The AI Toolchain — issue -379, April 30, 2024

Aider v0.29.0 adds direct LLM provider connections and new model-selection flags including --opus, --sonnet, and --weak-model.

$ git clone --branch v0.29.0 https://github.com/Aider-AI/aider.git
# already have the repo? check out this version:
$ git checkout v0.29.0

$ aider --opus --weak-model claude-3-haiku-20240307

$ aider --sonnet --anthropic-api-key <your-api-key>

• ›Adds --weak-model <model-name> flag to specify which model handles commit messages and chat history summarization independently of the main model.
• ›Adds --opus and --sonnet CLI flags for direct Anthropic model selection.
• ›Adds --4-turbo-vision CLI flag for GPT-4 Turbo with vision support.
• ›Adds --anthropic-api-key CLI flag to supply an Anthropic API key directly.
• ›Adds direct connection support for Anthropic, Cohere, Gemini, and many other LLM providers.

Aider v0.28.0 adds support for gpt-4-turbo-2024-04-09 and gpt-4-turbo models.

$ git clone --branch v0.28.0 https://github.com/Aider-AI/aider.git
# already have the repo? check out this version:
$ git checkout v0.28.0

$ aider --model gpt-4-turbo

• ›Adds support for the gpt-4-turbo-2024-04-09 and gpt-4-turbo models (61.7% on Exercism benchmark, 34.1% on refactoring/laziness benchmark); default remains gpt-4-1106-preview.

Zed v0.131.6 adds vim-surround, markdown preview enhancements, and new line_indicator_format and show_nav_history_buttons settings.

$ git clone --branch v0.131.6 https://github.com/zed-industries/zed.git
# already have the repo? check out this version:
$ git checkout v0.131.6

{
  "line_indicator_format": "short"
}

{
  "tab_bar": {
    "show_nav_history_buttons": false
  }
}

• ›Adds line_indicator_format setting (e.g. "short") to make the status bar line/column indicator more compact.
• ›Adds tab_bar.show_nav_history_buttons setting to show or hide navigation history buttons in the tab bar.
• ›Adds ui_font_family setting in settings.json to override the UI font (previously defaulted to Zed Sans, now the system UI font on macOS).
• ›Adds vim search motions in visual modes as targets for operators like d, c, y.
• ›Adds action to open markdown preview in the same pane.

Zed v0.130.4 adds git blame, tab switcher, transparent backgrounds, Emmet, and new formatter/ESLint options.

$ git clone --branch v0.130.4 https://github.com/zed-industries/zed.git
# already have the repo? check out this version:
$ git checkout v0.130.4

// In Zed settings.json
{"languages": {"JavaScript": {"formatter": {"code_actions": {"source.fixAll.eslint": true}}}}}

// In Zed settings.json
{"lsp": {"eslint": {"settings": {"problems": {"shortenToSingleLine": true}}}}}

• ›Adds editor: toggle git blame command (bound to cmd-alt-g b by default) to toggle a sidebar showing git blame information for the current buffer.
• ›Adds a new formatter/format_on_save option code_actions that uses language server code actions to format a buffer — e.g. {"languages": {"JavaScript": {"formatter": {"code_actions": {"source.fixAll.eslint": true}}}}} — allowing ESLint-only formatting without running Prettier.
• ›Adds support for configuring ESLint problems settings via {"lsp": {"eslint": {"settings": {"problems": {"shortenToSingleLine": true}}}}}.
• ›Adds support for transparent and blurred window backgrounds on macOS via a new background.appearance theme key accepting opaque, transparent, or blurred, with alpha values in colors.
• ›Adds a tab switcher accessible via ctrl-tab and ctrl-shift-tab, working in both the main workspace and terminal panel.

Zed v0.129.1 adds Emmet support, assistant toggle, gopls binary config, fast file-switch, and regex newline/tab replacements.

$ git clone --branch v0.129.1 https://github.com/zed-industries/zed.git
# already have the repo? check out this version:
$ git checkout v0.129.1

{
  "lsp": {
    "gopls": {
      "binary": {
        "path": "/home/user/go/bin/gopls",
        "arguments": ["-debug=0.0.0.0:8080"]
      }
    }
  }
}

{
  "lsp": {
    "elixir-ls": {
      "settings": {
        "mixEnv": "test"
      }
    }
  }
}

• ›Adds assistant.enabled setting to enable or disable the AI Assistant.
• ›Adds gopls binary path and arguments config under {"lsp": {"gopls": {"binary": {"path": "...", "arguments": [...]}}}} in user settings.
• ›Adds workspace configuration support for elixir-ls via {"lsp": {"elixir-ls": {"settings": {...}}}} in Zed settings.
• ›Adds support for inserting newlines (\n) and tabs (\t) in editor Regex search replacements.
• ›Adds a keybinding (shift-f12) for the editor::GoToImplementation action.

Haystack v2.0.1 adds streaming support to HuggingFaceLocalGenerator and introduces a new SparseEmbedding class.

$ git clone --branch v2.0.1 https://github.com/deepset-ai/haystack.git
# already have the repo? check out this version:
$ git checkout v2.0.1

• ›Adds streaming_callback parameter to HuggingFaceLocalGenerator to handle streaming responses.
• ›Introduces new SparseEmbedding class for storing sparse vector representations of a Document, laying groundwork for Sparse Embedding Retrieval with forthcoming Sparse Embedders and Sparse Embedding Retrievers.

Haystack v1.25.2 adds response_format, seed, and prompt-truncation toggle to OpenAI/Azure invocation layers.

$ git clone --branch v1.25.2 https://github.com/deepset-ai/haystack.git
# already have the repo? check out this version:
$ git checkout v1.25.2

• ›Adds response_format and seed parameters to the OpenAI and Azure OpenAI invocation layers, enabling structured output control and reproducible sampling.
• ›Adds a boolean parameter to toggle prompt truncation in invocation layers, giving callers explicit control over whether long prompts are silently cut.

LangChain v0.1.16 adds tool-call messages to core, Mustache prompt templates, a Chroma partner package, and updated agent tool-call support.

