Release Notes

NOTE: All packages have been bumped to handle the latest llama-index-core version.

llama-index-core [0.13.0]

• breaking: removed deprecated agent classes, including FunctionCallingAgent, the older ReActAgent implementation, AgentRunner, all step workers, StructuredAgentPlanner, OpenAIAgent, and more. All users should migrate to the new workflow based agents: FunctionAgent, CodeActAgent, ReActAgent, and AgentWorkflow (#19529)
• breaking: removed deprecated QueryPipeline class and all associated code (#19554)
• breaking: changed default index.as_chat_engine() to return a CondensePlusContextChatEngine. Agent-based chat engines have been removed (which was the previous default). If you need an agent, use the above mentioned agent classes. (#19529)
• fix: Update BaseDocumentStore to not return Nones in result (#19513)
• fix: Fix FunctionTool param doc parsing and signature mutation; update tests (#19532)
• fix: Handle empty prompt in MockLLM.stream_complete (#19521)

llama-index-embeddings-mixedbreadai [0.5.0]

• feat: Update mixedbread embeddings and rerank for latest sdk (#19519)

llama-index-instrumentation [0.4.0]

• fix: let wrapped exceptions bubble up (#19566)

llama-index-llms-google-genai [0.3.0]

• feat: Add Thought Summaries and signatures for Gemini (#19505)

llama-index-llms-nvidia [0.4.0]

• feat: add support for kimi-k2-instruct (#19525)

llama-index-llms-upstage [0.6.0]

• feat: add new upstage model(solar-pro2) (#19526)

llama-index-postprocessor-mixedbreadai-rerank [0.5.0]

• feat: Update mixedbread embeddings and rerank for latest sdk (#19519)

llama-index-readers-github [0.8.0]

• feat: Github Reader enhancements for file filtering and custom processing (#19543)

llama-index-readers-s3 [0.5.0]

• feat: add support for region_name via client_kwargs in S3Reader (#19546)

llama-index-tools-valyu [0.4.0]

• feat: Update Valyu sdk to latest version (#19538)

llama-index-voice-agents-gemini-live [0.2.0]

• feat(beta): adding first implementation of gemini live (#19489)

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

• feat: astradb get nodes + delete nodes support (#19544)

llama-index-vector-stores-milvus [0.9.0]

• feat: Add support for specifying partition_names in Milvus search configuration (#19555)

llama-index-vector-stores-s3 [0.2.0]

• fix: reduce some metadata keys from S3VectorStore to save space (#19550)

llama-index-vector-stores-postgres [0.6.0]

• feat: Add support for ANY/ALL postgres operators (#19553)

Release Notes

[2025-07-22]

llama-index-core [0.12.52.post1]

• fix: do not write system prompt to memory in agents (#19512)

llama-index-core [0.12.52]

• fix: Fix missing prompt in async MultiModalLLMProgram calls (#19504)
• fix: Properly raise errors from docstore, fixes Vector Index Retrieval for stores_text=True/False (#19501)

llama-index-indices-managed-bge-m3 [0.5.0]

• feat: optimize memory usage for BGEM3Index persistence (#19496)

llama-index-readers-web [0.4.5]

• feat: Add timeout to webpage readers, defaults to 60 seconds (#19503)

llama-index-tools-jira-issue [0.1.0]

• feat: added jira issue tool spec (#19457)

llama-index-vector-stores-azureaisearch [0.3.10]

• chore: add **kwargs into AzureAISearchVectorStore super init (#19500)

llama-index-vector-stores-neo4jvector [0.4.1]

• fix: Patch Neo4jVector Call version (#19498)

Release Notes

[2025-07-21]

llama-index-core [0.12.51]

• feat: Enhance FunctionTool with auto type conversion for basic Python types like date when using pydantic fields in functions (#19479)
• fix: Fix retriever KeyError when using FAISS and other vector stores that do no store text (#19476)
• fix: add system prompt to memory and use it also for structured generation (#19490)

llama-index-readers-azstorage-blob [0.3.2]

• fix: Fix metadata serialization issue in AzStorageBlobReader (#19491)

Release Notes

[2025-07-19]

llama-index-core [0.12.50]

