Backfill issue -393 — The AI Toolchain

EDITIONS tail grep head diff uniq

The daily firehose — everything the toolchain shipped today, already filtered.

// HOW THIS ISSUE IS MADE

We read every release from the 174 tools on our watchlist at the source — GitHub and GitLab release notes, vendor release pages and changelogs, project blogs and feeds, vendor press releases, and the source code behind the tag. Bug-fix-only releases and non-product newsroom noise are dropped; what's left is summarized down to the new capability, how to try it, and any screenshots or videos the release itself published. Every entry links to the sources it was built from.

VIEW

ISSUE VIEW full issue

Do you prefer this view?

$ tct list   # 9 tools matched

AI & LLM Tooling

◆ AI Agent Frameworks

deepset Haystack

v1.14.0 3 RELEASES · 2023-02-02 → 2023-02-28 NOTES STABLE

Haystack v1.14.0 adds Shaper, PromptNode run_batch/model_kwargs/top_k, IVF+PQ for OpenSearch, JsonConverter, and more.

└──▷ GET THIS VERSION

$ git clone --branch v1.14.0 https://github.com/deepset-ai/haystack.git
# already have the repo? check out this version:
$ git checkout v1.14.0

└──▷ USE IT

Pass model-specific generation parameters to PromptNode at initialisation, such as temperature and stop sequences.

python

from haystack.nodes import PromptNode
node = PromptNode('gpt-3.5-turbo', model_kwargs={'temperature': 0.2, 'stop': ['\n']})

Initialise OpenSearchDocumentStore with IVF+Product Quantization so the index is trained automatically on first use.

python

from haystack.document_stores import OpenSearchDocumentStore
store = OpenSearchDocumentStore(
    index='my_index',
    embedding_field='embedding',
    embedding_dim=768,
    ivf_train_size=10000
)

›Adds Shaper node to transform and reshape data between pipeline components, usable independently or as a PromptNode helper.
›Adds run_batch method to PromptNode for batch inference.
›Adds model_kwargs option to PromptNode for passing arbitrary model parameters.
›Adds top_k parameter to PromptNode.
›Exposes output_variable in PromptNode result.

+15 moreshow less

›Adds train_index method and ivf_train_size initialisation parameter to OpenSearchDocumentStore for IVF and IVF with Product Quantization index training.
›Adds JsonConverter node for converting JSON inputs in pipelines.
›Adds frontmatter-to-meta extraction in MarkdownConverter.
›Adds page range support to PDF converters.
›Adds use_prefiltering parameter to DeepsetCloudDocumentStore.
›Adds BM25 support for tables in InMemoryDocumentStore.
›Adds support for custom headers in document stores.
›Adds support for multiple RayPipeline instances running concurrently.
›Allows all training options for SentenceTransformers EmbeddingRetriever.
›Adds user-configurable timeout for remote APIs.
›Enables secure model loading by default.
›Adds OpenAIError to the retry mechanism.
›Warns users when max_tokens is too short for OpenAI models.
›Includes testing facilities in the haystack package for downstream consumers.
›Supports multiple document_ids in the Answer object for generative QA.

└──▷ BREAKING ON UPGRADE

!The REST API schema for tables has been updated to be consistent with Document.to_dict; existing table schema integrations may require adjustment.
!The Answer object now supports multiple document_ids (previously a single value); code that assumes a single document_id field will need to be updated.
!Defaults for OpenAIAnswerGenerator have changed; existing pipelines relying on previous defaults may behave differently after upgrade.

2 more releases in this issue · 2023-02-02 → 2023-02-28

v1.13.2 NOTES STABLE

Haystack v1.13.2 adds use_prefiltering parameter to DeepsetCloudDocumentStore

└──▷ GET THIS VERSION

$ git clone --branch v1.13.2 https://github.com/deepset-ai/haystack.git
# already have the repo? check out this version:
$ git checkout v1.13.2

›Adds use_prefiltering parameter to DeepsetCloudDocumentStore to control whether pre-filtering is applied during document retrieval.

v1.13.1 NOTES STABLE

Haystack v1.13.1 adds the Shaper component and frontmatter-to-meta extraction in MarkdownConverter.

└──▷ GET THIS VERSION

$ git clone --branch v1.13.1 https://github.com/deepset-ai/haystack.git
# already have the repo? check out this version:
$ git checkout v1.13.1

›Adds Shaper component for reshaping and transforming data between pipeline nodes.
›Adds frontmatter extraction to meta in MarkdownConverter, surfacing YAML/TOML front matter as structured document metadata.

Was this useful?

LangChain

v0.0.96 21 RELEASES · 2023-02-02 → 2023-02-28 NOTES STABLE

LangChain v0.0.96 adds image file and iFixit document loaders plus partial variables for prompts.

