chore(deps): update dependency llama-index to v0.10.13 [security]

[plural-renovate]

This PR contains the following updates:

Package	Update	Change
llama-index (source)	minor	==0.7.4 -> ==0.10.13

GitHub Vulnerability Alerts

CVE-2023-39662

An issue in llama_index v.0.7.13 and before allows a remote attacker to execute arbitrary code via the exec parameter in PandasQueryEngine function.

CVE-2024-4181

A command injection vulnerability exists in the RunGptLLM class of the llama_index library, version 0.9.47, used by the RunGpt framework from JinaAI to connect to Language Learning Models (LLMs). The vulnerability arises from the improper use of the eval function, allowing a malicious or compromised LLM hosting provider to execute arbitrary commands on the client's machine. This issue was fixed in version 0.10.13. The exploitation of this vulnerability could lead to a hosting provider gaining full control over client machines.

---

Release Notes

run-llama/llama_index (llama-index)

v0.10.13

Compare Source

New Features

• Added a llama-pack for KodaRetriever, for on-the-fly alpha tuning (#​11311)
• Added support for mistral-large (#​11398)
• Last token pooling mode for huggingface embeddings models like SFR-Embedding-Mistral (#​11373)
• Added fsspec support to SimpleDirectoryReader (#​11303)

Bug Fixes / Nits

• Fixed an issue with context window + prompt helper (#​11379)
• Moved OpenSearch vector store to BasePydanticVectorStore (#​11400)
• Fixed function calling in fireworks LLM (#​11363)
• Made cohere embedding types more automatic (#​11288)
• Improve function calling in react agent (#​11280)
• Fixed MockLLM imports (#​11376)

v0.10.12

Compare Source

New Features

• Added llama-index-postprocessor-colbert-rerank package (#​11057)
• MyMagicAI LLM (#​11263)
• MariaTalk LLM (#​10925)
• Add retries to github reader (#​10980)
• Added FireworksAI embedding and LLM modules (#​10959)

Bug Fixes / Nits

• Fixed string formatting in weaviate (#​11294)
• Fixed off-by-one error in semantic splitter (#​11295)
• Fixed download_llama_pack for multiple files (#​11272)
• Removed BUILD files from packages (#​11267)
• Loosened python version reqs for all packages (#​11267)
• Fixed args issue with chromadb (#​11104)

v0.10.11

Compare Source

Bug Fixes / Nits

• Fixed multi-modal LLM for async acomplete (#​11064)
• Fixed issue with llamaindex-cli imports (#​11068)

v0.10.10

Compare Source

I'm still a bit wonky with our publishing process -- apologies. This is just a version
bump to ensure the changes that were supposed to happen in 0.10.9 actually
did get published. (AF)

v0.10.9

Compare Source

• add llama-index-cli dependency

v0.10.8

Compare Source

v0.10.7

Compare Source

New Features

• Added Self-Discover llamapack (#​10951)

Bug Fixes / Nits

• Fixed linting in CICD (#​10945)
• Fixed using remote graph stores (#​10971)
• Added missing LLM kwarg in NoText response synthesizer (#​10971)
• Fixed openai import in rankgpt (#​10971)
• Fixed resolving model name to string in openai embeddings (#​10971)
• Off by one error in sentence window node parser (#​10971)

v0.10.6

Compare Source

First, apologies for missing the changelog the last few versions. Trying to figure out the best process with 400+ packages.

At some point, each package will have a dedicated changelog.

But for now, onto the "master" changelog.

