias: litellm_llm.simple
+ model: gpt-4o-mini
pipelines:
- name: sql_generation
llm: litellm_llm.default
@@ -28,6 +43,7 @@
- name: intent_classification
- llm: litellm_llm.default
+ llm: litellm_llm.simple
109K+ traces analyzed
Up to 78% cost reduction
< 1 hr to pull request

Get started in minutes. Jetty does the rest.

Once connected, Jetty works in the background analyzing your traces and delivering optimization PRs. Your team stays focused on building while your LLM costs stay in check.

1Connect your traces

Create an account and connect your trace source. Jetty analyzes your traces in read-only mode — revoke anytime.

2Connect GitHub

Link your repo so Jetty can deliver pull requests directly. Takes less than a minute.

3Get your first PR

Jetty analyzes your traces and delivers your first optimization PR. Review the diff, the savings estimate, and the evidence — merge what makes sense.

∞Jetty keeps working

New traces come in, new savings get found, new pull requests appear — automatically.

Built for trust

Jetty connects in read-only mode. Your data is encrypted, never shared, and deleted when you disconnect.

Read-only access
We only read LLM inputs and outputs. No write access to your systems.
Encrypted in transit & at rest
All data encrypted using industry-standard protocols.
Your data stays yours
Trace data is analyzed in isolation and never shared with third parties or used to train models.
Revoke anytime
Disconnect in one click. We delete your data.

See what you could save

Free during early access. Takes 2 minutes to connect. No credit card required.

Get your free analysis