$ git clone --branch v0.1.16 https://github.com/langchain-ai/langchain.git
# already have the repo? check out this version:
$ git checkout v0.1.16

• ›Adds Mustache prompt template support to core via mustache prompt templates, enabling Mustache syntax alongside existing template formats.
• ›Adds a new tool calls message type to core, with tool_calls included in AI message chunk serialization, giving agents and chains a standardized way to represent tool invocations.
• ›Updates agents to use tool-call messages, aligning agent execution with the new core tool-call message format.
• ›Adds langchain-chroma as a new Chroma partner package, providing a dedicated integration path for the Chroma vector store.
• ›Adds IDs to tool calls in the MistralAI integration, bringing it in line with the tool-call message standard.

LangChain v0.1.15 adds Mermaid graph rendering, Groq tool calling, Anthropic tool use, async document loaders, and a new Postgres chat history package.

$ git clone --branch v0.1.15 https://github.com/langchain-ai/langchain.git
# already have the repo? check out this version:
$ git checkout v0.1.15

png_bytes = chain.get_graph().draw_mermaid_png()
with open('graph.png', 'wb') as f:
    f.write(png_bytes)

from langchain_community.document_loaders import TextLoader

loader = TextLoader('data.txt')
docs = await loader.aload()

from langchain_groq import ChatGroq
from langchain_core.tools import tool

@tool
def get_weather(city: str) -> str:
    'Get the weather for a city.'
    return f'Sunny in {city}'

llm = ChatGroq(model='llama3-70b-8192')
llm_with_tools = llm.bind_tools([get_weather])
for chunk in llm_with_tools.stream('What is the weather in Paris?'):
    print(chunk)

• ›Adds aload method to document loaders in langchain-core for async document loading.
• ›Adds aformat method to FewShotPromptTemplate for async prompt formatting.
• ›Adds aformat_messages to ChatMessagePromptTemplate for async message formatting.
• ›Adds aformat_prompt and ainvoke to BasePromptTemplate for async prompt formatting and invocation.
• ›Adds aformat_document async method to core document formatting utilities.

LangChain v0.1.14 adds DuckDB vector store, AI21 semantic text splitter, GigaChat embeddings, async memory support, and Cohere as a partner package.

$ git clone --branch v0.1.14 https://github.com/langchain-ai/langchain.git
# already have the repo? check out this version:
$ git checkout v0.1.14

from langchain_community.document_loaders import RecursiveUrlLoader

loader = RecursiveUrlLoader(
    url="https://docs.example.com/api",
    base_url="https://docs.example.com/api"
)
docs = loader.load()

from langchain_community.vectorstores import DuckDB
from langchain_openai import OpenAIEmbeddings

vectorstore = DuckDB.from_documents(
    documents=docs,
    embedding=OpenAIEmbeddings()
)
results = vectorstore.similarity_search("threat actor lateral movement", k=4)

• ›Adds base_url option to RecursiveUrlLoader to control crawl scope.
• ›Adds mode and post_processors arguments to S3FileLoader, exposing unstructured loader options.
• ›Adds DuckDB as a vector store via langchain-community.
• ›Adds langchain_cohere as a new partner package with Cohere chat/embedding support.
• ›Adds AI21 Labs Semantic Text Splitter as a partner integration.

Letta 0.3.11 adds CLI streaming support for OpenAI and OpenAI-compatible endpoints via memgpt run --stream

$ git clone --branch 0.3.11 https://github.com/letta-ai/letta.git
# already have the repo? check out this version:
$ git checkout 0.3.11

$ memgpt run --stream

• ›Adds --stream flag to memgpt run to enable streaming output in the CLI when using OpenAI or OpenAI-compatible (proxy) endpoints.

Letta 0.3.10 adds support for Anthropic Claude, Cohere Command-R+, and Groq LLM APIs.

$ git clone --branch 0.3.10 https://github.com/letta-ai/letta.git
# already have the repo? check out this version:
$ git checkout 0.3.10

• ›Adds Anthropic Claude API support as a new LLM backend.
• ›Adds Cohere API support, including the Command-R+ model.

Letta 0.3.9 adds Google AI Gemini Pro as an LLM provider, REST API tool creation, a dev portal, and Python 3.12 support.

$ git clone --branch 0.3.9 https://github.com/letta-ai/letta.git
# already have the repo? check out this version:
$ git checkout 0.3.9

$ memgpt configure
# When prompted:
# Select LLM inference provider: google_ai
# Enter your Google AI (Gemini) API key: <your-api-key>
# Enter your Google AI (Gemini) service endpoint: generativelanguage
# Select default model: gemini-pro

• ›Adds google_ai as a selectable LLM inference provider in memgpt configure, with support for the gemini-pro model (30720-token context window) via the generativelanguage service endpoint.
• ›Adds REST API support for tool creation, enabling programmatic management of agent tools.
• ›Adds a dev portal for local development and inspection.
• ›Adds Python 3.12 compatibility.

Letta 0.3.8 adds Docker Compose server support, Groq integration, and richer source metadata.

$ git clone --branch 0.3.8 https://github.com/letta-ai/letta.git
# already have the repo? check out this version:
$ git checkout 0.3.8

$ docker compose up

• ›Supports spinning up a MemGPT server with a PostgreSQL database via docker compose up using compose.yaml.
• ›Adds Groq as a supported LLM provider via the local option with authentication.
• ›Returns num_passages in Source.metadata_ from the REST list sources endpoint.
• ›Adds a description field to Source objects.
• ›Moves quickstart configuration to use inference.memgpt.ai as the default inference endpoint.

LlamaIndex v0.10.30 adds LATS agent pack, two new embedding integrations, OR filter support, and intermediate QueryPipeline outputs.

