Brief
LlamaIndex 0.14.24: parser, ingestion and memory fixes land in core
LlamaIndex published release notes for v0.14.24 dated 2026-08-19, led by fixes to core splitting, ingestion and tool handling plus a handful of small features. The evidence is a list of one-line changelog entries, so the practical weight of each change is not established here.
The release notes for llama-index 0.14.24, dated 2026-08-19, list changes across llama-index-core and several integrations, including embeddings for Bedrock and VoyageAI and a FalkorDB graph store.
The named features are narrow: property-graph LLM path extractors gain a raise_on_error option, Memory can accept any AsyncDBChatStore, and async support was added for LLMRerank.
Most entries are fixes — CodeSplitter preserving oversized leaf nodes, IngestionPipeline upserts keeping all nodes per document, MetadataFilters defaulting to AND when the condition is None, and tool-call handling no longer marking *args/**kwargs as required parameters.
The notes give one-line summaries only, without reproduction steps or per-change detail, so how much each fix alters behaviour cannot be judged from them.
Our reading
For readers running retrieval or agent pipelines on LlamaIndex, the ingestion and splitting entries are the ones that touch data already in production: a splitter that dropped oversized leaves or an upsert that lost nodes per document would quietly change what a retriever sees. The Memory and async rerank additions matter mainly to teams building long-running or asynchronous agent workflows. Anyo…
What to do or watch
Upgrade in a test environment first and re-run splitting and ingestion over a sample corpus to compare node counts before and after. Watch for a follow-up issue or commit detail if your symptom matches the one-line fix, since the notes alone do not confirm the underlying defect.
Source details and supporting facts
Each line is stated by the page named above it.
Stated by github.com
- Release notes for llama-index-core 0.14.24 are dated 2026-08-19.
- The release includes a fix to preserve oversized leaf nodes in CodeSplitter (#22119).
- The release includes a fix to keep all nodes per document in IngestionPipeline upserts (#22133).
- The release adds raise_on_error to property-graph LLM path extractors (#22195).
- The release allows Memory to accept any AsyncDBChatStore (#22541).
- The release implements async for LLMRerank (#22597).
- The release defaults the MetadataFilters condition to AND when it is None (#22155).
Sources
- LlamaIndex releasesText stored 16 September 2026
How this story was checked. Written from the 1 page listed above, stored 16 September 2026; claims checked against that stored text on 17 September 2026.
What that means
- 7 of 7 reported statements were confirmed against the page that carries them; the rest were removed rather than published.
- Figures in the text were required to appear in the stored source text: yes. Identifiers: yes.
- The check reads stored text only: no claim rests on a fresh look that did not happen.
- Where the reporting was silent, the text says so instead of filling the gap.