New Features

• Added NomicHFEmbedding (#​10762)
• Added MinioReader (#​10744)

Bug Fixes / Nits

• Various fixes for clickhouse vector store (#​10799)
• Fix index name in neo4j vector store (#​10749)
• Fixes to sagemaker embeddings (#​10778)
• Fixed performance issues when splitting nodes (#​10766)
• Fix non-float values in reranker + b25 (#​10930)
• OpenAI-agent should be a dep of openai program (#​10930)
• Add missing shortcut imports for query pipeline components (#​10930)
• Fix NLTK and tiktoken not being bundled properly with core (#​10930)
• Add back llama_index.core.__version__ (#​10930)

v0.10.5

• added dead simple FnAgentWorker for custom agents (#​14329)
• Pass the kwargs on when build_index_from_nodes (#​14341)
• make async utils a bit more robust to nested async (#​14356)

v0.10.4

• Added PropertyGraphIndex and other supporting abstractions. See the full guide for more details (#​13747)
• Updated AutoPrevNextNodePostprocessor to allow passing in response mode and LLM (#​13771)
• fix type handling with return direct (#​13776)
• Correct the method name to _aget_retrieved_ids_and_texts in retrievval evaluator (#​13765)
• fix: QueryTransformComponent incorrect call self._query_transform (#​13756)
• implement more filters for SimpleVectorStoreIndex (#​13365)

v0.10.3

Compare Source

Bug Fixes / Nits

• Fixed passing in LLM to as_chat_engine (#​10605)
• Fixed system prompt formatting for anthropic (#​10603)
• Fixed elasticsearch vector store error on __version__ (#​10656)
• Fixed import on openai pydantic program (#​10657)
• Added client back to opensearch vector store exports (#​10660)
• Fixed bug in SimpleDirectoryReader not using file loaders properly (#​10655)
• Added lazy LLM initialization to RankGPT (#​10648)
• Fixed bedrock embedding from_credentials passing ing the model name (#​10640)
• Added back recent changes to TelegramReader (#​10625)

v0.10.1

Compare Source

New Features

• Anthropic support for new models (#​11623, #​11612)
• Easier creation of chat prompts (#​11583)
• Added a raptor retriever llama-pack (#​11527)
• Improve batch cohere embeddings through bedrock (#​11572)
• Added support for vertex AI embeddings (#​11561)

Bug Fixes / Nits

• Ensure order in async embeddings generation (#​11562)
• Fixed empty metadata for csv reader (#​11563)
• Serializable fix for composable retrievers (#​11617)
• Fixed milvus metadata filter support (#​11566)
• FIxed pydantic import in clickhouse vector store (#​11631)
• Fixed system prompts for gemini/vertext-gemini (#​11511)

v0.10.0

Compare Source

Breaking Changes

• Several changes are introduced. See the full blog post for complete details.

v0.9.48

Compare Source

Bug Fixes / Nits

• Add back deprecated API for BedrockEmbdding (#​10581)

v0.9.47

Compare Source

Last patch before v0.10!

New Features

• add conditional links to query pipeline (#​10520)
• refactor conditional links + add to cookbook (#​10544)
• agent + query pipeline cleanups (#​10563)

Bug Fixes / Nits

• Add sleep to fix lag in chat stream (#​10339)
• OllamaMultiModal kwargs (#​10541)
• Update Ingestion Pipeline to handle empty documents (#​10543)
• Fixing minor spelling error (#​10516)
• fix elasticsearch async check (#​10549)
• Docs/update slack demo colab (#​10534)
• Adding the possibility to use the IN operator for PGVectorStore (#​10547)
• fix agent reset (#​10562)
• Fix MD duplicated Node id from multiple docs (#​10564)

v0.9.46

Compare Source

New Features

• Update pooling strategy for embedding models (#​10536)
• Add Multimodal Video RAG example (#​10530)
• Add SECURITY.md (#​10531)
• Move agent module guide up one-level (#​10519)

Bug Fixes / Nits

• Deeplake fixes (#​10529)
• Add Cohere section for llamaindex (#​10523)
• Fix md element (#​10510)

v0.9.45.post1

Compare Source

New Features

• Upgraded deeplake vector database to use BasePydanticVectorStore (#​10504)

Bug Fixes / Nits

• Fix MD parser for inconsistency tables (#​10488)
• Fix ImportError for pypdf in MetadataExtractionSEC.ipynb (#​10491)

v0.9.45

Compare Source

New Features

• Upgraded deeplake vector database to use BasePydanticVectorStore (#​10504)