$ git clone --branch v0.10.30 https://github.com/run-llama/llama_index.git
# already have the repo? check out this version:
$ git checkout v0.10.30

from llama_index.core.vector_stores.types import MetadataFilters, MetadataFilter, FilterCondition

filters = MetadataFilters(
    filters=[
        MetadataFilter(key="source", value="arxiv"),
        MetadataFilter(key="source", value="pubmed"),
    ],
    condition=FilterCondition.OR,
)
results = index.as_retriever(filters=filters).retrieve("transformer models")

from azure.identity import DefaultAzureCredential, get_bearer_token_provider
from llama_index.embeddings.azure_openai import AzureOpenAIEmbedding

token_provider = get_bearer_token_provider(
    DefaultAzureCredential(), "https://cognitiveservices.azure.com/.default"
)
embed_model = AzureOpenAIEmbedding(
    model="text-embedding-ada-002",
    deployment_name="my-deployment",
    azure_endpoint="https://<your-resource>.openai.azure.com/",
    azure_ad_token_provider=token_provider,
)

• ›Adds OR filter condition support to the simple vector store, enabling more flexible metadata filtering alongside existing AND conditions.
• ›Exposes azure_ad_token_provider argument in both llama-index-embeddings-azure-openai and llama-index-llms-azure-openai to support token expiration/refresh scenarios.
• ›Adds httpx_async_client option to llama-index-embeddings-cohere for async HTTP client customization.
• ›New llama-index-embeddings-ipex-llm integration (v0.1.0) adds embedding support via Intel IPEX-LLM.
• ›New llama-index-embeddings-octoai integration (v0.1.0) adds embedding support via OctoAI.

LlamaIndex v0.10.29 adds OpenVINO LLMs and reranking, Couchbase and Bedrock vector/retrieval integrations, Chain-of-Abstraction agent pack, and Mistral Large on Bedrock.

$ git clone --branch v0.10.29 https://github.com/run-llama/llama_index.git
# already have the repo? check out this version:
$ git checkout v0.10.29

• ›Adds llama-index-llms-openvino (0.1.0) — new OpenVino LLM integration installable via pip install llama-index-llms-openvino.
• ›Adds llama-index-postprocessor-openvino-rerank OpenVINO reranking postprocessor support.
• ›Adds llama-index-retrievers-bedrock (0.1.0) — Amazon Bedrock knowledge base integration as a retriever.
• ›Adds llama-index-retrievers-mongodb-atlas-bm25-retriever (0.1.3) — MongoDB Atlas BM25 retriever.
• ›Adds llama-index-vector-stores-couchbase (0.1.0) — Couchbase as a vector store.

LlamaIndex v0.10.28 adds Anthropic tool calling, OpenVINO embeddings, ipex-llm integration, and multilingual Wikipedia support.

$ git clone --branch v0.10.28 https://github.com/run-llama/llama_index.git
# already have the repo? check out this version:
$ git checkout v0.10.28

from llama_index.core.tools import FunctionTool

def lookup_price(ticker: str) -> str:
    return f"${ticker}: 142.00"

price_tool = FunctionTool.from_defaults(
    fn=lookup_price,
    return_direct=True,
)

from llama_index.readers.wikipedia import WikipediaReader

reader = WikipediaReader()
docs = reader.load_data(pages=["Louvre"], lang="fr")

• ›Adds return_direct option to tool metadata in llama-index-core, letting tools short-circuit the agent loop and return their output directly to the caller.
• ›Adds async_postprocess_nodes to the RankGPT postprocessor in llama-index-core, enabling fully async reranking pipelines.
• ›Adds thread-safe and coroutine-safe instrumentation spans in llama-index-core, making telemetry safe for concurrent and async workloads.
• ›Adds in-memory loading for non-default filesystems in PDFReader (llama-index-core), enabling PDF ingestion from remote or custom storage backends.
• ›Adds SynthesizeComponent to shortcut imports in llama-index-core.

LlamaIndex v0.10.27 adds Databricks, Cloudflare Workers AI, and Neptune Analytics integrations alongside Cohere Command R+ and RankGPT support.

$ git clone --branch v0.10.27 https://github.com/run-llama/llama_index.git
# already have the repo? check out this version:
$ git checkout v0.10.27

• ›Adds span_id attribute to Events in the instrumentation layer (llama-index-core).
• ›Adds node-postprocessors support to retriever_tool (llama-index-core).
• ›Adds FLAREInstructQueryEngine delegation to the retriever API when the query engine supports it (llama-index-core).
• ›New llama-index-llms-databricks [0.1.0] integration with the Databricks LLM API.
• ›New llama-index-embeddings-cloudflar-workersai [0.1.0] text embedding integration with Cloudflare Workers AI.

AutoGen v0.2.26 adds PGVector support for RAG, selective carryover in initiate_chats, and sk-proj- OpenAI API key format.

$ git clone --branch v0.2.26 https://github.com/microsoft/autogen.git
# already have the repo? check out this version:
$ git checkout v0.2.26

• ›Adds vector_db as a settable parameter in retrieval-augmented chat contrib, enabling customizable vector database backends including PGVector.
• ›Enhances initiate_chats to support selective carryover of context between chats.
• ›Supports OpenAI sk-proj- API key format.
• ›New integration example with promptflow in samples/apps/promptflow-autogen.

AutoGen v0.2.25 adds Gemini model support and custom Bing Search base URL for the browser agent.

$ git clone --branch v0.2.25 https://github.com/microsoft/autogen.git
# already have the repo? check out this version:
$ git checkout v0.2.25

• ›Adds support for a custom base URL for Bing Search in the browser agent, enabling use of proxy or regional endpoints.
• ›Adds Google Gemini as a supported model provider for AutoGen agents.

AutoGen v0.2.24 adds Anthropic Claude function calling, a customizable vectordb module, and CosmosDB support.

$ git clone --branch v0.2.24 https://github.com/microsoft/autogen.git
# already have the repo? check out this version:
$ git checkout v0.2.24

$ pip install pyautogen[cosmosdb]

• ›Adds extra_require for cosmosdb in setup.py, enabling optional CosmosDB installation as a vector store backend.
• ›Adds a vectordb module with a customizable vector database interface for RAG pipelines.
• ›Adds function call support for Anthropic Claude via the latest Anthropic API.
• ›Adds llm_config support in AgentOptimizer, allowing LLM configuration to be passed directly to the optimizer.
• ›Adds 'py' as a recognized language tag in ConversableAgent code execution, enabling Python code blocks to be detected and run.