• feat: support html table extraction in MarkdownElementNodeParser (#19449)
• fix/slightly breaking: make get_cache_dir() function more secure by changing default location (#19415)
• fix: resolve race condition in SQLAlchemyChatStore with precise timestamps (#19432)
• fix: update document store import to use BaseDocumentStore in DocumentContextExtractor (#19466)
• fix: improve empty retrieval check in vector index retriever (#19471)
• fix: Fix running workflow agents as MCP servers by adding start event handling to workflow agents (#19470)
• fix: handle ID type mismatch in various retrievers (#19448)
• fix: add structured output to multi agent also from secondary constructor + tests (#19435)
• fix: duplicated session_id metadata_filter in VectorMemoryBlock (#19427)
• fix: make sure to stop agent on function tool return direct (#19413)
• fix: use a private folder to store NTLK cache (#19444)
• fix: Update ReAct agent parse error message (#19431)

llama-index-instrumentation [0.3.0]

• feat: Improve instrumentation span name (#19454)

llama-index-llms-bedrock-converse [0.7.6]

• chore: added llama 4 models in Bedrock Converse, remove llama 3.2 1b and 3b from function calling models (#19434)

llama-index-llms-cloudflare-ai-gateway [0.1.0]

• feat: introduce cloudflare ai gateway (#19395)

llama-index-llms-google-genai [0.2.5]

• feat: Add google_search Tool Support to GoogleGenAI LLM Integration (#19406)

llama-index-readers-confluence [0.3.2]

• refactor: various Confluence reader enhancements (logging, error handling) (#19424)

llama-index-readers-service-now [0.1.0]

• feat: added service-now reader (#19429)

llama-index-protocols-ag-ui [0.1.4]

• chore: remove some stray debug prints from AGUI (#19469)

llama-index-tools-wikipedia [0.3.1]

• fix: Remove load_kwargs from WikipediaToolSpec.load_data tool (#19464)

llama-index-vector-stores-baiduvectordb [0.3.1]

• fix: pass **kwargs to super().__init__ in BaiduVectorDB (#19436)

llama-index-vector-stores-moorcheh [0.1.1]

• fix: Update Moorcheh Vector Store namespace resolution (#19461)

llama-index-vector-stores-s3 [0.1.0]

• feat: s3 vectors support (#19456)

Release Notes

[2025-07-14]

llama-index-core [0.12.49]

• fix: skip tests on CI (#19416)
• fix: fix structured output (#19414)
• Fix: prevent duplicate triplets in SimpleGraphStore.upsert_triplet (#19404)
• Add retry capability to workflow agents (#19393)
• chore: modifying raptors dependencies with stricter rules to avoid test failures (#19394)
• feat: adding a first implementation of structured output in agents (#19337)
• Add tests for and fix issues with Vector Store node serdes (#19388)
• Refactor vector index retrieval (#19382)
• Retriever Query Engine should use async node postprocessors (#19380)

llama-index-llms-bedrock-converse [0.7.5]

• Fix BedrockConverse streaming token counting by handling messageStop … (#19369)

llama-index-llms-nvidia [0.3.5]

• nvidia-llm : Adding support to use llm models outside default list (#19366)

llama-index-llms-oci-genai [0.5.2]

• Fix bugs in tool calling for OCI generative AI Llama models (#19376)

llama-index-postprocessor-flashrank-rerank [0.1.0]

• Fix bugs in tool calling for OCI generative AI Llama models (#19376)

llama-index-readers-web [0.4.4]

• fix: avoid SimpleWebPageReader and others to use url as a Document id (#19398)

llama-index-storage-docstore-duckdb [0.1.0]

• Add DuckDB KV, Document, and Index Store (#19282)

llama-index-storage-index-store-duckdb [0.1.0]

• Add DuckDB KV, Document, and Index Store (#19282)

llama-index-storage-kvstore-duckdb [0.1.3]

• DuckDB: Deadlocks-b-gone (#19401)
• Improvements for DuckDB thread safety and embed dimension handling (#19391)
• Add DuckDB KV, Document, and Index Store (#19282)

llama-index-vector-stores-duckdb [0.4.6]

• DuckDB: Deadlocks-b-gone (#19401)
• Improvements for DuckDB thread safety and embed dimension handling (#19391)
• DuckDB Async and Faster Cosine Similarity (#19383)
• DuckDB Small clean-up and add embeddings to returned nodes (#19377)

llama-index-vector-stores-moorcheh [0.1.0]

• feat: Add Moorcheh vector store integration (#19349)

Release Notes

[2025-07-09]

llama-index-core [0.12.48]

• fix: convert dict chat_history to ChatMessage objects in AgentWorkflowStartEvent (#19371)
• fix: Replace ctx.get/set with ctx.store.get/set in Context (#19350)
• Bump the pip group across 6 directories with 1 update (#19357)
• Make fewer trips to KV store during Document Hash Checks (#19362)
• Don't store Copy of document in metadata and properly return Nodes (#19343)
• Bump llama-index-core from 0.12.8 to 0.12.41 in /docs in the pip group across 1 directory (#19345)
• fix: Ensure CallbackManager is applied to default embed_model (#19335)
• fix publish sub-package workflow (#19338)

llama-index-embeddings-huggingface-optimum-intel [0.3.1]

