LlamaIndex v0.14.24

Release Notes

[2026-08-19]

llama-index-core [0.14.24]

• fix: close HotpotQA evaluator file handles (#22113)
• fix(core): preserve oversized leaf nodes in CodeSplitter (#22119)
• fix: preserve multiblock chat history writes (#22124)
• fix(core): skip empty nodes in HTMLNodeParser (#22125)
• fix(core): respect mmr_threshold=0 in MMR embedding search (#22126)
• fix: keep all nodes per document in IngestionPipeline upserts (re-submit of #22046, not a dup of #21456) (#22133)
• fix: don't mark *args/**kwargs as required tool parameters (#22135)
• fix(core): default MetadataFilters condition to AND when None (#22155)
• fix: honor agent structured_output_fn/output_cls inside AgentWorkflow (#22162)
• fix(core): use word tokenization for stopword removal in SemanticDoubleMergingSplitterNodeParser (#22167)
• fix(core): restore prompt helper fallback to supplied LLM metadata (#22177)
• populate response text on streaming chat response when writing to memory (#22179)
• feat(property_graph): add raise_on_error to LLM path extractors (#22195)
• Clarify FactExtractionMemoryBlock condense prompt to return a full snapshot (#22213)
• fix a few typos (#22512)
• fix(core): give each CitationQueryEngine citation node its own id and offsets (#22537)
• fix(core): persist SimpleChatStore without escaping non-ascii (#22538)
• feat(core): allow Memory to accept any AsyncDBChatStore (#22541)
• test(node_parser): enable pytest discovery for sentence_window tests (#22545)
• Implement async for LLMRerank (#22597)

llama-index-embeddings-bedrock [0.8.3]

• fix: make aioboto3 optional in llama-index-embeddings-bedrock (#21915)
• fix(embeddings): validate non-empty Bedrock embedding payload before AWS call (#22527)

llama-index-embeddings-voyageai [0.6.1]

• fix: set chunk_size=32000 and fix input format for contextualizedembeddings (#22144)

llama-index-graph-stores-falkordb [0.6.0]

• fix(falkordb): correct schema, vector ranking, filters and batching in property graph store (#22522)

llama-index-indices-managed-dashscope [0.5.1]

• fix(dashscope): send query_history in retriever payload (#22212)

llama-index-indices-managed-lancedb [0.3.2]

• fix(lancedb): repair image ingestion crash, embedding_model typo, and… (#22544)
• lint (#22572)

llama-index-llms-anthropic [0.11.10]

• feat(anthropic): Add support for Claude Sonnet 5 (#22202)
• Fix/sonnet 5 function calling (#22217)
• feat(anthropic, bedrock-converse): add Claude Opus 5 to model allowlists (#22451)
• fix(anthropic): correct Opus 4.6 context window to 1M (#22555)

llama-index-llms-bedrock-converse [0.14.18]

• fix: make aioboto3 optional in llama-index-llms-bedrock-converse (#21916)
• feat(bedrock-converse): Add support for Claude Sonnet 5 (#22201)
• feat(anthropic, bedrock-converse): add Claude Opus 5 to model allowlists (#22451)
• feat(bedrock-converse): Support thinking type 'disabled' (#22534)

llama-index-llms-google-genai [0.10.0]

• fix: dont overwrite thought signatures (#21562)
• fix(google-genai): don't send default params in gemini genconfig (#22270)
• feat(google-genai): use gemini 3.7 Flash by default in lib and docs (#22734)

llama-index-llms-llama-cpp [0.6.1]

• fix(llama-cpp): report the model's effective context window (#22539)

llama-index-llms-openai [0.7.10]

• Add GPT-5.6 models to supported OpenAI models (#22385)

llama-index-node-parser-slide [0.3.0]

• fix a few typos (#22512)

llama-index-observability-otel [0.6.4]

• chore: bump instrumentation version in otel integration to 0.5.0 as minimum (#22151)
• fix: improve otel span ID handling (#22485)

llama-index-protocols-ag-ui [0.4.0]

• fix(ag-ui): raise ValueError instead of fabricating tool_call_id (#22103)
• fix: persist AG-UI frontend tool messages (#22109)
• Isolate AG-UI initial state copies (#22189)
• feat(ag-ui): support multimodal user input (images, audio, video, documents) (#22678)

llama-index-readers-alibabacloud-aisearch [0.4.1]

• fix: close file handles and http connections in reader integrations (#22529)

llama-index-readers-couchbase [0.5.1]

• Add telemetry (#22473)

llama-index-readers-file [0.6.0]

• fix: bound HWP section decompression (#22106)

llama-index-readers-gcs [0.6.1]

• docs: fix GCS reader README import paths (#22139)

llama-index-readers-github [0.11.2]

• fix(github-reader): add asyncio_mode=auto to unblock async tests in CI (#21937)

llama-index-readers-google [0.7.3]

• fix: GoogleDriveReader returns None instead of a list when an error is swallowed (#22556)

llama-index-readers-huggingface-fs [0.5.1]

• fix: close file handles and http connections in reader integrations (#22529)

llama-index-readers-hwp [0.5.0]

• fix: bound HWP section decompression (#22106)

llama-index-readers-memos [0.5.1]

• Fix MemosReader default endpoint and metadata (#22199)

llama-index-readers-minio [0.5.1]

• fix: don't hardcode verify=False for S3 TLS in BotoMinioReader (#22530)

llama-index-readers-sec-filings [0.5.1]

• fix: close file handles and http connections in reader integrations (#22529)

llama-index-readers-stripe-docs [0.5.1]

• fix: close file handles and http connections in reader integrations (#22529)

llama-index-retrievers-galaxia [0.2.2]

• fix: close file handles and http connections in reader integrations (#22529)

llama-index-retrievers-pathway [0.5.0]

• fix a few typos (#22512)

llama-index-storage-docstore-couchbase [0.4.1]

• Add telemetry (#22473)

llama-index-storage-index-store-couchbase [0.4.1]

• Add telemetry (#22473)

llama-index-storage-kvstore-couchbase [0.4.1]

• Add telemetry (#22473)

llama-index-tools-desearch [0.1.1]

• docs: remove leftover generation text from desearch tool README (#22149)

llama-index-tools-mcp [0.5.0]

• feat: migrate llama-index-tools-mcp to mcp 2.x (#22557)

llama-index-tools-typecast [0.1.0]

• Document Typecast unsynthesizable text errors (#22514)

llama-index-tools-vectara-query [0.5.0]

• fix a few typos (#22512)

llama-index-vector-stores-azureaisearch [0.5.1]

• fix(azureaisearch): preserve falsy metadata values when indexing nodes (#22154)

llama-index-vector-stores-couchbase [0.7.1]

• Add telemetry (#22473)

llama-index-vector-stores-pinecone [0.8.1]

• fix(pinecone): support pinecone client v8 and v9 (#22511)

llama-index-vector-stores-qdrant [0.10.3]

• fix(qdrant): aquery() drops falsy shard_identifier values (e.g. 0) (#22169)
• fix(qdrant): restore compatibility with qdrant-client 1.19.0 (#22630)

llama-index-vector-stores-vertexaivectorsearch [0.5.0]

• feat(vector-stores/vertexai): Expand support of V2 API for VertexAIVectorStore (#22085)

llama-index-vector-stores-weaviate [1.6.2]

• fix(weaviate): return real properties as node metadata for pre-existing collections (#22540)