Bug Fixes / Nits

• Fix MD parser for inconsistency tables (#​10488)
• Fix ImportError for pypdf in MetadataExtractionSEC.ipynb (#​10491)

v0.9.44

Compare Source

New Features

• ollama vision cookbook (#​10438)
• Support Gemini "transport" configuration (#​10457)
• Add Upstash Vector (#​10451)

v0.9.43

Compare Source

New Features

• Add multi-modal ollama (#​10434)

Bug Fixes / Nits

• update base class for astradb (#​10435)

v0.9.42.post2

Compare Source

v0.9.42.post1

Compare Source

New Features

• Add Async support for Base nodes parser (#​10418)

v0.9.42

Compare Source

New Features

• Add Async support for Base nodes parser (#​10418)

v0.9.41

Compare Source

New Features

• Nomic Embedding (#​10388)
• Dashvector support sparse vector (#​10386)
• Table QA with MarkDownParser and Benchmarking (#​10382)
• Simple web page reader (#​10395)

Bug Fixes / Nits

• fix full node content in KeywordExtractor (#​10398)

v0.9.40

Compare Source

New Features

• Improve and fix bugs for MarkdownElementNodeParser (#​10340)
• Fixed and improve Perplexity support for new models (#​10319)
• Ensure system_prompt is passed to Perplexity LLM (#​10326)
• Extended BaseRetrievalEvaluator to include an optional PostProcessor (#​10321)

v0.9.39

Compare Source

New Features

• Support for new GPT Turbo Models (#​10291)
• Support Multiple docs for Sentence Transformer Fine tuning(#​10297)

Bug Fixes / Nits

• Marvin imports fixed (#​9864)

v0.9.38

Compare Source

New Features

• Support for new OpenAI v3 embedding models (#​10279)

Bug Fixes / Nits

• Extra checks on sparse embeddings for qdrant (#​10275)

v0.9.37.post1

Compare Source

v0.9.37

Compare Source

New Features

• Added a RAG CLI utility (#​10193)
• Added a textai vector store (#​10240)
• Added a Postgresql based docstore and index store (#​10233)
• specify tool spec in tool specs (#​10263)

Bug Fixes / Nits

• Fixed serialization error in ollama chat (#​10230)
• Added missing fields to SentenceTransformerRerank (#​10225)
• Fixed title extraction (#​10209, #​10226)
• nit: make chainable output parser more exposed in library/docs (#​10262)
• 🐛 summary index not carrying over excluded metadata keys (#​10259)

v0.9.36

Compare Source

New Features

• Added support for SageMakerEmbedding (#​10207)

Bug Fixes / Nits

• Fix duplicated file_id on openai assistant (#​10223)
• Fix circular dependencies for programs (#​10222)
• Run TitleExtractor on groups of nodes from the same parent document (#​10209)
• Improve vectara auto-retrieval (#​10195)

v0.9.35

Compare Source

New Features

• beautifulsoup4 dependency to new optional extra html (#​10156)
• make BaseNode.hash an @property (#​10163)
• Neutrino (#​10150)
• feat: JSONalyze Query Engine (#​10067)
• [wip] add custom hybrid retriever notebook (#​10164)
• add from_collection method to ChromaVectorStore class (#​10167)
• CLI experiment v0: ask (#​10168)
• make react agent prompts more editable (#​10154)
• Add agent query pipeline (#​10180)

Bug Fixes / Nits

• Update supabase vecs metadata filter function to support multiple fields (#​10133)
• Bugfix/code improvement for LanceDB integration (#​10144)
• beautifulsoup4 optional dependency (#​10156)
• Fix qdrant aquery hybrid search (#​10159)
• make hash a @​property (#​10163)
• fix: bug on poetry install of llama-index[postgres] (#​10171)
• [doc] update jaguar vector store documentation (#​10179)
• Remove use of not-launched finish_message (#​10188)
• Updates to Lantern vector stores docs (#​10192)
• fix typo in multi_document_agents.ipynb (#​10196)

v0.9.34

Compare Source

New Features

• Added SageMakerEndpointLLM (#​10140)
• Added support for Qdrant filters (#​10136)

