LlamaIndex v0.14.25 Fixes Security Alerts Across Its AI Developer Integrations

LlamaIndex v0.14.25 Fixes Security Alerts Across Its AI Developer Integrations

AI-generated editorial illustration; not a documentary photograph.

LlamaIndex released version 0.14.25 on September 21, 2026, with changes spanning its core package, model integrations, readers, retrievers and storage connectors. The release notes emphasize security-alert fixes across a large portion of the project, alongside targeted behavior and dependency updates.

Broad security maintenance

The release records security-alert fixes for packages including AgentMesh, Azure, multiple callback integrations, embeddings providers and the instrumentation package. Many other components received dependency maintenance instead, including re-locking 93 manifests affected by stuck Dependabot alerts. The notes do not identify the alerts, their severity or whether a particular vulnerability was exploitable in deployed applications.

Core behavior changes

llama-index-core moves to 0.14.25 and includes several functional changes. It now falls back when a metadata replacement target is None, restores compact and refine streaming, avoids retrying failed function tools and adds native async support to StructuredLLMRerank. The release also adds test coverage for empty arguments in streamed tool calls.

Model and API integrations

Several provider packages received focused updates. llama-index-llms-google-genai 0.11.2 fixes Gemini tool request IDs and preserves images on Gemini function responses, while llama-index-llms-openai 0.8.1 keeps images and files on Responses API tool results. llama-index-llms-perplexity 0.6.1 adds an integration attribution header. The release notes also raise the llama-index-llms-openai-like pin to 0.8.x across 26 consumers.

Compatibility considerations

The update removes the deprecated ipex-llm and optimum-intel IPEX integrations from the core package. Some packages also raise Python requirement floors to avoid keeping vulnerable dependency forks alive. Teams using those integrations or relying on older dependency constraints should review the affected package-specific notes before upgrading; the supplied release entry does not provide migration instructions or a complete compatibility matrix.

Source: LlamaIndex — v0.14.25

Share this article
FacebookThreadsLinkedInX

Leave a Reply

Your email address will not be published. Required fields are marked *