AutoGen v0.2.22 adds TransformMessages capability, Anthropic Claude support, GroupChat speaker customization, and an in-memory cache class.

$ git clone --branch v0.2.22 https://github.com/microsoft/autogen.git
# already have the repo? check out this version:
$ git checkout v0.2.22

retrieve_user_proxy = RetrieveUserProxyAgent(
    name="retrieve_proxy",
    retrieve_config={
        "docs_path": "./docs",
        "context_max_tokens": 2000,
    }
)

• ›Adds TransformMessages capability as a generalized replacement for previous long-context handling — prior long-context capabilities are now deprecated.
• ›Adds support for Anthropic Claude models, including system message support in Claude-based workflows.
• ›Adds an in-memory cache class (Add in memory cache class) for LLM response caching without disk I/O.
• ›Adds context_max_tokens support in RetrieveUserProxyAgent via retrieve_config, giving fine-grained control over retrieval context size.
• ›Adds ability to specify the role field for select-speaker messages in GroupChat, enabling Mistral and other non-OpenAI models to function correctly in group chat speaker selection.

Semantic Kernel .NET 1.9.0 adds OpenAI Assistant Agent support, XML tag chat prompts, and Google connector API version selection.

$ git clone --branch dotnet-1.9.0 https://github.com/microsoft/semantic-kernel.git
# already have the repo? check out this version:
$ git checkout dotnet-1.9.0

• ›Adds Google connector API version selection, enabling callers to target a specific Google API version from the connector configuration.
• ›Introduces the OpenAI Assistant Agent, adding a new agent type backed by the OpenAI Assistants API.
• ›Supports XML tags in chat prompts, allowing prompt templates to use XML-style tag syntax alongside existing formats.

Semantic Kernel 1.8.0 adds AgentGroupChat, function call content types, HuggingFace TGI chat, and custom OpenAI-compatible endpoints.

$ git clone --branch dotnet-1.8.0 https://github.com/microsoft/semantic-kernel.git
# already have the repo? check out this version:
$ git checkout dotnet-1.8.0

• ›Introduces AgentGroupChat to the Agent Framework, enabling multi-agent group chat orchestration.
• ›Adds function call content model classes (FunctionCallContent and related types) for structured handling of LLM function call payloads.
• ›Makes OpenAPI operation metadata and extension metadata available at function invocation time.
• ›Supports custom OpenAI-compatible chat message API endpoints via the OpenAI connector.
• ›Adds HuggingFace TGI (Text Generation Inference) Chat Completion Message API support.

Semantic Kernel python-0.9.6b1 redesigns plugin/function registration with new kernel methods and modular import paths.

$ git clone --branch python-0.9.6b1 https://github.com/microsoft/semantic-kernel.git
# already have the repo? check out this version:
$ git checkout python-0.9.6b1

from semantic_kernel import Kernel
from semantic_kernel.functions import kernel_function

class MyPlugin:
    @kernel_function(name="greet", description="Greet a user")
    def greet(self, name: str) -> str:
        return f"Hello, {name}!"

kernel = Kernel()
kernel.add_plugin(MyPlugin(), plugin_name="MyPlugin")

• ›Adds kernel.add_plugin() and kernel.add_plugins() for registering plugins directly as a KernelPlugin instance, as a custom class with @kernel_function-decorated methods, or as a decorated dictionary.
• ›Adds kernel.add_function() and kernel.add_functions() for registering individual functions with the kernel.
• ›Adds kernel.add_plugin_from_openapi() to load an OpenAPI plugin into the kernel.
• ›Adds kernel.add_plugin_from_openai() to load an OpenAI plugin into the kernel.
• ›Restructures imports for faster load performance: only the Kernel is exposed at the root; all other components live in sub-packages (e.g., semantic_kernel.functions), with OpenAI and Azure OpenAI accessed via from semantic_kernel.connectors.ai.open_ai import ....

Semantic Kernel 1.7.1 adds optional chat history resumption in the stepwise planner and custom Bing Search endpoint support.

$ git clone --branch dotnet-1.7.1 https://github.com/microsoft/semantic-kernel.git
# already have the repo? check out this version:
$ git checkout dotnet-1.7.1

• ›Adds optional chatHistory parameter to the stepwise planner, enabling execution to be resumed mid-flight from a prior conversation state.
• ›Supports custom Bing Search endpoints alongside improved response formatting for Bing Search results.

Semantic Kernel 1.7.0 adds Gemini connector, BERT ONNX embeddings, OpenAI TokenCredentials, Azure file-service endpoint, and CJK text-splitter support.

$ git clone --branch dotnet-1.7.0 https://github.com/microsoft/semantic-kernel.git
# already have the repo? check out this version:
$ git checkout dotnet-1.7.0

var plugin = await kernel.ImportPluginFromApiManifestAsync(
    pluginName: "MyMultiApiPlugin",
    filePath: "./plugins/myPlugin/apimanifest.json",
    new ApiManifestPluginParameters());

• ›Introduces ApiManifestPluginParameters to support multiple API dependencies when loading API Manifest plugins.
• ›Adds Name property to ChatMessageContent for identifying message authors in multi-agent chat scenarios.
• ›Publishes Microsoft.SemanticKernel.Plugins.OpenApi.Extensions as a standalone NuGet package for OpenAPI plugin extensibility.
• ›Adds BERT ONNX embedding generation service, enabling local on-device embedding without a cloud API.
• ›Adds experimental Gemini connector, bringing Google Gemini models into the SK connector ecosystem.

Semantic Kernel Python 0.9.5b1 adds OpenAI/OpenAPI plugin operations with auth, AzureOpenAI stepwise planner support, and enhanced chat message content handling.

$ git clone --branch python-0.9.5b1 https://github.com/microsoft/semantic-kernel.git
# already have the repo? check out this version:
$ git checkout python-0.9.5b1

• ›Enables the function calling stepwise planner to use AzureOpenAI chat service as a backend.
• ›Introduces operations to handle OpenAI plugins, improves OpenAPI plugin support, and allows authentication for plugin calls.
• ›Adds a messages custom function helper for Handlebars templates, and removes Jinja2 built-in helpers from the custom helpers namespace.
• ›Honors configured function calling options when executing kernel functions.
• ›Enhances ChatMessageContent creation and parsing with richer structured support.