Bug Fixes / Nits

• Update bedrock utils for Claude 2:1 (#​10139)
• BugFix: deadlocks using multiprocessing (#​10125)

v0.9.33

Compare Source

New Features

• Added RankGPT as a postprocessor (#​10054)
• Ensure backwards compatibility with new Pinecone client version bifucation (#​9995)
• Recursive retriever all the things (#​10019)

Bug Fixes / Nits

• BugFix: When using markdown element parser on a table containing comma (#​9926)
• extend auto-retrieval notebook (#​10065)
• Updated the Attribute name in llm_generators (#​10070)
• jaguar vector store add text_tag to add_kwargs in add() (#​10057)

v0.9.32

Compare Source

New Features

• added query-time row retrieval + fix nits with query pipeline over structured data (#​10061)
• ReActive Agents w/ Context + updated stale link (#​10058)

v0.9.31

Compare Source

New Features

• Added selectors and routers to query pipeline (#​9979)
• Added sparse-only search to qdrant vector store (#​10041)
• Added Tonic evaluators (#​10000)
• Adding async support to firestore docstore (#​9983)
• Implement mongodb docstore put_all method (#​10014)

Bug Fixes / Nits

• Properly truncate sql results based on max_string_length (#​10015)
• Fixed node.resolve_image() for base64 strings (#​10026)
• Fixed cohere system prompt role (#​10020)
• Remove redundant token counting operation in SentenceSplitter (#​10053)

v0.9.30

Compare Source

New Features

• Implements a Node Parser using embeddings for Semantic Splitting (#​9988)
• Add Anyscale Embedding model support (#​9470)

Bug Fixes / Nits

• nit: fix pandas get prompt (#​10001)
• Fix: Token counting bug (#​9912)
• Bump jinja2 from 3.1.2 to 3.1.3 (#​9997)
• Fix corner case for qdrant hybrid search (#​9993)
• Bugfix: sphinx generation errors (#​9944)
• Fix: language used before assignment in CodeSplitter (#​9987)
• fix inconsistent name "text_parser" in section "Use a Text Splitter… (#​9980)
• 🐛 fixing batch size (#​9982)
• add auto-async execution to query pipelines (#​9967)
• 🐛 fixing init (#​9977)
• Parallel Loading with SimpleDirectoryReader (#​9965)
• do not force delete an index in milvus (#​9974)

v0.9.29

Compare Source

New Features

• Added support for together.ai models (#​9962)
• Added support for batch redis/firestore kvstores, async firestore kvstore (#​9827)
• Parallelize IngestionPipeline.run() (#​9920)
• Added new query pipeline components: function, argpack, kwargpack (#​9952)

Bug Fixes / Nits

• Updated optional langchain imports to avoid warnings (#​9964)
• Raise an error if empty nodes are embedded (#​9953)

v0.9.28.post2

Compare Source

v0.9.28.post1

Compare Source

v0.9.28

Compare Source

New Features

• Added support for Nvidia TenorRT LLM (#​9842)
• Allow tool_choice to be set during agent construction (#​9924)
• Added streaming support for QueryPipeline (#​9919)

Bug Fixes / Nits

• Set consistent doc-ids for llama-index readers (#​9923, #​9916)
• Remove unneeded model inputs for HuggingFaceEmbedding (#​9922)
• Propagate tool_choice flag to downstream APIs (#​9901)
• Add chat_store_key to chat memory from_defaults() (#​9928)

v0.9.27

Compare Source

New Features

• add query pipeline (#​9908)
• Feature: Azure Multi Modal (fixes: #​9471) (#​9843)
• add postgres docker (#​9906)
• Vectara auto_retriever (#​9865)
• Redis Chat Store support (#​9880)
• move more classes to core (#​9871)

