v0.1.1-alpha.9

• Query LLM costs from DB, so that no costs are hard-coded anymore
• Write tool use blocks into outputs too, because previously they were not instrumented and/or only stored as attributes
• Store provider keys encrypted in the DB, so we can run online evals with them
• Sheet view for text fields in the spans/traces view
• Small improvements around the onboarding and self-hosting experience

v0.1.1-alpha.10

Minor fix to redirects for onboarding in local development

v0.1.1-alpha.8

Update the version of Node in Next, fix the build

v0.1.1-alpha.7

This release brings all the features that were previously kept closed source, but now are just disabled runtime with env vars

v0.1.1-alpha.6

Various enhancements:

• Traces page load time and separate input and output tokens
• Better project api key management
• Build for different platforms
• Force pull newest images in docker-compose

v0.1.1-alpha.5

This release brings in:

• Support for Vercel AI SDK
• Some initial rethinking on evaluate events (gone) vs automatic labels
• Input and output costs in the traces table view
• Improved contributing experience and new contributing guide

v0.1.1-alpha.4

Fix build (alpha.3) by upgrading rust to 1.81

v0.1.1-alpha.3

This release revamps evaluation and fixes numerous bugs

• Evaluation dashboard with score visualization and comparison
• Add filters by label values and event values in the traces page
• Return parent span with trace to the frontend, so we can vizualize it in the table later
• Upgrade dependabot alerts
• Image support for openai auto-instrumented logs
• additional front-end revamp
• smoother first time sign-up experience

v0.1.1-alpha.2

This release makes self-hosting easier by pulling pre-built images from ghcr

Evals, datasets, labels

This release brings:

• Traced evaluations
• Ability to label spans
• Revive datasets
• Multiple UI enhancements