└──▷ GET THIS VERSION

$ git clone --branch v0.0.96 https://github.com/langchain-ai/langchain.git
# already have the repo? check out this version:
$ git checkout v0.0.96

›Adds partial variables support for prompt templates, enabling pre-filling of template variables at definition time.
›Adds a new document loader for image files.
›Adds a new iFixit document loader for ingesting iFixit repair guides and wikis.

20 more releases in this issue · 2023-02-02 → 2023-02-28

v0.0.95 NOTES STABLE

LangChain v0.0.95 adds CoNLL-U loader, AtlasDB and Deep Lake vector store integrations, and Weaviate certainty search parameter

└──▷ GET THIS VERSION

$ git clone --branch v0.0.95 https://github.com/langchain-ai/langchain.git
# already have the repo? check out this version:
$ git checkout v0.0.95

›Adds certainty as a supported parameter for similarity_search in the Weaviate vector store integration.
›Adds a CoNLL-U document loader for ingesting CoNLL-U formatted corpus files.
›Adds AtlasDB as a supported vector store integration.
›Adds Deep Lake as a supported vector store integration.
›Adds an indexing pipeline capability.

+1 moreshow less

›Adds a copy-paste document loader.

v0.0.94 NOTES STABLE

LangChain v0.0.94 adds LLM integrations, new document loaders, and a SearxNG query suffix parameter.

└──▷ GET THIS VERSION

$ git clone --branch v0.0.94 https://github.com/langchain-ai/langchain.git
# already have the repo? check out this version:
$ git checkout v0.0.94

└──▷ USE IT

Append a site-scoping suffix to every SearxNG query to restrict results to a domain.

python

from langchain.utilities import SearxSearchWrapper

search = SearxSearchWrapper(
    searx_host="http://localhost:8080",
    query_suffix="site:docs.python.org"
)
result = search.run("asyncio event loop")
print(result)

Load a Jupyter Notebook as a LangChain Document for ingestion into a vector store.

python

from langchain.document_loaders import NotebookLoader

loader = NotebookLoader("analysis.ipynb")
docs = loader.load()
print(docs[0].page_content[:500])

Load a Word document for use in a retrieval-augmented generation pipeline.

python

from langchain.document_loaders import UnstructuredWordDocumentLoader

loader = UnstructuredWordDocumentLoader("report.docx")
docs = loader.load()
print(docs[0].page_content[:500])

›Adds query_suffix parameter to the SearxNG search integration, allowing extra terms to be appended to every search query.
›Adds new LLM provider integrations: Writer, Banana, Modal, and StochasticAI.
›Adds a document loader for Jupyter Notebook (.ipynb) files.
›Adds a document loader for Microsoft Word documents.
›Adds a Facebook data loader (Harrison/fb loader).

+3 moreshow less

›Exposes log probabilities (logprobs) from OpenAI LLM responses.
›Exposes additional Cohere generation parameters (Harrison/cohere params).
›Adds source document tracking in retrieval chains (Harrison/source docs).

v0.0.93 NOTES STABLE

LangChain v0.0.93 adds Aleph Alpha and DeepInfra LLM integrations plus an IFTTT tool.

└──▷ GET THIS VERSION

$ git clone --branch v0.0.93 https://github.com/langchain-ai/langchain.git
# already have the repo? check out this version:
$ git checkout v0.0.93

›Adds Aleph Alpha LLM integration, expanding the set of supported language model providers.
›Adds DeepInfra LLM integration, enabling inference through the DeepInfra platform.
›Adds an IFTTT tool, allowing agents to trigger IFTTT webhooks and automations.

v0.0.92 NOTES STABLE

LangChain v0.0.92 adds OpenSearch vector store, GitBook loader, StdIn tool, and a reworked callback system.

└──▷ GET THIS VERSION

$ git clone --branch v0.0.92 https://github.com/langchain-ai/langchain.git
# already have the repo? check out this version:
$ git checkout v0.0.92

└──▷ USE IT

Load GitBook documentation into LangChain for question-answering over internal or public wikis.

python

from langchain.document_loaders import GitbookLoader

loader = GitbookLoader('https://docs.example.com')
docs = loader.load()

›Adds OpenSearch as a supported vector database for similarity search and storage.
›Adds a StdIn interaction tool, enabling agents to prompt the user for input via standard input.
›Adds a GitBook document loader for ingesting GitBook content into LangChain pipelines.
›Adds reworked callback system via the callback changes RFC, enabling more flexible chain event handling.
›Adds ability to override default verbose and memory settings when loading a chain.

+1 moreshow less

›Adds add_documents support, enabling direct document ingestion into vector stores.

v0.0.91 NOTES STABLE

LangChain v0.0.91 adds a Markdown text splitter, custom prompt support for VectorDBQA, and a top-k context control for ChatVectorDBChain.

