Releases: epam/ai-dial-core
Releases · epam/ai-dial-core
0.23.0
Features
- Add server-specific files to the request for application with schema while proxying the completion request (#649)
- Support optional field 'audience' in Identity Provider #650 (#652)
Fixes
- Access service validates permission to user appdata in a wrong manner #674 (#676)
- Core fails to list file metadata of the root folder #662 (#663)
- Core fails to process chat completion request to interceptor #660 (#661)
- Publish request fails for quick apps #653 (#654)
- RandomizedWeightedBalancer chooses the first upstream with higher probability than others #666 (#667)
- fixed computation of assembled_response (#648)
- threadlock logs are debug (#673)
- update models with missing parameters (#664)
Other
0.21.3
0.22.3
0.21.2
0.22.2
0.22.1
0.22.0
Features
- Implement temperatureSupport and addonsSupport features to be able to hide these settings fields #637 (#638)
- Setting for tweaking merging strategy for JSON arrays in configs (#622)
- add cache max size setting to ResourceService (#628)
- add viewer url to custom app schemas (#632)
- add viewer url to the schemas listing (#634)
- applications type schemas for DIAL (#575)
- code interpreter (#633)
Fixes
- RouteController is not responsive #623 (#624)
- feature controller (#629)
- improve session id validation (#639)
- invalidate per-request API key in DeploymentFeatureController (#643)
- propagate headers to feature endpoints (#635)
- resource type to support code_interpreter_session (#641)
- revert initial scale for code interpreter (#642)
- use zero initial scale in code interpreter (#640)
Other
0.21.1
0.21.0
Features
- Add customViewState field to conversation api (#599)
- Cut off request/response body if its size exceeds limit #612 (#613)
- Get the Project name from JWT token claims (#605)
- Implement Randomized Weighted Balancer #589 (#596)
- Implement fallback strategy in tier balancer #588 (#593)
- Return aggregated error code if max attempts limit is exceeded #587 (#597)
- Return retry-after header in case if rate limit is exceeded #615 (#617)
- add week/month settings for tokens rate-limiting (#619)
- include query params when rewritePath is enabled (#598)
Fixes
- Application/Model listing API returns wrong value for max_retry_attemps #616 (#618)
- Core sets parent_deployment to a wrong value #530 (#604)
- GfLogStore writes corrupted prompt log record #607 (#608)
- ORT fails with 'No locally installed toolchains match and toolchain download repositories have not been configured.' (#595)
- Prompt logs are not saved for RouteController #611 (#614)
- Rate limit is not updated in RouteController #600 (#601)
- RouteController doesn't mark upstream as failed in case of unsuccessful http code #602 (#603)
- execution path must contain deployment's name only (#610)