• Fix IntelEmbedding base.py (#19351)

llama-index-indices-managed-lancedb [0.1.0]

• Fix broken lancedb tests (#19352)

llama-index-indices-managed-llamacloud [0.7.10]

• vbump llama-cloud (#19355)
• Fix async retrieval of page figure nodes (#19334)

llama-index-llms-google-genai [0.2.4]

• Add Cached Content Support to GoogleGenAI LLM Integration (#19361)

llama-index-llms-oci-genai [0.5.1]

• Add support of Image prompt for OCI generative AI Llama models (#19306)

llama-index-readers-file [0.4.11]

• swamp xml for defusedxml (#19342)

llama-index-storage-chat-stores-postgres [0.2.2]

• Update asyncpg (#19365)

Release Notes

llama-index-core [0.12.47]

• feat: add default max_iterations arg to .run() of 20 for agents (#19035)
• feat: set tool_required to True for FunctionCallingProgram and structured LLMs where supported (#19326)
• fix: fix missing raw in agent workflow events (#19325)
• fix: fixed parsing of empty list in parsing json output (#19318)
• chore: Deprecate Multi Modal LLMs (#19115)
  • All existing multi-modal llms are now extensions of their base LLM counterpart
  • Base LLM classes support multi-modal features in llama-index-core
  • Base LLM classes use ImageBlock internally to support multi-modal features

llama-index-cli [0.4.4]

• fix: prevent command injection vulnerability in RAG CLI --clear flag (#19322)

llama-index-indices-managed-lancedb [0.1.0]

• feat: Adding an integration for LanceDB MultiModal AI LakeHouse (#19232)

llama-index-llms-anthropic [0.7.6]

• feat: anthropic citations support (#19316)

llama-index-llms-oci-genai [0.5.1]

• feat: Add support of Image prompt for OCI generative AI Llama models (#19306)

llama-index-readers-web [0.4.3]

• chore: Add firecrawl integration source (#19203)

Release Notes

[2025-07-02]

llama-index-core [0.12.46]

• feat: Add async delete and insert to vector store index (#19281)
• fix: Fixing ChatMessage to str handling of empty inputs (#19302)
• fix: fix function tool context detection with typed context (#19309)
• fix: inconsistent ref node handling (#19286)
• chore: simplify citation block schema (#19308)

llama-index-embeddings-google-genai [0.2.1]

• chore: bump min google-genai version (#19304)

llama-index-embeddings-nvidia [0.3.4]

• fix: embedding model with custom endpoints 404 error (#19295)

llama-index-llms-google-genai [0.2.3]

• chore: bump min google-genai version (#19304)

llama-index-tools-mcp [0.2.6]

• fix: configuring resources from the mcp server correctly (#19307)

llama-index-voice-agents-elevenlabs [0.3.0-beta]

• fix: Migrating Elevenlabs to adjust it to framework standard (#19273)

Release Notes

[2025-06-30]

llama-index-core [0.12.45]

• feat: allow tools to output content blocks (#19265)
• feat: Add chat UI events and models to core package (#19242)
• fix: Support loading Node from ingestion cache (#19279)
• fix: Fix SemanticDoubleMergingSplitterNodeParser not respecting max_chunk_size (#19235)
• fix: replace get_doc_id() with id_ in base index (#19266)
• chore: remove usage and references to deprecated Context get/set API (#19275)
• chore: deprecate older agent packages (#19249)

llama-index-llms-anthropic [0.7.5]

• feat: Adding new AWS Claude models available on Bedrock (#19233)

llama-index-embeddings-azure-openai [0.3.9]

• feat: Add dimensions parameter to AzureOpenAIEmbedding (#19239)

llama-index-embeddings-bedrock [0.5.2]

• feat: Update aioboto3 dependency (#19237)

llama-index-llms-bedrock-converse [0.7.4]

• feat: Update aioboto3 dependency (#19237)

llama-index-llms-dashscope [0.4.1]

• fix: Fix dashscope qwen assistant api Error response problem, extract tool_calls info from ChatMessage kwargs to top level (#19224)

llama-index-memory-mem0 [0.3.2]

• feat: Adapting Mem0 to new framework memory standard (#19234)

llama-index-tools-google [0.5.0]