└──▷ GET THIS VERSION

$ git clone --branch v0.0.91 https://github.com/langchain-ai/langchain.git
# already have the repo? check out this version:
$ git checkout v0.0.91

└──▷ USE IT

Limit retrieved context chunks when building a chat-over-docs chain, reducing token usage while keeping answers grounded.

python

from langchain.chains import ChatVectorDBChain

chain = ChatVectorDBChain.from_llm(
    llm=llm,
    vectorstore=vectorstore,
    top_k_docs_for_context=3
)

Split a Markdown document on its natural headings and sections rather than fixed character counts.

python

from langchain.text_splitter import MarkdownTextSplitter

splitter = MarkdownTextSplitter(chunk_size=500, chunk_overlap=50)
docs = splitter.create_documents([markdown_text])

›Adds top_k_docs_for_context parameter to ChatVectorDBChain to control how many retrieved chunks are used as context.
›Supports passing custom prompts into VectorDBQA chains.
›Adds a Markdown-aware text splitter (MarkdownTextSplitter) for more semantically coherent document chunking.
›Improves DirectoryLoader with enhancements to how directories of documents are loaded.

v0.0.90 NOTES STABLE

LangChain v0.0.90 adds a Constitutional AI chain and self-hosted Runhouse integration.

└──▷ GET THIS VERSION

$ git clone --branch v0.0.90 https://github.com/langchain-ai/langchain.git
# already have the repo? check out this version:
$ git checkout v0.0.90

›Adds a Constitutional chain, enabling Constitutional AI-style critique-and-revision pipelines over LLM outputs.
›Adds self-hosted Runhouse integration as a new LLM/compute backend option.

v0.0.89 NOTES STABLE

LangChain v0.0.89 adds HN and SRT loaders, .ppt support, source document returns, and a new ToolKit concept.

└──▷ GET THIS VERSION

$ git clone --branch v0.0.89 https://github.com/langchain-ai/langchain.git
# already have the repo? check out this version:
$ git checkout v0.0.89

›Adds .ppt file support to UnstructuredPowerPointLoader (previously only .pptx was supported).
›Introduces HNLoader for loading Hacker News content.
›Adds an SRT (subtitle) file loader for ingesting subtitle documents.
›Enables ChatVectorDBChain to return source documents alongside answers.
›Introduces a ToolKit concept and makes Tools its own model, enabling grouped tool management for agents.

v0.0.88 NOTES STABLE

LangChain v0.0.88 adds Google Search via serper.dev, SearxNG meta-search, FAISS vector search, async PromptLayer, and new Telegram/Evernote loaders

└──▷ GET THIS VERSION

$ git clone --branch v0.0.88 https://github.com/langchain-ai/langchain.git
# already have the repo? check out this version:
$ git checkout v0.0.88

└──▷ USE IT

Retrieve documents from a FAISS index using a raw embedding vector instead of a text query — useful when you already have an embedding from another model.

python

# existing_embedding is a list[float] produced by your embedding model
docs = vectorstore.similarity_search_by_vector(existing_embedding, k=5)

›Adds SearxNG meta search API helper for querying multiple search engines through a self-hosted SearxNG instance.
›Adds Google Search API integration via serper.dev wrapper, enabling Google search tool use without a direct Google API key.
›Adds similarity search by vector in FAISS, allowing retrieval using a raw embedding vector rather than a query string.
›Adds async API support to PromptLayerOpenAI LLM, enabling non-blocking LLM calls with prompt logging.
›Adds element metadata to the Unstructured document loader, surfacing richer per-element context from parsed files.

+4 moreshow less

›Adds a Telegram document loader for ingesting Telegram chat exports.
›Adds an Evernote document loader for ingesting Evernote content.
›Adds chat QA with sources, enabling question-answering chains over chat history that return source attribution.
›Adds semantic subset support for working with semantically filtered subsets of documents.

v0.0.87 NOTES STABLE

LangChain v0.0.87 enables streaming responses for the OpenAI LLM integration.

└──▷ GET THIS VERSION

$ git clone --branch v0.0.87 https://github.com/langchain-ai/langchain.git
# already have the repo? check out this version:
$ git checkout v0.0.87

›Enables streaming support for the OpenAI LLM, allowing token-by-token output as the model generates responses.

v0.0.86 NOTES STABLE

LangChain v0.0.86 adds Chroma persistence and four new LLM integrations: GooseAI, CerebriumAI, Petals, and ForefrontAI.

└──▷ GET THIS VERSION

$ git clone --branch v0.0.86 https://github.com/langchain-ai/langchain.git
# already have the repo? check out this version:
$ git checkout v0.0.86

›Adds GooseAI, CerebriumAI, Petals, and ForefrontAI as new LLM integrations.
›Adds persistence support for the Chroma vector store.
›Adds automatic retry on openai.error.ServiceUnavailableError for OpenAI calls.

v0.0.85 NOTES STABLE

LangChain v0.0.85 adds Chroma as a supported vector store integration.

└──▷ GET THIS VERSION

$ git clone --branch v0.0.85 https://github.com/langchain-ai/langchain.git
# already have the repo? check out this version:
$ git checkout v0.0.85

›Adds Chroma vector store integration, enabling Chroma as a retrieval backend for LangChain chains and agents.
›Adds a Knowledge Graph (KG) chain capability.

v0.0.84 NOTES STABLE

LangChain v0.0.84 adds a fake LLM for testing, PDFMiner loader, and unstructured document support.

