●Change8

v0.14.4

📦 llamaindexView on GitHub →
✨ 7 features🐛 10 fixes🔧 12 symbols

Summary

This release introduces support for Claude 4.5 and structured outputs for OpenAI-like models, while primarily addressing dependency issues and bug fixes across the ecosystem.

Migration Steps

  1. Update llama-index-core to 0.14.4 to resolve installation issues.
  2. Update specific provider packages (e.g., anyscale, azure-openai, fireworks) to resolve OpenAI dependency conflicts.

✨ New Features

  • Added support for Anthropic Claude Sonnet 4.5 in Anthropic and Bedrock Converse LLMs.
  • Added structured outputs support for OpenAILike and OpenAI LLMs.
  • Introduced Bedrock AgentCore Memory integration.
  • Introduced ApacheSolrVectorStore integration.
  • Updated MistralAI LLM list.
  • Updated ScrapegraphAI tool integration.
  • Added py.typed to OpenRouter and Anthropic packages for better type checking support.

🐛 Bug Fixes

  • Fixed OpenAI dependency issues across multiple embedding, LLM, tool, and vector store packages.
  • Fixed pre-release installation issues in llama-index-core.
  • Fixed authorization header setup logic in text-embeddings-inference.
  • Fixed ValueError in Google GenAI when ChatMessage contains multiple blocks.
  • Fixed unhashable type error in Ollama stream chat when using tools.
  • Fixed Sarvam integration bugs and typos.
  • Fixed ConfluenceReader to respect the cloud parameter when fetching child pages.
  • Fixed ServiceNow reader issue where pages with empty/null latest content could not be fetched.
  • Fixed index creation logic in Postgres vector store.
  • Fixed playwright tests and ChromaVectorStore docstrings.

🔧 Affected Symbols

AnthropicBedrockConverseOpenAILikeBedrockAgentCoreMemoryApacheSolrVectorStoreOllama.stream_chatConfluenceReaderServiceNowReaderPostgresVectorStoreChromaVectorStore.queryGoogleGenAISarvamLLM