Bug Fixes / Nits / Smaller Features

• Propagate tool_choice flag to downstream APIs (#​9901)
• filter out negative indexes from faiss query (#​9907)
• added NE filter for qdrant payloads (#​9897)
• Fix incorrect id assignment in MyScale query result (#​9900)
• Qdrant Text Match Filter (#​9895)
• Fusion top k for hybrid search (#​9894)
• Fix (#​9867) sync_to_async to avoid blocking during asynchronous calls (#​9869)
• A single node passed into compute_scores returns as a float (#​9866)
• Remove extra linting steps (#​9878)
• add vectara links (#​9886)

v0.9.26

Compare Source

New Features

• Added a BaseChatStore and SimpleChatStore abstraction for dedicated chat memory storage (#​9863)
• Enable custom tree_sitter parser to be passed into CodeSplitter (#​9845)
• Created a BaseAutoRetriever base class, to allow other retrievers to extend to auto modes (#​9846)
• Added support for Nvidia Triton LLM (#​9488)
• Added DeepEval one-click observability (#​9801)

Bug Fixes / Nits

• Updated the guidance integration to work with the latest version (#​9830)
• Made text storage optional for doctores/ingestion pipeline (#​9847)
• Added missing sphinx-automodapi dependency for docs (#​9852)
• Return actual node ids in weaviate query results (#​9854)
• Added prompt formatting to LangChainLLM (#​9844)

v0.9.25.post1

Compare Source

v0.9.25

Compare Source

New Features

• Added concurrancy limits for dataset generation (#​9779)
• New deepeval one-click observability handler (#​9801)
• Added jaguar vector store (#​9754)
• Add beta multimodal ReAct agent (#​9807)

Bug Fixes / Nits

• Changed default batch size for OpenAI embeddings to 100 (#​9805)
• Use batch size properly for qdrant upserts (#​9814)
• _verify_source_safety uses AST, not regexes, for proper safety checks (#​9789)
• use provided LLM in element node parsers (#​9776)
• updated legacy vectordb loading function to be more robust (#​9773)
• Use provided http client in AzureOpenAI (#​9772)

v0.9.24

Compare Source

New Features

• Add reranker for BEIR evaluation (#​9743)
• Add Pathway integration. (#​9719)
• custom agents implementation + notebook (#​9746)

Bug Fixes / Nits

• fix beam search for vllm: add missing parameter (#​9741)
• Fix alpha for hrbrid search (#​9742)
• fix token counter (#​9744)
• BM25 tokenizer lowercase (#​9745)

v0.9.23

Compare Source

Bug Fixes / Nits

• docs: fixes qdrant_hybrid.ipynb typos (#​9729)
• make llm completion program more general (#​9731)
• Refactor MM Vector store and Index for empty collection (#​9717)
• Adding IF statement to check for Schema using "Select" (#​9712)
• allow skipping module loading in download_module and download_llama_pack (#​9734)

v0.9.22

Compare Source

New Features

• Added .iter_data() method to SimpleDirectoryReader (#​9658)
• Added async support to Ollama LLM (#​9689)
• Expanding pinecone filter support for in and not in (#​9683)

Bug Fixes / Nits

• Improve BM25Retriever performance (#​9675)
• Improved qdrant hybrid search error handling (#​9707)
• Fixed None handling in ChromaVectorStore (#​9697)
• Fixed postgres schema creation if not existing (#​9712)

v0.9.21

Compare Source

New Features

• Added zilliz cloud as a managed index (#​9605)

Bug Fixes / Nits

• Bedrock client and LLM fixes (#​9671, #​9646)

v0.9.20

Compare Source

New Features

• Added insert_batch_size to limit number of embeddings held in memory when creating an index, defaults to 2048 (#​9630)
• Improve auto-retrieval (#​9647)
• Configurable Node ID Generating Function (#​9574)
• Introduced action input parser (#​9575)
• qdrant sparse vector support (#​9644)
• Introduced upserts and delete in ingestion pipeline (#​9643)
• Add Zilliz Cloud Pipeline as a Managed Index (#​9605)
• Add support for Google Gemini models via VertexAI (#​9624)
• support allowing additional metadata filters on autoretriever (#​9662)