• feat: Add proper async google search to tool spec (#19250)
• fix: Clean up results in GoogleSearchToolSpec (#19246)

llama-index-vector-stores-postgres [0.5.4]

• fix: Fix pg vector store sparse query (#19241)

Release Notes

llama-index-core [0.12.44]

• feat: Adding a CachePoint content block for caching chat messages (#19193)
• fix: fix react system header formatting in workflow agent (#19158)
• fix: fix ReActOutputParser when no "Thought:" prefix is produced by the LLM (#19190)
• fix: Fixed string striping in react output parser (#19192)
• fix: properly handle system prompt for CodeAct agent (#19191)
• fix: Exclude raw field in AgentStream event to fix potential serialization issue (#19150)
• chore: Mark older agent architectures in core as deprecated (#19205)
• chore: deprecate query pipelines in code (#19206)

llama-index-embeddings-fastembed [0.3.5]

• feat: Add Batch Support for FastEmbed (#19147)

llama-index-embeddings-huggingface [0.5.5]

• feat: Add async batching for huggingface using asyncio.to_thread (#19207)

llama-index-llms-anthropic [0.7.4]

• fix: update kwargs for anthropic bedrock (#19169)

llama-index-llms-google-genai [0.2.2]

• fix: Setting up System instruction properly for google genai client (#19196)

llama-index-llms-mistralai [0.6.1]

• fix: Fix image url handling in Mistral AI (#19139)

llama-index-llms-perplexity [0.3.7]

• fix: make api_key use PPLX_API_KEY in perplexity llm integration (#19217)

llama-index-postprocessor-bedrock-rerank [0.4.0]

• fix: Avoid changing 'top_n' self attribute at runtime (#19221)

llama-index-postprocessor-sbert-rerank [0.3.2]

• feat: add cross_encoder_kwargs parameter for advanced configuration (#19148)

llama-index-utils-workflow [0.3.5]

• feat: Adding visualization functions for single/multi agent workflows (#19101)

llama-index-vector-stores-azureaisearch [0.3.8]

• feat: Enable forwarding of arbitrary Azure Search SDK parameters in AzureAISearchVectorStore for document retrieval (#19173)

llama-index-vector-stores-db2 [0.1.0]

• feat: add IBM Db2 vector store (#19195)

llama-index-vector-stores-duckdb [0.4.0]

• feat: refactor DuckDB VectorStore (#19106)

llama-index-vector-stores-pinecone [0.6.0]

• feat: support pinecone v7 (#19163)
• fix: support python version >=3.9,<4.0 for llama-index-vector-stores-pinecone (#19186)

llama-index-vector-stores-qdrant [0.6.1]

• fix: fix types with IN/NIN filters in qdrant (#19159)

llama-index-voice-agents-openai [0.1.1-beta]

• feat: Adding beta OpenAI Realtime Conversation integration (#19010)

Release Notes

llama-index-core [0.12.43]

• feat: Make BaseWorkflowAgent a workflow itself (#19052)
• fix: make the progress bar of title extractor unified (#19131)
• fix: Use get_tqdm_iterable in SimpleDirectoryReader (#18722)
• chore: move out Workflows code to llama-index-workflows and keeping backward compatibility (#19043)
• chore: move instrumentation code out to its own package llama-index-instrumentation (#19062)

llama-index-llms-bedrock-converse [0.7.2]

• fix: improve empty tool call handling in bedrock converse (#19084)

llama-index-llms-openai [0.4.7]

• fix: fix building openai responses dicts (#19089, #19094)

llama-index-llms-perplexity [0.3.6]

• feat: Perf/improve robustness of perplexity llm integration (#19022)

llama-index-postprocessor-sbert-rerank [0.3.1]

• feat: Added cache dir to Sentence Transformers post processor (#19097)

llama-index-protocols-ag-ui [0.1.2]

• feat: add ag-ui protocol support (#19104, #19103, #19102, #18898)

llama-index-readers-google [0.6.2]

• fix: Fix error getting metadata file IDs in google drive reader (#19118)

llama-index-readers-hive [0.3.1]

• chore: deprecate and remove hive reader (#18990)

llama-index-readers-mongodb [0.3.2]

• feat: Added Async driver for alazy_load_data for mongodb reader (#19038)

llama-index-storage-chat-store-sqlite [0.1.1]

• fix: sqlite chat store compatibility with sqlalchemy 1.4 (#19067)

llama-index-tools-hive [0.1.0]

• feat: Add Hive Intelligence search tool (#19029)

llama-index-utils-workflow [0.3.4]