Jan v0.4.11 adds a Mistral inference engine extension, API prefix settings, and moves logs into a monitoring extension.

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

• ›Adds API prefix settings, letting users configure a custom URL prefix for the local API server.
• ›Adds a Mistral inference engine extension for remote Mistral model inference.
• ›Moves log output into the monitoring extension for centralized observability.
• ›Adds GPU driver and toolkit status display in the UI.
• ›Adds an 'open log directory' shortcut to the troubleshooting modal.

Jan v0.4.10 adds a Groq Inference Extension for cloud-accelerated inference alongside homepage and download page redesigns.

$ git clone --branch v0.4.10 https://github.com/janhq/jan.git
# already have the repo? check out this version:
$ git checkout v0.4.10

• ›Adds Groq Inference Extension, enabling Groq-hosted cloud inference as a backend alongside local engines.
• ›Revamps the Jan homepage with a new design.
• ›Adds a dedicated download page to the Jan website.

KoboldCpp v1.62.2 adds Img2Img via /sdapi/v1/img2img, a --chatcompletionsadapter flag, and an embedded Horde Worker priority system.

$ git clone --branch v1.62.2 https://github.com/LostRuins/koboldcpp.git
# already have the repo? check out this version:
$ git checkout v1.62.2

$ koboldcpp.exe --model my_model.gguf --chatcompletionsadapter my_adapter.json

• ›Adds --chatcompletionsadapter CLI flag to specify OpenAI Chat Completions adapter files at load time, enabling any instruct tag format via the Chat Completions API.
• ›Emulates the A1111-compatible /sdapi/v1/img2img endpoint, enabling image-to-image generation directly from KoboldCpp.
• ›Expands /api/extra/perf/ with additional usage statistics.
• ›Adds /docs endpoint as an alias for /api, serving built-in API documentation.
• ›Embedded Horde Workers now prioritise the local user, automatically pausing and resuming when a local client is active so local and horde workloads can coexist.

LocalAI v2.12.3 adds Assistant API, OpenVINO runtime, Swagger UI, Vision support for AutoGPTQ, and an Intel AIO profile.

$ git clone --branch v2.12.3 https://github.com/mudler/LocalAI.git
# already have the repo? check out this version:
$ git checkout v2.12.3

• ›Adds Assistant and AssistantFiles API endpoints, enabling OpenAI-compatible assistant workflows.
• ›Adds Swagger UI at the LocalAI front page for interactively exploring and testing API calls directly in the browser.
• ›Adds OpenVINO runtime support for the transformer backend, enabling hardware-accelerated inference on Intel hardware.
• ›Adds token streaming support for the transformer backend (including OpenVINO and CUDA paths).
• ›Adds Intel GPU profile for AIO (all-in-one) images, enabling out-of-the-box Intel GPU acceleration.

LocalAI v2.12.1 adds Assistant API, OpenVINO runtime, Swagger UI, vision support for AutoGPTQ, and an Intel AIO profile.

$ git clone --branch v2.12.1 https://github.com/mudler/LocalAI.git
# already have the repo? check out this version:
$ git checkout v2.12.1

• ›Adds Assistant and AssistantFiles API endpoints, enabling OpenAI-compatible assistant workflows.
• ›Adds Swagger UI at the LocalAI front page for interactively exploring and testing API calls directly in the browser.
• ›Adds a landing welcome page served at the LocalAI front page root.
• ›Adds OpenVINO runtime support for the transformer backend, including token streaming for both OpenVINO and CUDA.
• ›Adds token streaming support for the transformer backend.

LocalAI v2.12.0 adds Assistant API, OpenVINO runtime, Swagger UI, Vision for AutoGPTQ, and an Intel AIO GPU profile.

$ git clone --branch v2.12.0 https://github.com/mudler/LocalAI.git
# already have the repo? check out this version:
$ git checkout v2.12.0

• ›Adds Assistant and AssistantFiles API endpoints, enabling assistant-style workflows against a local inference server.
• ›Adds Swagger UI for interactively exploring and testing LocalAI API calls directly from the browser.
• ›Adds OpenVINO runtime support for the transformer backend, with token streaming for both OpenVINO and CUDA.
• ›Adds Vision (VL model) support to the AutoGPTQ backend.
• ›Adds an Intel GPU profile for AIO (all-in-one) images.

Adds min_p sampling preset as default, Ascend NPU support, and HF_ENDPOINT awareness for model downloads.

$ git clone --branch snapshot-2024-04-14 https://github.com/oobabooga/textgen.git
# already have the repo? check out this version:
$ git checkout snapshot-2024-04-14

• ›Respects model and LoRA directory settings when downloading files, so downloads land in the configured paths rather than defaults.
• ›Reads the HF_ENDPOINT environment variable when downloading models, enabling use of Hugging Face mirror endpoints.
• ›Adds Ascend NPU as a supported hardware backend.
• ›Adds a min_p sampling preset and makes it the default generation preset.

vLLM v0.4.0.post1 adds Intel CPU inference backend, lookahead scheduling for speculative decoding, and new AutoGPTQ checkpoint format support.

$ git clone --branch v0.4.0.post1 https://github.com/vllm-project/vllm.git
# already have the repo? check out this version:
$ git checkout v0.4.0.post1

• ›Adds Intel CPU inference backend, enabling vLLM inference on CPU without a GPU.
• ›Adds lookahead scheduling for speculative decoding, improving throughput for speculative decode workloads.
• ›Adds support for the new AutoGPTQ checkpoint_format, broadening quantized model compatibility.
• ›Restores sm70/sm75 (Volta/Turing GPU) binary support dropped in v0.4.0.

Ollama v0.1.32 adds five new models including WizardLM 2 and Mixtral 8x22B, plus smarter GPU/CPU model splitting on macOS.