Bug Fixes / Nits

• Fix pip install commands in LM Format Enforcer notebooks (#​9648)
• Fixing some more links and documentations (#​9633)
• some bedrock nits and fixes (#​9646)

v0.9.19

Compare Source

New Features

• new llama datasets LabelledEvaluatorDataset & LabelledPairwiseEvaluatorDataset (#​9531)

v0.9.18

Compare Source

New Features

• multi-doc auto-retrieval guide (#​9631)

Bug Fixes / Nits

• fix(vllm): make Vllm's 'complete' method behave the same as other LLM class (#​9634)
• FIx Doc links and other documentation issue (#​9632)

v0.9.17

Compare Source

New Features

• [example] adding user feedback (#​9601)
• FEATURE: Cohere ReRank Relevancy Metric for Retrieval Eval (#​9495)

Bug Fixes / Nits

• Fix Gemini Chat Mode (#​9599)
• Fixed types-protobuf from being a primary dependency (#​9595)
• Adding an optional auth token to the TextEmbeddingInference class (#​9606)
• fix: out of index get latest tool call (#​9608)
• fix(azure_openai.py): add missing return to subclass override (#​9598)
• fix mix up b/w 'formatted' and 'format' params for ollama api call (#​9594)

v0.9.16.post1

Compare Source

v0.9.16

Compare Source

New Features

• agent refactor: step-wise execution (#​9584)
• Add OpenRouter, with Mixtral demo (#​9464)
• Add hybrid search to neo4j vector store (#​9530)
• Add support for auth service accounts for Google Semantic Retriever (#​9545)

Bug Fixes / Nits

• Fixed missing default=None for LLM.system_prompt (#​9504)
• Fix #​9580 : Incorporate metadata properly (#​9582)
• Integrations: Gradient[Embeddings,LLM] - sdk-upgrade (#​9528)
• Add mixtral 8x7b model to anyscale available models (#​9573)
• Gemini Model Checks (#​9563)
• Update OpenAI fine-tuning with latest changes (#​9564)
• fix/Reintroduce WHERE filter to the Sparse Query for PgVectorStore (#​9529)
• Update Ollama API to ollama v0.1.16 (#​9558)
• ollama: strip invalid formatted option (#​9555)
• add a device in optimum push #​9541 (#​9554)
• Title vs content difference for Gemini Embedding (#​9547)
• fix pydantic fields to float (#​9542)

v0.9.15.post2

Compare Source

v0.9.15.post1

Compare Source

v0.9.15

Compare Source

New Features

• Added full support for Google Gemini text+vision models (#​9452)
• Added new Google Semantic Retriever (#​9440)
• added from_existing() method + async support to OpenAI assistants (#​9367)

Bug Fixes / Nits

• Fixed huggingface LLM system prompt and messages to prompt (#​9463)
• Fixed ollama additional kwargs usage (#​9455)

v0.9.14.post3

Compare Source

v0.9.14.post2

Compare Source

v0.9.14.post1

Compare Source

v0.9.14

Compare Source

New Features

• Add MistralAI LLM (#​9444)
• Add MistralAI Embeddings (#​9441)
• Add Ollama Embedding class (#​9341)
• Add FlagEmbeddingReranker for reranking (#​9285)
• feat: PgVectorStore support advanced metadata filtering (#​9377)
• Added sql_only parameter to SQL query engines to avoid executing SQL (#​9422)