• feat: support drawing mermaid diagrams of workflows (#19083)

llama-index-vector-stores-lancedb [0.3.3]

• fix: create table with pre-defined schema (#19064)

llama-index-vector-stores-milvus [0.8.5]

• fix: Connections.connect() got multiple values for argument alias (#19119)

llama-index-vector-stores-opengauss [0.1.0]

• feat: add openGauss integration (#19024)

Release Notes

llama-index-core [0.12.42]

• fix: pass input message to memory get (#19054)
• fix: use async memory operations within async functions (#19032)
• fix: Using uuid instead of hashing for broader compatibility in SQLTableNodeMapping (#19011)

llama-index-embeddings-bedrock [0.5.1]

• feat: Update aioboto3 dependency (#19015)

llama-index-indices-managed-llama-cloud [0.7.7]

• feat: figure retrieval SDK integration (#19017)
• fix: Return empty list when argument raw_figure_nodes is None type in page_figure_nodes_to_node_with_score (#19053)

llama-index-llms-mistralai [0.6.0]

• feat: Add reasoning support to mistralai LLM + magistral (#19048)

llama-index-llms-openai [0.4.5]

• feat: O3 pro day 0 support (#19030)
• fix: skip tool description length check in openai response api (#18956)

llama-index-llms-perplexity [0.3.5]

• fix: perplexity llm integration bug fix (#19007)

llama-index-multi-modal-llms-openai-like [0.1.0]

• feat: add openai like multi-modal LLM (#18997)

llama-index-postprocessor-bedrock-rerank [0.3.3]

• feat: Prefer 'BedrockRerank' over 'AWSBedrockRerank' (#19016)

llama-index-readers-papers [0.3.1]

• fix: make filename hashing more robust (#18318)

llama-index-tools-artifact-editor [0.1.0]

• feat: Create ArtifactEditorToolSpec for editing pydantic objects (#18989)

llama-index-utils-workflow [0.3.3]

• feat: Add label truncation to workflow visualization (#19027)

llama-index-vector-stores-opensearch [0.5.6]

• feat: Add ability to exclude source fields from query response (#19018)

llama-index-voice-agents-elevenlabs [0.2.0-beta]

• fix: Docs corrections + integrating tools for ElevenLabs integration (#19014)

Release Notes

llama-index-core [0.12.41]

• feat: Add MutableMappingKVStore for easier caching (#18893)
• fix: async functions in tool specs (#19000)
• fix: properly apply file limit to SimpleDirectoryReader (#18983)
• fix: overwriting of LLM callback manager from Settings (#18951)
• fix: Adding warning in the docstring of JsonPickleSerializer for the user to deserialize only safe things, rename to PickleSerializer (#18943)
• fix: ImageDocument path and url checking to ensure that the input is really an image (#18947)
• chore: remove some unused utils from core (#18985)

llama-index-embeddings-azure-openai [0.3.8]

• fix: Azure api-key and azure-endpoint resolution fixes (#18975)
• fix: api_base vs azure_endpoint resolution fixes (#19002)

llama-index-graph-stores-ApertureDB [0.1.0]

• feat: Aperturedb propertygraph (#18749)

llama-index-indices-managed-llama-cloud [0.7.4]

• fix: resolve retriever llamacloud index (#18949)
• chore: composite retrieval add ReRankConfig (#18973)

llama-index-llms-azure-openai [0.3.4]

• fix: api_base vs azure_endpoint resolution fixes (#19002)

llama-index-llms-bedrock-converse [0.7.1]

• fix: handle empty message content to prevent ValidationError (#18914)

llama-index-llms-litellm [0.5.1]

• feat: Add DocumentBlock support to LiteLLM integration (#18955)

llama-index-llms-ollama [0.6.2]

• feat: Add support for the new think feature in ollama (#18993)

llama-index-llms-openai [0.4.4]

• feat: add OpenAI JSON Schema structured output support (#18897)
• fix: skip tool description length check in openai response api (#18956)

llama-index-packs-searchain [0.1.0]

• feat: Add searchain package (#18929)

llama-index-readers-docugami [0.3.1]

• fix: Avoid hash collision in XML parsing (#18986)

llama-index-readers-file [0.4.9]

• fix: pin llama-index-readers-file pandas for now (#18976)

llama-index-readers-gcs [0.4.1]

• feat: Allow newer versions of gcsfs (#18987)

llama-index-readers-obsidian [0.5.2]

• fix: Obsidian reader checks and skips hardlinks (#18950)

llama-index-readers-web [0.4.2]

