●Change8

v1.80.0.dev6

Breaking Changes
📦 litellmView on GitHub →
⚠ 1 breaking✨ 12 features🐛 13 fixes🔧 12 symbols

Summary

The release adds several UI enhancements, expands Prompt Management capabilities, updates cost maps, and introduces support for new models, while fixing numerous bugs and a breaking change that makes gpt-5 models default to response mode.

⚠️ Breaking Changes

  • All gpt-5 and reasoning models now default to responses mode, which may break code expecting completion responses. To retain previous behavior, explicitly set `response_format='completion'` or use the `model_type='completion'` parameter.

Migration Steps

  1. If you rely on completion responses from gpt-5 or reasoning models, explicitly set `response_format='completion'` or `model_type='completion'` after the default was changed to responses.
  2. Update any custom deployment name handling in cost tracking to align with the new cost tracking logic.
  3. Review Prompt Management usage if you only passed `prompt_id`; ensure versioning works as expected.
  4. No code changes required for the js-yaml version bump.

✨ New Features

  • UI: Added additional model settings to chat models in test key.
  • UI: Deactivated delete button on model table for config models.
  • Added Azure models 2025 11 to cost maps.
  • Added extra_body support for response-supported API params from chat completion.
  • Team vector-store permissions are now respected for key access.
  • Prompt Management: Allow specifying just prompt_id in a request to a model.
  • Prompt Management: Added support for versioning prompts.
  • Responses API: Added support for gpt-5.1-codex model (github-copilot).
  • Prompt Management: Allow storing prompt version in DB.
  • Prompt Management: Added UI for editing the prompts.
  • Migration job labels feature added.
  • LiteLLM key alias used as fallback Noma applicationId in NomaGuardrail.

🐛 Bug Fixes

  • Fixed Swagger UI resolver errors for chat completion endpoints caused by Pydantic v2 `$defs` not being properly exposed in the OpenAPI schema.
  • Mocked Tremor's Tooltip to fix flaky UI tests.
  • Fixed e2e UI Playwright test.
  • Fixed Responses API cost tracking with custom deployment names.
  • Fixed IBM Guardrails optional parameters and added `extra_headers` field.
  • Fixed support for GPT-5.1 reasoning.effort='none' in proxy.
  • Added CSS to truncate long request IDs in request viewer.
  • Trimmed logged response strings in spend logs.
  • Removed `azure/` prefix placeholder for Azure in Add Model UI.
  • Fixed UI edit logging settings not appearing.
  • Ensured consistent chunk IDs in Bedrock streaming responses.
  • Fixed vector store create issue.
  • Fixed NomaGuardrail fallback applicationId using LiteLLM key alias.

🔧 Affected Symbols

openapi_schemaUIResponsesAPIIBMGuardrailscost_mapsproxyspend_logsvector_storebedrock_streamingPromptManagementNomaGuardrailjs_yaml