v0.12.1

What's Changed

• Using actual usage/token infos from streaming responses when available (else falling back to estimations)
• Several dependency version bumps, esp. fastapi, uvicorn, redis, pyyaml, httpx, azure-identity

Full Changelog: v0.12.0...v0.12.1

v0.12.0

• Fallback to DefaultAzureCredential authentication against Azure OpenAI if received request contains valid API key (configured in PowerProxy's config) but endpoint configuration has no API key configured -> enables config without AOAI key, authentication against AOAI via Managed Identity, Workload Identity, Azure CLI credentials etc.
• Updated GPT-4-Turbo examples to GPT-4o
• Fixes a problem where some streaming responses are not returned as stream
• Version bump for openai and jsonschema

Full Changelog: v0.11.2...v0.12.0

v0.11.2

• Added DefaultAzureCredential authentication to LogUsageToLogAnalytics plugin -> enables support for workload identity authentication (for example when Azure Kubernetes Service/AKS is used)
• Logging for Whisper. Note: Whisper does not have a concept of tokens. Therefore, token counts will be zero for Whisper models.
• Documentation updates

Full Changelog: v0.11.1...v0.11.2

v0.11.1

What's Changed

• Fixes a problem when the openai package sends both API key and Authorization header
• Bump openai package version for tests to latest version (1.35.4)
• Bump redis[hiredis] from 5.0.6 to 5.0.7 by @dependabot in #95

Full Changelog: v0.11.0...v0.11.1

v0.11.0

What's Changed

• Added Azure OpenAI API version to data logged by LogUsage plugin
• Bump azure-identity from 1.17.0 to 1.17.1 by @dependabot in #93

Full Changelog: v0.10.5...v0.11.0

v0.10.5

What's Changed

• Jumps to next endpoint/deployment if HTTP error 408/Gateway Timeout is encountered
• Bump azure-identity from 1.16.1 to 1.17.0 by @dependabot in #91

Full Changelog: v0.10.4...v0.10.5

v0.10.4

What's Changed

• Added option to customize limits and timeouts when connecting to Azure OpenAI (use with care, defaults should be sufficient in most cases, see example config for details)
• Removed console logging for 401 errors
• Bugfix: Error message logging not working for streaming requests
• Multiple version bumps: redis[hiredis], azure-identity, azure-monitor-ingestion

Full Changelog: v0.10.3...v0.10.4

v0.10.3

What's Changed:

• Increased connect timeout

Full Changelog: v0.10.2...v0.10.3

v0.10.2

What's Changed

• Fixed a bug where a 500 HTTP error response was given instead of a 429
• Improved error message for requests that try to use a non-existing virtual deployment (if virtual deployments are configured)
• uvicorn version bumps

Full Changelog: v0.10.1...v0.10.2

v0.10.1

What's Changed

• Update tests to latest openai package 1.30.3
• Improved error message when Entra ID request is sent but PowerProxy's configuration misses a client use in that case
• Test refactorings and minor fixes
• Bump jsonschema from 4.21.1 to 4.22.0 by @dependabot in #78

Full Changelog: v0.10.0...v0.10.1