• fix: Use httpx instead of urllib in llama-index-readers-web (#18945)

llama-index-storage-kvstore-postgres [0.3.5]

• fix: Remove unnecessary psycopg2 from llama-index-storage-kvstore-postgres dependencies (#18964)

llama-index-tools-mcp [0.2.5]

• fix: actually format the workflow args into a start event instance (#19001)
• feat: Adding support for log recording during MCP tool calls (#18927)

llama-index-vector-stores-chroma [0.4.2]

• fix: Update ChromaVectorStore port field and argument types (#18977)

llama-index-vector-stores-milvus [0.8.4]

• feat: Upsert Entities supported in Milvus (#18962)

llama-index-vector-stores-redis [0.5.2]

• fix: Correcting Redis URL/Client handling (#18982)

llama-index-voice-agents-elevenlabs [0.1.0-beta]

• feat: ElevenLabs beta integration (#18967)

Release Notes

llama-index-core [0.12.40]

• feat: Add StopEvent step validation so only one workflow step can handle StopEvent (#18932)
• fix: Add compatibility check before providing tool_required to LLM args (#18922)

llama-index-embeddings-cohere [0.5.1]

• fix: add batch size validation with 96 limit for Cohere API (#18915)

llama-index-llms-anthropic [0.7.2]

• feat: Support passing static AWS credentials to Anthropic Bedrock (#18935)
• fix: Handle untested no tools scenario for anthropic tool config (#18923)

llama-index-llms-google-genai [0.2.1]

• fix: use proper auto mode for google-genai function calling (#18933)

llama-index-llms-openai [0.4.2]

• fix: clear up some field typing issues of OpenAI LLM API (#18918)
• fix: migrate broken reasoning_effort kwarg to reasoning_options dict in OpenAIResponses class (#18920)

llama-index-tools-measurespace [0.1.0]

• feat: Add weather, climate, air quality and geocoding tool from Measure Space (#18909)

llama-index-tools-mcp [0.2.3]

• feat: Add headers handling to BasicMCPClient (#18919)

Release Notes

[2025-05-30]

llama-index-core [0.12.39]

• feat: Adding Resource to perform dependency injection in Workflows (docs coming soon!) (#18884)
• feat: Add tool_required param to function calling LLMs (#18654)
• fix: make prefix and response non-required for hitl events (#18896)
• fix: SelectionOutputParser when LLM chooses no choices (#18886)

llama-index-indices-managed-llama-cloud [0.7.2]

• feat: add non persisted composite retrieval (#18908)

llama-index-llms-bedrock-converse [0.7.0]

• feat: Update aioboto3 dependency to allow latest version (#18889)

llama-index-llms-ollama [0.6.1]

• Support ollama 0.5.0 SDK, update ollama docs (#18904)

llama-index-vector-stores-milvus [0.8.3]

• feat: Multi language analyzer supported in Milvus (#18901)

Release Notes

llama-index-core [0.12.38]

• feat: Adding a very simple implementation of an embeddings cache (#18864)
• feat: Add cols_retrievers in NLSQLRetriever (#18843)
• feat: Add row, col, and table retrievers as args in NLSQLTableQueryEngine (#18874)
• feat: add configurable allow_parallel_tool_calls to FunctionAgent (#18829)
• feat: Allow ctx in BaseToolSpec functions, other ctx + tool calling overhauls (#18783)
• feat: Optimize get_biggest_prompt for readability and efficiency (#18808)
• fix: prevent DoS attacks in JSONReader (#18877)
• fix: SelectionOutputParser when LLM chooses no choices (#18886)
• fix: resuming AgentWorkflow from ctx during hitl (#18844)
• fix: context serialization during AgentWorkflow runs (#18866)
• fix: Throw error if content block resolve methods yield empty bytes (#18819)
• fix: Reduce issues when parsing "Thought/Action/Action Input" ReActAgent completions (#18818)
• fix: Strip code block backticks from QueryFusionRetriever llm response (#18825)
• fix: Fix get_function_tool in function_program.py when schema doesn't have "title" key (#18796)

llama-index-agent-azure-foundry [0.1.0]

• feat: add azure foundry agent integration (#18772)

llama-index-agent-llm-compiler [0.3.1]

• feat: llm-compiler support stream_step/astream_step (#18809)

llama-index-embeddings-google-genai [0.2.0]

• feat: add gemini embeddings tests and retry configs (#18846)

llama-index-embeddings-openai-like [0.1.1]

• fix: Pass http_client & async_http_client to parent for OpenAILikeEmbedding (#18881)

llama-index-embeddings-voyageai [0.3.6]