└──▷ GET THIS VERSION

$ git clone --branch v0.0.84 https://github.com/langchain-ai/langchain.git
# already have the repo? check out this version:
$ git checkout v0.0.84

›Adds a fake LLM implementation for deterministic testing and development workflows without real model calls.
›Adds PDFMiner document loader for extracting text from PDF files.
›Adds unstructured document loader support for ingesting a broader range of document formats.

v0.0.83 NOTES STABLE

LangChain v0.0.83 adds an online PDF loader and an Airbyte integration.

└──▷ GET THIS VERSION

$ git clone --branch v0.0.83 https://github.com/langchain-ai/langchain.git
# already have the repo? check out this version:
$ git checkout v0.0.83

›Adds an online PDF document loader, enabling LangChain to ingest PDFs directly from URLs without downloading them first.
›Adds an Airbyte integration, allowing LangChain to load data from any Airbyte-supported source connector.

v0.0.82 NOTES STABLE

LangChain v0.0.82 adds UnstructuredURLLoader for loading documents directly from URLs.

└──▷ GET THIS VERSION

$ git clone --branch v0.0.82 https://github.com/langchain-ai/langchain.git
# already have the repo? check out this version:
$ git checkout v0.0.82

└──▷ USE IT

Load and parse web page content from a list of URLs for use in a retrieval pipeline.

python

from langchain.document_loaders import UnstructuredURLLoader

loader = UnstructuredURLLoader(urls=["https://example.com/report", "https://example.com/advisory"])
docs = loader.load()

›Adds UnstructuredURLLoader class for loading and parsing data from URLs into LangChain documents.
›Adds Evernote document loader integration.
›Adds batch embedding support to reduce API calls when embedding large document sets.

└──▷ BREAKING ON UPGRADE

!The sample_row_in_table_info parameter has been removed from the SQL database integration.

v0.0.81 NOTES STABLE

LangChain v0.0.81 adds webpage and Gutenberg book loading capabilities.

└──▷ GET THIS VERSION

$ git clone --branch v0.0.81 https://github.com/langchain-ai/langchain.git
# already have the repo? check out this version:
$ git checkout v0.0.81

›Adds webpage loading logic for ingesting web content as documents.
›Adds support for loading Gutenberg books as document sources.

v0.0.80 NOTES STABLE

LangChain v0.0.80 adds async support for OpenAI LLM, LLMChain, LLMMathChain, and Agent, plus a new Roam document loader.

└──▷ GET THIS VERSION

$ git clone --branch v0.0.80 https://github.com/langchain-ai/langchain.git
# already have the repo? check out this version:
$ git checkout v0.0.80

›Adds asyncio support for OpenAI LLM, LLMChain, LLMMathChain, and Agent, enabling non-blocking LLM calls in async Python applications.
›Adds a new Roam document loader for ingesting content from Roam Research databases.

v0.0.79 NOTES STABLE

LangChain v0.0.79 adds Anthropic, HuggingFace Inference Endpoint, GoogleDriveLoader, Obsidian loader, FAISS save/load, and document analysis.

└──▷ GET THIS VERSION

$ git clone --branch v0.0.79 https://github.com/langchain-ai/langchain.git
# already have the repo? check out this version:
$ git checkout v0.0.79

›Adds GoogleDriveLoader for loading documents directly from Google Drive.
›Adds Anthropic LLM integration as a new supported model provider.
›Adds HuggingFace Inference Endpoint as a new LLM backend.
›Adds Obsidian loader for ingesting notes from an Obsidian vault.
›Adds save/load support for FAISS vector stores, enabling persistence of indexed embeddings.

+5 moreshow less

›Adds analyze document chain for running analysis over a full document.
›Adds optional return of shell output on incorrect commands, surfacing error context from the shell tool.
›Adds i_end parameter to batch extraction for controlling extraction range.
›Adds prompt template prefix support for customizing how prompt templates are constructed.
›Adds configurable SQL row limits for SQL-based chains.

v0.0.78 NOTES STABLE

LangChain v0.0.78 adds a chat-over-documents chain, Unstructured file support, and prompt-from-string construction.

└──▷ GET THIS VERSION

$ git clone --branch v0.0.78 https://github.com/langchain-ai/langchain.git
# already have the repo? check out this version:
$ git checkout v0.0.78

›Adds a chat vector DB chain enabling conversational question-answering over vector store-backed document collections.
›Adds support for Unstructured document loading, allowing ingestion of a broader range of file formats into LangChain pipelines.
›Adds prompt template construction directly from a string, simplifying prompt creation without requiring a separate template file.

v0.0.77 NOTES STABLE

LangChain v0.0.77 adds token-based text splitting, automatic OpenAI retries, and Milvus vector store support.

