Release Notes

This update provides important bug fixes and improvements.

Threads

The UI will depend on the output_index recorded by the server to sort messages in threads instead of the created timestamp.
Introduces ThreadMessage subclass that offers some improvements over the original OpenAI SDK version:
- status can be None for user messages.
- created_at can be float for Next-Gen Assistants.
- output_index: The output index of the message, if applicable for Next-Gen Assistants.
- metadata dictionary: This field can also include boolean values, in addition to strings. This change resolves Pydantic serializer warnings introduced when the boolean is_current_user value was added in the metadata dictionary.
The Next-Gen Assistant Responses API handler will now stop streaming the response when a second assistant message is received on the same run. This change prevents an issue where the model may continue adding similar or identical messages of the same content in a single response.
The server will withhold additional messages from an assistant in a single run when returning thread messages in the web client to prevent the issue in existing runs.

Fixed: Exporting a group's thread messages may fail because of lazy loading code changes introduced.

The docs and openapi.json endpoints are no longer available in non-development environments.

Schema Upgrade	Migration Script	Permissions Update	Task Definition Update	Configuration Update
No	No	No	No	No

feat/ Improve thread message ordering by @ekassos in #1079
internal/ Hide OpenAPI and docs endpoints by @ekassos in #1080
fix/ Unable to export threads because of missing annotations by @ekassos in #1081
feat/ Patch OpenAI multiple response model issue by @ekassos in #1082

Full Changelog: 828-srv399-web245...832-srv403-web246