• feat: Introducing voyage-3.5 models (#18793)

llama-index-indices-managed-llama-cloud [0.7.1]

• feat: add client support for search_filters_inference_schema (#18867)
• feat: add async methods and blank index creation (#18859)

llama-index-llms-anthropic [0.6.19]

• feat: update for claude 4 support in Anthropic LLM (#18817)
• fix: thinking + tool calls in anthropic (#18834)
• fix: check thinking is non-null in anthropic messages (#18838)
• fix: update/fix claude-4 support (#18820)

llama-index-llms-bedrock-converse [0.6.0]

• feat: add-claude4-model-support (#18827)
• fix: fixing DocumentBlock usage within Bedrock Converse (#18791)
• fix: calling tools with empty arguments (#18786)

llama-index-llms-cleanlab [0.5.0]

• feat: Update package name and models (#18483)

llama-index-llms-featherlessai [0.1.0]

• feat: featherless-llm-integration (#18778)

llama-index-llms-google-genai [0.1.14]

• fix: Google GenAI token counting behavior, add basic retry mechanism (#18876)

llama-index-llms-ollama [0.5.6]

• feat: Attempt to automatically set context window in ollama (#18822)
• feat: use default temp in ollama models (#18815)

llama-index-llms-openai [0.3.44]

• feat: Adding new OpenAI responses features (image gen, mcp call, code interpreter) (#18810)
• fix: Update OpenAI response type imports for latest openai library compatibility (#18824)
• fix: Skip tool description length check in OpenAI agent (#18790)

llama-index-llms-servam [0.1.1]

• feat: add Servam AI LLM integration with OpenAI-like interface (#18841)

llama-index-observability-otel [0.1.0]

• feat: OpenTelemetry integration for observability (#18744)

llama-index-packs-raptor [0.3.2]

• Use global llama_index tokenizer in Raptor clustering (#18802)

llama-index-postprocessor-rankllm-rerank [0.5.0]

• feat: use latest rank-llm sdk (#18831)

llama-index-readers-azstorage-blob [0.3.1]

• fix: Metadata and filename in azstorageblobreader (#18816)

llama-index-readers-file [0.4.8]

• fix: reading pptx files from remote fs (#18862)

llama-index-storage-kvstore-postgres [0.3.1]

• feat: Create PostgresKVStore from existing SQLAlchemy Engine (#18798)

llama-index-tools-brightdata [0.1.0]

• feat: brightdata integration (#18690)

llama-index-tools-google [0.3.1]

• fix: GmailToolSpec.load_data() calls search with missing args (#18832)

llama-index-tools-mcp [0.2.2]

• feat: enhance SSE endpoint detection for broader compatibility (#18868)
• feat: overhaul BasicMCPClient to support all MCP features (#18833)
• fix: McpToolSpec fetch all tools given the empty allowed_tools list (#18879)
• fix: add missing BasicMCPClient.with_oauth() kwargs (#18845)

llama-index-tools-valyu [0.2.0]

• feat: Update to valyu 2.0.0 (#18861)

llama-index-vector-stores-azurecosmosmongo [0.6.0]

• feat: Add Vector Index Compression support for Azure Cosmos DB Mongo vector store (#18850)

llama-index-vector-stores-opensearch [0.5.5]

• feat: add filter support to check if a metadata key doesn't exist (#18851)
• fix: dont pass in both extra_info and metadata in vector store nodes (#18805)

Release Notes

llama-index-core [0.12.37]

• Ensure Memory returns at least one message (#18763)
• Separate text blocks with newlines when accessing message.content (#18763)
• reset next_agent in multi agent workflows (#18782)
• support sqlalchemy v1 in chat store (#18780)
• fix: broken hotpotqa dataset URL (#18764)
• Use get_tqdm_iterable in SimpleDirectoryReader (#18722)
• Pass agent workflow kwargs into start event (#18747)
• fix(chunking): Ensure correct handling of multi-byte characters during AST node chunking (#18702)

llama-index-llms-anthropic [0.6.14]

• Fixed DocumentBlock handling in OpenAI and Anthropic (#18769)

llama-index-llms-bedrock-converse [0.5.4]

• Fix tool call parsing for bedrock converse (#18781)
• feat: add missing client params for bedrock (#18768)
• fix merging multiple tool calls in bedrock converse (#18761)

llama-index-llms-openai [0.3.42]

• Fixed DocumentBlock handling in OpenAI and Anthropic (#18769)
• Remove tool-length check in openai (#18784)
• Add check for empty tool call delta, bump version (#18745)

llama-index-llms-openai-like [0.3.5]