$ git clone --branch v0.1.32 https://github.com/ollama/ollama.git
# already have the repo? check out this version:
$ git checkout v0.1.32

$ ollama run wizardlm2:7b

$ ollama pull snowflake-arctic-embed

• ›Adds wizardlm2:8x22b and wizardlm2:7b (WizardLM 2 from Microsoft AI) with improved performance on complex chat, multilingual, reasoning, and agent use cases.
• ›Adds snowflake-arctic-embed, a suite of text embedding models by Snowflake optimized for retrieval performance.
• ›Adds command-r-plus, a large language model purpose-built for RAG use cases.
• ›Adds dbrx, a 132B open general-purpose LLM created by Databricks.
• ›Adds mixtral:8x22b, Mistral AI's new Mixture of Experts base model.

Ollama v0.1.31 adds embedding model support with REST API, Python, and JavaScript library access for RAG workflows.

$ git clone --branch v0.1.31 https://github.com/ollama/ollama.git
# already have the repo? check out this version:
$ git checkout v0.1.31

• ›Adds embedding model support via the REST API, Python library (ollama-python), and JavaScript library (ollama-js), enabling retrieval-augmented generation (RAG) applications.
• ›Adds Qwen 1.5 32B (qwen:32b), a multilingual model competitive with larger models.
• ›Adds StarlingLM Beta (starling-lm:beta), a 7B model with Apache 2.0 license.
• ›Adds DolphinCoder StarCoder 7B (dolphincoder:7b), an uncensored coding-focused variant based on StarCoder2.
• ›Adds StableLM 1.6 Chat (stablelm2:chat), an instruction-tuned version of StableLM 1.6.

Phoenix Evals 0.8.0 adds a user frustration evaluator for detecting negative user sentiment in LLM conversations.

$ git clone --branch arize-phoenix-evals-v0.8.0 https://github.com/Arize-ai/phoenix.git
# already have the repo? check out this version:
$ git checkout arize-phoenix-evals-v0.8.0

• ›Adds a user frustration eval to detect signs of user frustration in LLM application traces.

Arize Phoenix v3.23.0 adds support for default_headers in Azure OpenAI integration.

$ git clone --branch arize-phoenix-v3.23.0 https://github.com/Arize-ai/phoenix.git
# already have the repo? check out this version:
$ git checkout arize-phoenix-v3.23.0

• ›Adds default_headers parameter support for the azure_openai integration, enabling custom HTTP headers to be sent with Azure OpenAI requests.

Phoenix v3.22.0 adds log_traces method to send TraceDataset traces directly to Phoenix.

$ git clone --branch arize-phoenix-v3.22.0 https://github.com/Arize-ai/phoenix.git
# already have the repo? check out this version:
$ git checkout arize-phoenix-v3.22.0

import phoenix as px

dataset = px.TraceDataset(dataframe)
px.log_traces(trace_dataset=dataset)

• ›Adds log_traces method that sends a TraceDataset of traces to Phoenix programmatically.

Phoenix Evals v0.7.0 adds SQL and code functionality evaluation templates.

$ git clone --branch arize-phoenix-evals-v0.7.0 https://github.com/Arize-ai/phoenix.git
# already have the repo? check out this version:
$ git checkout arize-phoenix-evals-v0.7.0

• ›Adds SQL and Code Functionality Eval Templates for evaluating SQL queries and code outputs.

Arize Phoenix v3.21.0 adds SQL and Code functionality eval templates for LLM evaluation.

$ git clone --branch arize-phoenix-v3.21.0 https://github.com/Arize-ai/phoenix.git
# already have the repo? check out this version:
$ git checkout arize-phoenix-v3.21.0

• ›Adds SQL and Code functionality eval templates for evaluating LLM outputs against structured query and code generation tasks.

Langfuse v2.36.0 adds ENABLE_EVENT_LOG environment variable to control event logging.

$ git clone --branch v2.36.0 https://github.com/langfuse/langfuse.git
# already have the repo? check out this version:
$ git checkout v2.36.0

$ ENABLE_EVENT_LOG=true

• ›Adds ENABLE_EVENT_LOG environment variable to control whether event logging is active.

Langfuse v2.34.0 expands eval observability with an eval log, score table integration, config-page logs, and a log table toolbar.

$ git clone --branch v2.34.0 https://github.com/langfuse/langfuse.git
# already have the repo? check out this version:
$ git checkout v2.34.0

• ›Adds an eval log for tracking evaluation runs and their outcomes.
• ›Adds eval scores to the score table, surfacing evaluation results alongside other scores.
• ›Adds logs directly to the eval config page for in-context observability of evaluation execution.
• ›Adds a toolbar to the log table for faster navigation and filtering.
• ›Links eval templates within eval configs for easier cross-reference.

Langfuse v2.32.0 adds bring-your-own LLM API key support and managed eval prompt templates.

$ git clone --branch v2.32.0 https://github.com/langfuse/langfuse.git
# already have the repo? check out this version:
$ git checkout v2.32.0

• ›Adds 'bring your own LLM API key' capability, letting users supply their own API key for LLM-powered features.
• ›Adds Langfuse-managed eval prompt templates, providing ready-to-use templates for LLM-as-a-judge evaluations.
• ›Updates OpenAI model definitions to include newer models for cost and usage tracking.

Langfuse v2.33.0 adds an eval log and refreshes the API key design.

$ git clone --branch v2.33.0 https://github.com/langfuse/langfuse.git
# already have the repo? check out this version:
$ git checkout v2.33.0

• ›Adds an eval log to surface evaluation execution details in the UI.
• ›Improves the API key design for clearer key management.

Langfuse v2.31.0 launches a Playground feature, available under the Enterprise Edition.

$ git clone --branch v2.31.0 https://github.com/langfuse/langfuse.git
# already have the repo? check out this version:
$ git checkout v2.31.0

• ›Adds a Playground for interactively testing and iterating on LLM prompts, available in the Enterprise Edition.

Langfuse v2.30.1 adds countObservations and nullable model to the metrics API endpoint.

$ git clone --branch v2.30.1 https://github.com/langfuse/langfuse.git
# already have the repo? check out this version:
$ git checkout v2.30.1

• ›Adds countObservations field and makes model nullable in the metrics API endpoint.