└──▷ GET THIS VERSION

$ git clone --branch v0.0.77 https://github.com/langchain-ai/langchain.git
# already have the repo? check out this version:
$ git checkout v0.0.77

›Adds a token-based text splitter as an alternative to character-based splitting for more accurate chunking of LLM inputs.
›Adds automatic retry logic to the OpenAI LLM integration to handle transient API errors.
›Adds Milvus as a supported vector store integration.

v0.0.76 NOTES STABLE

LangChain v0.0.76 adds truncate param for CohereEmbeddings, InstructEmbeddings, PAL context passing, and a from-string method.

└──▷ GET THIS VERSION

$ git clone --branch v0.0.76 https://github.com/langchain-ai/langchain.git
# already have the repo? check out this version:
$ git checkout v0.0.76

└──▷ USE IT

Truncate long inputs at the end when generating embeddings with Cohere, avoiding token-limit errors in bulk pipelines.

python

from langchain.embeddings import CohereEmbeddings

embeddings = CohereEmbeddings(truncate='END')
vectors = embeddings.embed_documents([very_long_text])

›Adds truncate parameter to CohereEmbeddings to control how input text is truncated before embedding.
›Adds from_string class method for constructing chains directly from a string.
›Updates PAL to support passing local and global context to PythonREPL, enabling richer execution environments.
›Adds instruct embeddings support as a new embedding type.
›Enables PAL to return the generated code in addition to the result.

Was this useful?

LlamaIndex

v0.4.13 15 RELEASES · 2023-02-01 → 2023-02-25 NOTES STABLE

LlamaIndex 0.4.13 adds embedding-based KG Index queries and multi-file LlamaHub loader support via download_loader.

└──▷ GET THIS VERSION

$ git clone --branch v0.4.13 https://github.com/run-llama/llama_index.git
# already have the repo? check out this version:
$ git checkout v0.4.13

›Extends download_loader integration with LlamaHub to support complex loaders that require multiple files, enabling integrations such as the GitHub loader.
›Enables embedding-based querying of the Knowledge Graph (KG) Index as an alternative to exact keyword matching.

14 more releases in this issue · 2023-02-01 → 2023-02-25

v0.4.12 NOTES STABLE

LlamaIndex v0.4.12 lets you pass a nested index as table context to the SQL index, tackling large schema prompts.

└──▷ GET THIS VERSION

$ git clone --branch v0.4.12 https://github.com/run-llama/llama_index.git
# already have the repo? check out this version:
$ git checkout v0.4.12

›Adds support for passing table context — including another index as the context source — to the SQL index, enabling text-to-SQL over databases with too many tables and columns to fit in a single prompt.

v0.4.11 NOTES STABLE

LlamaIndex 0.4.11 adds async vector index construction and decouples vector storage from index logic.

└──▷ GET THIS VERSION

$ git clone --branch v0.4.11 https://github.com/run-llama/llama_index.git
# already have the repo? check out this version:
$ git checkout v0.4.11

›Adds async support to vector index construction, enabling non-blocking index builds.
›Decouples vector storage from index build and query logic, laying the groundwork for new vector store integrations.

v0.4.8 NOTES STABLE

LlamaIndex v0.4.8 adds customizable text splitters per index and a use_gpt_index_import option for LlamaHub loaders.

└──▷ GET THIS VERSION

$ git clone --branch v0.4.8 https://github.com/run-llama/llama_index.git
# already have the repo? check out this version:
$ git checkout v0.4.8

└──▷ USE IT

Retain legacy gpt_index imports in LlamaHub loaders while migrating to llama_index at your own pace.

python

from llama_index import download_loader

SimpleWebPageReader = download_loader('SimpleWebPageReader', use_gpt_index_import=True)

›Adds use_gpt_index_import option to download_loader — set to True to retain gpt_index imports when LlamaHub loaders now default to llama_index.
›Adds ability to customize the text splitter for a given index.

└──▷ BREAKING ON UPGRADE

!All LlamaHub loaders now import from llama_index instead of gpt_index by default; code relying on gpt_index imports from download_loader will break unless use_gpt_index_import=True is set.

v0.4.7 NOTES STABLE

LlamaIndex v0.4.7 adds a Playground module for comparing indexes, models, and embeddings side by side.

└──▷ GET THIS VERSION

$ git clone --branch v0.4.7 https://github.com/run-llama/llama_index.git
# already have the repo? check out this version:
$ git checkout v0.4.7

›New Playground module lets practitioners test multiple indexes, models, and embeddings simultaneously and compare results in one place.

v0.4.6 NOTES STABLE

LlamaIndex v0.4.6 adds async tree_summarize queries and embedding batching for 3-5x faster responses and faster vector index construction.

└──▷ GET THIS VERSION

$ git clone --branch v0.4.6 https://github.com/run-llama/llama_index.git
# already have the repo? check out this version:
$ git checkout v0.4.6

›Adds async support for tree_summarize queries, delivering 3-5x faster query responses.
›Adds embedding batching to accelerate vector index construction.

v0.4.5 NOTES STABLE

LlamaIndex 0.4.5 adds KG index triplet tracking in sources and reduces index JSON file size.

