Streaming Usage Finalization + Public Model Lists

We shipped a focused update to make streaming usage complete, improve SDK compatibility, and surface richer model metadata in the dashboard.

#

Streaming usage is now finalized

Streaming responses can now update execution usage after the stream ends so your token counts, finish reasons, and costs are complete.

What changed:

• Provider‑specific stream parsing is now consistent across OpenAI, Anthropic, and Google.
• Executions now update with the final model, finish reason, and costs once the stream is done.
• Optional stream summary storage can be disabled via STREAM_STORE_SUMMARY=false.

#

Public model list endpoints

Model list endpoints are now public (no auth headers required) to align with how SDKs discover available models:

• GET /v1/openai/models
• GET /v1/openai/{projectId}/models
• GET /v1/anthropic/models
• GET /v1/anthropic/{projectId}/models
• GET /v1/google/models
• GET /v1/google/{projectId}/models

These return the models supported by Proxed, along with display names and pricing metadata.

Example:

curl https://api.proxed.ai/v1/openai/{your-project-id}/models

#

Richer execution details in the dashboard

We now surface model display names, badges, and pricing context in execution tables and detail views, plus faster model filters and selectors.

#

Docs updates

We updated the docs to reflect public model lists and streaming usage finalization:

• https://docs.proxed.ai/api-reference
• https://docs.proxed.ai/authentication

If you have feedback on streaming behavior or model list compatibility, let us know on GitHub.