Langfuse v2.30.0 adds concise JSON field rendering in tables and improved sign-in error visibility.

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

• ›Renders input, output, metadata, and other JSON fields concisely in tables at small row height for easier scanning.
• ›Displays errors on the sign-in page and logs unknown errors to Sentry for faster diagnosis.

Langfuse v2.29.0 adds metadata to datasets and dataset items, plus a score value filter on the scores API.

$ git clone --branch v2.29.0 https://github.com/langfuse/langfuse.git
# already have the repo? check out this version:
$ git checkout v2.29.0

• ›Adds a score value filter to the scores API endpoint, enabling programmatic retrieval of scores filtered by value.
• ›Adds metadata field support to datasets and dataset items, allowing structured metadata to be attached to both objects.

Langfuse v2.28.0 adds timestamp filtering to the daily metrics API and moves evals to its own main menu entry.

$ git clone --branch v2.28.0 https://github.com/langfuse/langfuse.git
# already have the repo? check out this version:
$ git checkout v2.28.0

$ curl -X GET 'https://<your-langfuse-host>/api/public/metrics/daily?fromTimestamp=2024-01-01T00:00:00Z&toTimestamp=2024-01-07T23:59:59Z' \
  -H 'Authorization: Basic <base64-encoded-credentials>'

• ›Adds fromTimestamp and toTimestamp query parameters to the GET /metrics/daily API endpoint, enabling scoped time-range queries against daily metrics.
• ›Moves evals to a dedicated top-level main menu entry in the UI, surfacing them as a first-class navigation destination.

Langfuse v2.27.0 adds domain-specific SSO provider support for enterprise deployments.

$ git clone --branch v2.27.0 https://github.com/langfuse/langfuse.git
# already have the repo? check out this version:
$ git checkout v2.27.0

• ›Adds domain-specific SSO providers (enterprise edition), enabling different identity providers to be mapped per email domain.

Langfuse v2.25.0 removes row limits on trace/generation input-output views in session displays.

$ git clone --branch v2.25.0 https://github.com/langfuse/langfuse.git
# already have the repo? check out this version:
$ git checkout v2.25.0

• ›Removes the previous cap on the number of trace and generation rows visible when viewing input/output data in a session.

Langfuse Docker images now available on Docker Hub in addition to GitHub Container Registry.

$ git clone --branch v2.24.1 https://github.com/langfuse/langfuse.git
# already have the repo? check out this version:
$ git checkout v2.24.1

• ›Langfuse Docker images are now published to Docker Hub (langfuse/langfuse) in addition to the existing GitHub Container Registry distribution.

Langfuse v2.24.0 adds a UI for eval config creation and OpenAI GPT-4 Turbo model support.

$ git clone --branch v2.24.0 https://github.com/langfuse/langfuse.git
# already have the repo? check out this version:
$ git checkout v2.24.0

• ›Adds a UI for creating evaluation configurations, enabling teams to set up eval configs without direct API or config-file access.
• ›Adds OpenAI GPT-4 Turbo as a supported model for evaluations.
• ›Inactive eval configs are now skipped during processing, reducing unnecessary worker load.
• ›Adds graceful shutdown handling for the evaluation worker.
• ›Adds security headers to the evaluation worker service.

Langfuse v2.23.0 adds adjustable table row heights and visual output cell highlighting in the UI.

$ git clone --branch v2.23.0 https://github.com/langfuse/langfuse.git
# already have the repo? check out this version:
$ git checkout v2.23.0

• ›Adds adjustable row height on tables with improved rendering of JSON and data cells.
• ›Applies a light-green background to output table cells for quicker visual distinction between input and output data.

Adds AUTH_DISABLE_SIGNUP environment variable to block all new user registrations.

$ git clone --branch v2.21.2 https://github.com/langfuse/langfuse.git
# already have the repo? check out this version:
$ git checkout v2.21.2

$ AUTH_DISABLE_SIGNUP=true

• ›Adds AUTH_DISABLE_SIGNUP environment variable to disable all new signups, enabling operators to lock down self-hosted instances to existing users only.

Langfuse v2.21.1 adds up/down navigation on traces and dataset items when viewed from dataset run items.

$ git clone --branch v2.21.1 https://github.com/langfuse/langfuse.git
# already have the repo? check out this version:
$ git checkout v2.21.1

• ›Enables up/down keyboard navigation on traces and dataset items when accessed from dataset run items.

Langfuse v2.21.0 adds a failIfNoEventsInLastMinute health-check flag, dataset I/O visibility, source trace links, and a CodeMirror JSON editor across datasets, models, and prompts.

$ git clone --branch v2.21.0 https://github.com/langfuse/langfuse.git
# already have the repo? check out this version:
$ git checkout v2.21.0

$ curl -f 'https://<your-langfuse-host>/health?failIfNoEventsInLastMinute=true'

• ›Adds failIfNoEventsInLastMinute flag to the /health API endpoint, enabling liveness checks that fail when no events have been ingested in the last minute.
• ›Introduces column visibility controls on all dataset tables in the UI.
• ›Surfaces input/output and expected output fields directly on dataset run items in the UI.
• ›Adds links from dataset run items back to their source trace or observation in the UI.
• ›Replaces the JSON editor across datasets, model tokenizer config, and prompt config with a CodeMirror-based editor (react-codemirror).

Langfuse v2.20.0 adds prompt tags with filtering/sorting, dashboard tag filters, and richer dataset API references.

$ git clone --branch v2.20.0 https://github.com/langfuse/langfuse.git
# already have the repo? check out this version:
$ git checkout v2.20.0

• ›Adds runName and datasetName reference fields to all dataset API endpoints, enabling richer cross-referencing in dataset responses.
• ›Adds tags to prompts with support for filtering, sorting, and pagination in the prompt management UI.
• ›Adds tag filtering to the dashboard for narrowing observability views by tag.

GET /observations now exposes promptId in API responses.

$ git clone --branch v2.19.0 https://github.com/langfuse/langfuse.git
# already have the repo? check out this version:
$ git checkout v2.19.0

$ curl -X GET 'https://<your-langfuse-host>/api/public/observations' \
  -H 'Authorization: Bearer <secret-key>'

• ›Exposes promptId field via the GET /observations API endpoint, enabling callers to trace which prompt produced a given observation.

