sourcebot-dev/sourcebot v6.0.2 on GitHub

Removed the SOURCEBOT_EE_AUDIT_LOGGING_ENABLED environment variable. Audit logging is now always enabled, and audit logs are always pruned according to SOURCEBOT_EE_AUDIT_RETENTION_DAYS.
A request whose Authorization token is rejected now falls back to a valid X-Sourcebot-Api-Key header. Previously, only an unknown or expired OAuth token or an invalid Bearer API key fell back.
Upgraded the AI SDK to ai ^6.0.309 and the latest v6 provider packages.
[EE] Knowledge Base stages without a chosen reasoning effort now use high when their model supports it, and Discovery and Overview no longer inherit the Survey effort.

[EE] Added reasoning effort for Amazon Bedrock Claude and OpenAI models in Knowledge Base stages.

Fixed Amazon Bedrock Claude responses being cut off at 4,096 output tokens, with thinking now on by default and a new maxOutputTokens option to override the output limit.
[EE] Fixed Knowledge Base scans retrying output-limit failures and re-running failing full scans every day, with repeatedly failing full scans now backing off while scheduled scans run as delta scans.
[EE] Fixed a React error when choosing a reasoning effort on the Knowledge Base configuration page.
Fixed oversized Ask Sourcebot tool results causing model requests to fail.