└──▷ GET THIS VERSION

$ git clone --branch v0.4.5 https://github.com/run-llama/llama_index.git
# already have the repo? check out this version:
$ git checkout v0.4.5

›Tracks triplets for the Knowledge Graph (KG) index in sources, making graph relationships visible in query provenance.
›Significantly reduces index JSON file size by removing unnecessary information.

v0.4.4 NOTES STABLE

LlamaIndex v0.4.4 adds QueryBundle and QueryTransform abstractions for finer control over query embedding and transformation.

└──▷ GET THIS VERSION

$ git clone --branch v0.4.4 https://github.com/run-llama/llama_index.git
# already have the repo? check out this version:
$ git checkout v0.4.4

›Adds QueryBundle abstraction to separate the query string from the string used for embedding lookup, enabling independent control of retrieval vs. generation queries.
›Adds QueryTransform class to transform queries within data structures, with HyDE (Hypothetical Document Embeddings) as the first implementation.

v0.4.3 NOTES STABLE

LlamaIndex v0.4.3 adds a Knowledge Graph index for triplet extraction and query-time KG traversal.

└──▷ GET THIS VERSION

$ git clone --branch v0.4.3 https://github.com/run-llama/llama_index.git
# already have the repo? check out this version:
$ git checkout v0.4.3

›Adds a Knowledge Graph index that builds a KG by extracting triplets from documents and leverages it at query time.

v0.4.2 NOTES STABLE

GPT Index 0.4.2 adds caching to download_loader and exposes Pinecone kwargs across all index operations.

└──▷ GET THIS VERSION

$ git clone --branch v0.4.2 https://github.com/run-llama/llama_index.git
# already have the repo? check out this version:
$ git checkout v0.4.2

›Adds caching to download_loader so loaders are served from local cache instead of re-downloading from llamahub.ai on every call.
›Exposes Pinecone kwargs on all index operations for the Pinecone index, enabling fine-grained control over Pinecone API calls.

v0.4.1 NOTES STABLE

LlamaIndex v0.4.1 adds an Azure OpenAI example notebook, a GitHub repository loader, and an updated OpenAI retry policy.

└──▷ GET THIS VERSION

$ git clone --branch v0.4.1 https://github.com/run-llama/llama_index.git
# already have the repo? check out this version:
$ git checkout v0.4.1

›Adds a GitHub repository loader for ingesting code and content directly from GitHub repositories.
›Adds an Azure OpenAI example notebook demonstrating integration with Azure-hosted OpenAI endpoints.
›Updates the retry policy for OpenAI API calls.

v0.4.0 NOTES STABLE

LlamaIndex v0.4.0 replaces print statements with full Python logger support throughout the codebase.

└──▷ GET THIS VERSION

$ git clone --branch v0.4.0 https://github.com/run-llama/llama_index.git
# already have the repo? check out this version:
$ git checkout v0.4.0

›Adds full Python logging module support throughout the codebase, replacing all print statements and enabling standard log routing, filtering, and formatting.

└──▷ BREAKING ON UPGRADE

!The verbose parameter has been removed from all APIs; configure output verbosity using Python's standard logging module instead.

v0.3.6 NOTES STABLE

LlamaIndex v0.3.6 adds an mbox parser so email archives can be fed directly into an index.

└──▷ GET THIS VERSION

$ git clone --branch v0.3.6 https://github.com/run-llama/llama_index.git
# already have the repo? check out this version:
$ git checkout v0.3.6

›Adds a parser/reader for .mbox files, enabling email archives to be ingested as index documents.

v0.3.5 NOTES STABLE

LlamaIndex v0.3.5 adds save/load from string for indices and graphs, and drops required query_configs for recursive queries.

└──▷ GET THIS VERSION

$ git clone --branch v0.3.5 https://github.com/run-llama/llama_index.git
# already have the repo? check out this version:
$ git checkout v0.3.5

›Adds save/load from string for indices and graphs, enabling persistence to sources beyond disk.
›Removes the requirement to specify query_configs for recursive queries — default configs are used automatically.

v0.3.2 NOTES STABLE

LlamaIndex v0.3.2 adds Qdrant as both a data source reader and a vector index store.

└──▷ GET THIS VERSION

$ git clone --branch v0.3.2 https://github.com/run-llama/llama_index.git
# already have the repo? check out this version:
$ git checkout v0.3.2

›Adds Qdrant integration, enabling Qdrant to be used both as a data reader (source) and as a vector store for your index.

Was this useful?

◆ AI Model & Data Infrastructure

Microsoft ONNX Runtime

v1.14.0 NOTES

ONNX Runtime v1.14.0 adds NUMA-aware threading, multi-stream GPU execution, Azure EP, Rust API preview, and ONNX opset 18 support.