Bug Fixes / Nits

• Feat/PgVector Support custom hnsw.ef_search and ivfflat.probes (#​9420)
• fix F1 score definition, update copyright year (#​9424)
• Change more than one image input for Replicate Multi-modal models from error to warning (#​9360)
• Removed GPT-Licensed aiostream dependency (#​9403)
• Fix result of BedrockEmbedding with Cohere model (#​9396)
• Only capture valid tool names in react agent (#​9412)
• Fixed top_k being multiplied by 10 in azure cosmos (#​9438)
• Fixed hybrid search for OpenSearch (#​9430)

Breaking Changes

• Updated the base LLM interface to match LLMPredictor (#​9388)
• Deprecated LLMPredictor (#​9388)

v0.9.13

Compare Source

New Features

• Added batch prediction support for LabelledRagDataset (#​9332)

Bug Fixes / Nits

• Fixed save and load for faiss vector store (#​9330)

v0.9.12

Compare Source

New Features

• Added an option reuse_client to openai/azure to help with async timeouts. Set to False to see improvements (#​9301)
• Added support for vLLM llm (#​9257)
• Add support for python 3.12 (#​9304)
• Support for claude-2.1 model name (#​9275)

Bug Fixes / Nits

• Fix embedding format for bedrock cohere embeddings (#​9265)
• Use delete_kwargs for filtering in weaviate vector store (#​9300)
• Fixed automatic qdrant client construction (#​9267)

v0.9.11.post1

Compare Source

v0.9.11

Compare Source

New Features

• Make reference_contexts optional in LabelledRagDataset (#​9266)
• Re-organize download module (#​9253)
• Added document management to ingestion pipeline (#​9135)
• Add docs for LabelledRagDataset (#​9228)
• Add submission template notebook and other doc updates for LabelledRagDataset (#​9273)

Bug Fixes / Nits

• Convert numpy to list for InstructorEmbedding (#​9255)

v0.9.10

Compare Source

New Features

• Advanced Metadata filter for vector stores (#​9216)
• Amazon Bedrock Embeddings New models (#​9222)
• Added PromptLayer callback integration (#​9190)
• Reuse file ids for OpenAIAssistant (#​9125)

Breaking Changes / Deprecations

• Deprecate ExactMatchFilter (#​9216)

v0.9.9

Compare Source

New Features

• Add new abstractions for LlamaDataset's (#​9165)
• Add metadata filtering and MMR mode support for AstraDBVectorStore (#​9193)
• Allowing newest scikit-learn versions (#​9213)

Breaking Changes / Deprecations

• Added LocalAI demo and began deprecation cycle (#​9151)
• Deprecate QueryResponseDataset and DatasetGenerator of evaluation module (#​9165)

Bug Fixes / Nits

• Fix bug in download_utils.py with pointing to wrong repo ([#​9215](https://togithub.com/run-llama/l

---

Configuration

📅 Schedule: Branch creation - "" (UTC), Automerge - At any time (no schedule defined).

🚦 Automerge: Disabled by config. Please merge this manually once you are satisfied.

♻ Rebasing: Whenever PR becomes conflicted, or you tick the rebase/retry checkbox.

🔕 Ignore: Close this PR and you won't be reminded about this update again.

---

• If you want to rebase/retry this PR, check this box

---

This PR has been generated by Renovate Bot.

[stoat-app]

Easy and customizable dashboards for your build system. Learn more about Stoat ↗︎

Static Hosting

Name	Link	Commit	Status
api-coverage	Visit	e4268c9	✅
rtc-coverage	Visit	e4268c9	✅
core-coverage	Visit	e4268c9	✅
cron-coverage	Visit	e4268c9	✅
email-coverage	Visit	e4268c9	✅
worker-coverage	Visit	e4268c9	✅
api-test-results	Visit	e4268c9	✅
graphql-coverage	Visit	e4268c9	✅
rtc-test-results	Visit	e4268c9	✅
core-test-results	Visit	e4268c9	✅
cron-test-results	Visit	e4268c9	✅
email-test-results	Visit	e4268c9	✅
worker-test-results	Visit	e4268c9	✅
graphql-test-results	Visit	e4268c9	✅

Job Runtime