What LlamaIndex shipped
Each line links to LlamaIndex's own note; the short summaries are FoxPlug's. Not affiliated with LlamaIndex.
Last 90 days, as of November 21, 2026
Changes to act on (last 90 days)
- November 13, 2026 · Breaking change · Breaking change: Version v0.14.26 was released on 13 Nov at 22:38
LlamaIndex is at the low end of the 9 ai infrastructure products we track for how often it ships: 9th of 9, at 0.3 entries a week over the days its feed covers. How the rank is worked out.
November 2026
Python security relocks and a raised requires-python breaking change affected the AgentMesh, Argilla and Arize Phoenix integration packages.
- November 13, 2026 · v0.14.26
Integration lock files were relocked for Python security fixes, and requires-python was raised across integrations in a breaking change. AgentMesh, Argilla and Arize Phoenix packages are affected.
September 2026
- September 21, 2026 · v0.14.25
Security alerts were resolved across numerous integration packages, including AgentMesh, Azure, Argilla, Arize Phoenix, HoneyHive and Langfuse.
August 2026
- August 19, 2026 · v0.14.24
Core fixes address closing HotpotQA evaluator file handles, preserving oversized leaf nodes in CodeSplitter, multiblock chat history writes, skipping empty nodes in HTMLNodeParser, honoring mmr_threshold=0, keeping all nodes per document in upserts, and tool p
June 2026
- June 24, 2026 · v0.14.23
Dependency bumps were applied across many directories for the uv and pip groups in several integration packages, including Argilla callbacks.
May 2026
- May 14, 2026 · v0.14.22
Lock files were upgraded across integrations, with dependency bumps in the pip group across many directories.
April 2026
- April 21, 2026 · v0.14.21
Dependency bumps were applied to the pip group across many directories, and core fixes cover a DocumentSummaryIndex.delete_nodes KeyError, structured output failures, UTF-8 encoding in the persistence layer, and message block buffer resolution.
- April 3, 2026 · v0.14.20
An nltk vulnerability was fixed across integration packages, alongside routine dependency bumps in the uv group.
March 2026
- March 25, 2026 · v0.14.19
Version 0.14.19 released. The entry lists only the version number and date, with no further detail on user-facing changes.
- March 16, 2026 · v0.14.18
Python 3.9 support was deprecated across integration packages including AgentMesh, Azure, AgentOps, Argilla, Arize Phoenix and HoneyHive.
- March 10, 2026 · v0.14.16
Core adds a token-bucket rate limiter for API calls and a multimodal LLM reranker, with an optional
embed_modelon SemanticDoubleMergingSplitterNodeParser. Fixes covertool_choicepassthrough and async retry backoff.
February 2026
- February 18, 2026 · v0.14.15
Core adds basic operations for multimodal types, recursive type support, and multimodal prompt templates and template variable formatting. A redundant
metadata_seperatorfield was removed from TextNode. - February 10, 2026 · v0.14.14
Core catches pydantic ValidationError in VectorStoreQueryOutputParser, distinguishes empty strings from None in MediaResource.hash, and falls back to a bundled nltk cache. Adds LangChain 1.x support and a TokenBudgetHandler callback.
January 2026
- January 21, 2026 · v0.14.13
Core adds an
early_stopping_methodparameter to agent workflows, token-based code splitting in CodeSplitter, and a RayIngestionPipeline integration for distributed ingestion. ChatMemoryBuffer is replaced with Memory.
December 2025
- December 30, 2025 · v0.14.12
Core gains async tool spec support, an element node parser, and improvements to MockFunctionCallingLLM. Fixes include sanitizing Pydantic schema names for OpenAI and a crash when result.ids is None.
- December 4, 2025 · v0.14.10
Core adds a mock function-calling LLM. A new Airweave tool integration provides search features, and typo fixes land in the qianfan LLM and utils packages.
- December 2, 2025 · v0.14.9
MultiModalVectorStoreIndex now returns a multi-modal ContextChatEngine, and ingestion guarantees _node-content is readable. Fixes cover context copying in
run_asyncand memory queue ordering after flush.
November 2025
- November 10, 2025 · v0.14.8
Core fixes ReActOutputParser looping when the answer contains "Action:", handles multi-block ChatMessages in ReActAgent, and preserves exceptions in ToolOutput. Buffers were added to image, audio, video and document blocks.
October 2025
- October 30, 2025 · v0.14.7
Core adds a Serpex tool integration and ensures the full docstring is used in FunctionTool. Updates include the VoyageAI and Anthropic integrations and an outdated error message about setting an LLM.
- October 26, 2025 · v0.14.6
Core adds
allow_parallel_tool_callsfor non-streaming use and fixes rare cases where sentence splits exceeded chunk size. BedrockEmbedding now supports the Cohere v4 response format, and Isaacus embeddings are added. - October 15, 2025 · v0.14.5
Core removes a debug print, safely initializes RefDocInfo in Docstore, and fixes duplicate node positions when identical text repeats. Multiprocess loading gains a progress bar.
- October 3, 2025 · v0.14.4
Core fixes pre-release installs. Several embeddings packages get corrected OpenAI dependencies, and authorization header setup is fixed in the text embeddings inference package.
Ship updates like these to your users
LlamaIndex keeps a public record of what it ships. FoxPlug gives your product the same: a changelog page, an in-app widget and release posts, all from one place.
See FoxPlugSee a live changelog and widgetYour changelog, written for you
FoxPlug writes your changelog from your releases and commits in plain English and publishes it as a page and a feed your users can follow. The page above shows what that looks like for a product people really use.
Write my changelog See the changelog FoxPlug wrote from its own commits
Free to start. Pro is a month.
FoxPlug Pro
- A public changelog page and feed for each project
- Written from your releases and commits, in plain English your users read
- Connect GitHub and new releases become new entries