└──▷ GET THIS VERSION

$ git clone --branch v1.14.0 https://github.com/microsoft/onnxruntime.git
# already have the repo? check out this version:
$ git checkout v1.14.0

›Adds new custom operator APIs enabling a custom op to wrap an entire model for inferencing with an external API or runtime.
›Adds new API to set intra-op thread affinity for fine-grained CPU pinning.
›Changes cudnn_conv_use_max_workspace default value to 1, improving CUDA EP transformer model performance.
›Supports ONNX 1.13 (opset 18), expanding the set of recognized operators and model formats.
›ORT thread pool is now NUMA-aware, improving multi-socket CPU performance.

+19 moreshow less

›Multi-stream execution provider refactoring improves GPU utilization and memory reuse across parallel inference streams for CUDA, TensorRT, and ROCm EPs.
›Adds Stream interface in ExecutionProvider API, allowing EP developers to customize stream implementations.
›New [Preview] Rust API for ORT, available to build from the main branch.
›Adds [Preview] Azure Execution Provider supporting AzureML-hosted models via Triton for hybrid on-device and on-cloud inferencing.
›Adds quantization support with AMX on Sapphire Rapids processors.
›ROCm EP now supports ROCm 5.4 and is GA-ready.
›TensorRT EP adds support for TensorRT 8.5 GA versions.
›OpenVINO EP adds support for OpenVINO 2022.3.
›DirectML EP updated to DML 1.10.1, adding operators NonZero, Shape, Size, Attention, EmbedLayerNorm, SkipLayerNorm, and BiasGelu.
›DirectML EP adds additional data type support for Abs, Sign, and Where operators.
›DirectML EP enables SetOptimizedFilePath export/reload.
›XNNPACK EP adds iOS build support and is now included in the onnxruntime-c iOS package.
›XNNPACK EP adds support for using the ORT allocator in XNNPACK kernels to minimize memory usage.
›XNNPACK EP adds support for additional commonly used operators.
›Web build now includes onnxruntime-extensions (NLP-centric) by default.
›Adds experimental utility functions for exchanging data between images and tensors in the JavaScript API.
›FP16 optimizer exposed in torch-ort for use with DeepSpeed/Apex.
›Mobile pre/post processing pipeline supports updating MobileNet and super-resolution models to embed pre/post processing (including JPG/PNG conversion custom ops) directly into the model via the onnxruntime-extensions Python package.
›Stable Diffusion training optimizations, including support for Resize and InstanceNorm gradients, with ORT-enabled examples added to the Hugging Face diffusers library.

└──▷ BREAKING ON UPGRADE

!Building ORT from source now requires cmake version >=3.24 instead of >=3.18.

Was this useful?

NVIDIA Triton Inference Server

v2.30.0 NOTES

Triton v2.30.0 adds user-defined custom batching strategies to the dynamic batcher.

└──▷ GET THIS VERSION

$ git clone --branch v2.30.0 https://github.com/triton-inference-server/server.git
# already have the repo? check out this version:
$ git checkout v2.30.0

›Enables user-defined batching constraints in the dynamic batcher via custom batching strategies, letting operators control how requests are grouped into batches.
›Relaxes the Python client gRPC version requirement, broadening compatible client environments.

Was this useful?

◆ AI Coding Agents

Zed

v0.74.2 4 RELEASES · 2023-02-01 → 2023-02-22 NOTES STABLE

Zed v0.74.2 adds toggle soft wrap, Reveal in Finder, and a setting to disable automatic updates.

└──▷ GET THIS VERSION

$ git clone --branch v0.74.2 https://github.com/zed-industries/zed.git
# already have the repo? check out this version:
$ git checkout v0.74.2

›Adds toggle soft wrap command to the editor.
›Adds 'Reveal in Finder' option to the project panel context menu.
›Adds a setting to turn off automatic updates.
›Improves autocompletion sorting by taking casing into consideration.

3 more releases in this issue · 2023-02-01 → 2023-02-22

v0.73.3 NOTES STABLE

Zed v0.73.3 adds YAML and Lua language support plus a menu bar screen-sharing indicator.

└──▷ GET THIS VERSION

$ git clone --branch v0.73.3 https://github.com/zed-industries/zed.git
# already have the repo? check out this version:
$ git checkout v0.73.3

›Adds YAML language support.
›Adds Lua language support.
›Adds a system menu bar icon that indicates when screen sharing is active during collaboration sessions.

v0.72.2 NOTES STABLE

Zed v0.72.2 adds contact removal, window position restoration, and an open licenses command.

└──▷ GET THIS VERSION

$ git clone --branch v0.72.2 https://github.com/zed-industries/zed.git
# already have the repo? check out this version:
$ git checkout v0.72.2

›Adds open licenses command to display Zed's dependency list and their licenses.
›Adds the ability to remove contacts from the contact list.

v0.71.3 NOTES STABLE

Zed v0.71.3 adds CLI file creation, a confirm_quit setting, Finder/drag-drop file opening, and in-app feedback.