• Remove tool-length check in openai (#18784)

llama-index-retrievers-vectorize [0.1.0]

• Add Vectorize retriever (#18685)

llama-index-tools-desearch [0.1.0]

• Feature/desearch integration (#18738)

Release Notes

llama-index-core [0.12.35]

• add support for prefilling partial tool kwargs on FunctionTool (#18658)
• Fix/react agent max iterations skipping (#18634)
• handling for edge-case serialization in prebuilt workflows like AgentWorkflow (#18628)
• memory revamp with new base class (#18594)
• add prebuilt memory blocks (#18607)

llama-index-embeddings-autoembeddings [0.1.0]

• Support for AutoEmbeddings integration from chonkie (#18578)

llama-index-embeddings-huggingface-api [0.3.1]

• Fix dep versions for huggingface-hub (#18662)

llama-index-indices-managed-vectara [0.4.5]

• Bugfix in using cutoff argument with chain reranker in Vectara (#18610)

llama-index-llms-anthropic [0.6.12]

• anthropic citations and tool calls (#18657)

llama-index-llms-cortex [0.3.0]

• Cortex enhancements 2 for auth (#18588)

llama-index-llms-dashscope [0.3.3]

• Fix dashscope tool call parsing (#18608)

llama-index-llms-google-genai [0.1.12]

• Fix modifying object references in google-genai llm (#18616)
• feat(llama-index-llms-google-genai): 2.5-flash-preview tests (#18575)
• Fix last_msg indexing (#18611)

llama-index-llms-huggingface-api [0.4.3]

• Huggingface API fixes for task and deps (#18662)

llama-index-llms-litellm [0.4.2]

• fix parsing streaming tool calls (#18653)

llama-index-llms-meta [0.1.1]

• Support Meta Llama-api as an LLM provider (#18585)

llama-index-node-parser-docling [0.3.2]

• Fix/docling node parser metadata (#186390)

llama-index-node-parser-slide [0.1.0]

• add SlideNodeParser integration (#18620)

llama-index-readers-github [0.6.1]

• Fix: Add follow_redirects=True to GitHubIssuesClient (#18630)

llama-index-readers-markitdown [0.1.1]

• Fix MarkItDown Reader bugs (#18613)

llama-index-readers-oxylabs [0.1.2]

• Add Oxylabs readers (#18555)

llama-index-readers-web [0.4.1]

• Fixes improper invocation of Firecrawl library (#18646)
• Add Oxylabs readers (#18555)

llama-index-storage-chat-store-gel [0.1.0]

• Add Gel integrations (#18503)

llama-index-storage-docstore-gel [0.1.0]

• Add Gel integrations (#18503)

llama-index-storage-kvstore-gel [0.1.0]

• Add Gel integrations (#18503)

llama-index-storage-index-store-gel [0.1.0]

• Add Gel integrations (#18503)

llama-index-utils-workflow [0.3.2]

• Fix event colors of draw_all_possible_flows (#18660)

llama-index-vector-stores-faiss [0.4.0]

• Add Faiss Map Vector store and fix missing index_struct delete (#18638)

llama-index-vector-stores-gel [0.1.0]

• Add Gel integrations (#18503)

llama-index-vector-stores-postgres [0.5.2]

• add indexed metadata fields (#18595)

NOTE: Updating to v0.12.0 will require bumping every other llama-index-* package! Every package has had a version bump. Only notable changes are below.

llama-index-core [0.12.0]

• Dropped python3.8 support, Unpinned numpy (#16973)
• Kg/dynamic pg triplet retrieval limit (#16928)

llama-index-indices-managed-llama-cloud [0.6.1]

• Add ID support for LlamaCloudIndex & update from_documents logic, modernize apis (#16927)
• allow skipping waiting for ingestion when uploading file (#16934)
• add support for files endpoints (#16933)

llama-index-indices-managed-vectara [0.3.0]

• Add Custom Prompt Parameter (#16976)

llama-index-llms-bedrock [0.3.0]

• minor fix for messages/completion to prompt (#15729)

llama-index-llms-bedrock-converse [0.4.0]

• Fix async streaming with bedrock converse (#16942)

llama-index-multi-modal-llms-nvidia [0.2.0]

• add vlm support (#16751)

llama-index-readers-confluence [0.3.0]

• Permit passing params to Confluence client (#16961)

llama-index-readers-github [0.5.0]

• Add base URL extraction method to GithubRepositoryReader (#16926)

llama-index-vector-stores-weaviate [1.2.0]

• Allow passing in Weaviate vector store kwargs (#16954)