└──▷ GET THIS VERSION

$ git clone --branch v0.71.3 https://github.com/zed-industries/zed.git
# already have the repo? check out this version:
$ git checkout v0.71.3

›Adds confirm_quit setting to control whether Zed prompts before quitting.
›Adds the ability to create new files via Zed's CLI.
›Adds an open licenses command listing all open-source dependencies and their licenses.
›Adds syntax highlighting for fenced code blocks in Markdown files.
›Adds support for opening files via 'Open With' in macOS Finder and by dragging files onto the application icon.

+2 moreshow less

›Adds in-app feedback: clicking 'Give Feedback' opens an editor tab to write and submit feedback without leaving Zed.
›Adds date stamps to Zed log file entries.

Was this useful?

Other / Uncategorized

◆ AI OBSERVABILITY

Arize Phoenix

v0.0.2rc1 NOTES

Arize Phoenix v0.0.2rc1 adds time-series graphs, timeslice selection, dataset time presets, record-level retrieval, and embedding validation.

└──▷ GET THIS VERSION

$ git clone --branch v0.0.2rc1 https://github.com/Arize-ai/phoenix.git
# already have the repo? check out this version:
$ git checkout v0.0.2rc1

›Adds embedding validation to catch schema issues before data is ingested.
›Adds time-series graphing to visualize model metrics over time.
›Adds a timeslice selection UI to drill into a specific point on the time-series graph.
›Adds a dataset time preset selector for quickly scoping the active dataset to a time window.
›Adds record-level retrieval to inspect individual records in the UI.

+1 moreshow less

›Adds server startup polling so the library waits for the Phoenix server to be ready before proceeding.

Was this useful?

◆ VECTOR DB RAG

Milvus

v2.2.3 NOTES

Milvus 2.2.3 adds rolling upgrades via Helm, coordinator active-standby HA, and new GetLoadState and WalkWithPrefix APIs.

└──▷ GET THIS VERSION

$ git clone --branch v2.2.3 https://github.com/milvus-io/milvus.git
# already have the repo? check out this version:
$ git checkout v2.2.3

└──▷ USE IT

Reload a collection index after a bulk-insert completes so newly imported vectors are immediately searchable without a full unload/reload cycle.

python

from pymilvus import connections, Collection

connections.connect(host='localhost', port='19530')
col = Collection('my_collection')
col.load(refresh=True)

›Adds GetLoadState API to query the load state of a collection or partition.
›Adds WalkWithPrefix API to the MetaKv interface for prefix-based metadata traversal.
›Adds a refresh option to LoadCollection and LoadPartition APIs to support reloading after bulk-insert.
›Adds a new segment metric counting the number of binlog files per segment to Milvus monitoring.
›Adds a segment seal policy triggered by the number of binlog files.

+8 moreshow less

›Adds GC for snapshot KV based on time-travel to reclaim stale metadata storage.
›Introduces rolling upgrade support for Milvus clusters installed via Helm charts, allowing the cluster to serve requests during upgrades without downtime.
›Introduces Coordinator HA (active-standby mode) for RootCoord and QueryCoord, ensuring failure recovery within 30 seconds.
›Adds bulk-insert progress reporting so operators can track import status in real time.
›Reduces memory usage during scalar field indexing and bulk-insert operations, and adds stream-reading for NumPy data imports.
›Surfaces explicit denial reasons when Milvus rejects read/write requests.
›Improves ListSegments performance for DataCoord catalog and LoadWithPrefix performance for SuffixSnapshot.
›Upgrades Knowhere to 1.3.8 for improved ANN search performance.

└──▷ BREAKING ON UPGRADE

!The maximum number of fields allowed in a collection is reduced from 256 to 64.

Was this useful?

Qdrant

v1.0.0 NOTES

Qdrant v1.0 adds collection alias listing, snapshot deletion, collection initialization from another collection, and opt-in read/write consistency guarantees.

└──▷ GET THIS VERSION

$ git clone --branch v1.0.0 https://github.com/qdrant/qdrant.git
# already have the repo? check out this version:
$ git checkout v1.0.0

›New API to list collection aliases.
›New API to delete snapshots.
›New API to initialize a collection from another collection, enabling quick experiments over different indexing parameters and seamless shard-count scaling when combined with aliases.
›Read operations gain opt-in consistency guarantees, ensuring consistent reads even from an inconsistent cluster.
›Write operations gain opt-in ordering guarantees, ensuring writes are ordered across parallel overlapping requests.

+3 moreshow less

›Telemetry data collection is now enabled by default and can be disabled via CLI or environment variable.
›Adds Windows platform support.
›Allows overriding the maximum number of CPUs via environment variable, useful for Docker containers and Kubernetes.

Was this useful?

// END OF TRANSMISSION

The AI Toolchain · curated daily · February 28, 2023
Releases & features from the open-source and commercial AI tools that practitioners actually run.

The AI Toolchain — issue -393, February 28, 2023