Langfuse v2.18.1 adds a worker health check endpoint.

$ git clone --branch v2.18.1 https://github.com/langfuse/langfuse.git
# already have the repo? check out this version:
$ git checkout v2.18.1

• ›Adds a worker health check to monitor the status of background workers.

Langfuse v2.18.0 adds a basic evaluation function capability.

$ git clone --branch v2.18.0 https://github.com/langfuse/langfuse.git
# already have the repo? check out this version:
$ git checkout v2.18.0

• ›Adds an evaluation basic function for assessing LLM outputs within the platform.

Langfuse v2.17.0 adds chat prompt support to the prompt management system.

$ git clone --branch v2.17.0 https://github.com/langfuse/langfuse.git
# already have the repo? check out this version:
$ git checkout v2.17.0

• ›Adds chat prompts, enabling prompt management for chat-style (multi-turn) message formats alongside existing text prompts.

Langfuse v2.16.0 adds dataset listing, trace-to-dataset-item creation, and descriptions for datasets and runs via API.

$ git clone --branch v2.16.0 https://github.com/langfuse/langfuse.git
# already have the repo? check out this version:
$ git checkout v2.16.0

$ curl -X GET 'https://cloud.langfuse.com/api/public/datasets' \
  -H 'Authorization: Bearer <secret_key>'

• ›Adds GET /datasets endpoint to list all datasets.
• ›Adds description field to datasets and runs; adds datasetName field to GET runs/[runName] response.
• ›Enables creating a dataset item directly from an existing trace via the datasets API.

Dataset item inputs can now be nullable in Langfuse v2.15.5

$ git clone --branch v2.15.5 https://github.com/langfuse/langfuse.git
# already have the repo? check out this version:
$ git checkout v2.15.5

• ›Allows dataset item input fields to be nullable, enabling dataset entries with no input value.

Langfuse v2.15.4 lets dataset run items link to full traces, not just observations.

$ git clone --branch v2.15.4 https://github.com/langfuse/langfuse.git
# already have the repo? check out this version:
$ git checkout v2.15.4

• ›Dataset run items can now be connected to traces in addition to observations, enabling full-trace context when reviewing dataset runs.

Langfuse v2.15.3 adds Auth0 as a supported authentication provider for self-hosted deployments.

$ git clone --branch v2.15.3 https://github.com/langfuse/langfuse.git
# already have the repo? check out this version:
$ git checkout v2.15.3

• ›Adds Auth0 as a supported authentication provider, configurable via self-host environment variables.

Langfuse v2.15.2 adds Okta as a supported authentication provider for self-hosted deployments.

$ git clone --branch v2.15.2 https://github.com/langfuse/langfuse.git
# already have the repo? check out this version:
$ git checkout v2.15.2

• ›Adds Okta as a supported SSO/auth provider, configured via environment variables in self-hosted deployments (see configuring-environment-variables in the self-host docs).

LanceDB v0.6.11 adds table renaming, richer index stats, and configurable index cache size on table open.

$ git clone --branch python-v0.6.11 https://github.com/lancedb/lancedb.git
# already have the repo? check out this version:
$ git checkout python-v0.6.11

table = db.open_table("my_vectors", index_cache_size=512)

db.rename_table("old_name", "new_name")

• ›Adds index_cache_size configuration option when opening a table, enabling tuning of in-memory index cache allocation.
• ›Adds rename_table function to rename tables in a LanceDB database.
• ›Expands data returned by index_stats to surface more index metadata.

LanceDB v0.6.8 adds storage_options for passing auth and config to object stores.

$ git clone --branch python-v0.6.8 https://github.com/lancedb/lancedb.git
# already have the repo? check out this version:
$ git checkout python-v0.6.8

• ›Adds storage_options argument to pass authentication and other configurations down to object stores.

LanceDB v0.4.17 exposes storage_options for passing auth and config to object stores.

$ git clone --branch v0.4.17 https://github.com/lancedb/lancedb.git
# already have the repo? check out this version:
$ git checkout v0.4.17

• ›Adds storage_options argument to pass authentication and other configuration directly to object stores.

LanceDB v0.6.7 adds filterable count_rows on the remote API and ships fp16 kernels in Python wheels.

$ git clone --branch python-v0.6.7 https://github.com/lancedb/lancedb.git
# already have the repo? check out this version:
$ git checkout python-v0.6.7

• ›Adds filter support to count_rows on the remote API, enabling row counts scoped to a query predicate.
• ›Ships fp16 kernels directly in Python wheels, enabling half-precision vector operations without extra installation.

LanceDB v0.4.16 adds filterable count_rows to the remote API and aligns search defaults with the Python SDK.

$ git clone --branch v0.4.16 https://github.com/lancedb/lancedb.git
# already have the repo? check out this version:
$ git checkout v0.4.16

• ›Adds filter support to count_rows on the remote API, enabling row counts scoped to a query predicate.
• ›Sets a default value for search.limit in the remote API to match the Python SDK's behavior.

LanceDB Python SDK gains an async API backed by the Rust SDK, aligning Python with long-term cross-SDK feature parity.

$ git clone --branch python-v0.6.6 https://github.com/lancedb/lancedb.git
# already have the repo? check out this version:
$ git checkout python-v0.6.6

• ›Introduces an async Python API that replaces the pylance backend with the Rust SDK, enabling asynchronous database operations from Python.

Milvus 2.4.0 adds MinIO TLS, AutoIndex for scalar fields, new observability metrics, and import task improvements.

$ git clone --branch v2.4.0 https://github.com/milvus-io/milvus.git
# already have the repo? check out this version:
$ git checkout v2.4.0

• ›Adds new metrics for QueryCoord current target channel check point lag latency.
• ›Adds new db label to common metrics, plus new metrics for deleted, indexed, and loaded entity counts with collectionName and dbName labels.
• ›Supports AutoIndex for scalar fields, extending automatic index selection beyond vector fields.
• ›Supports MinIO TLS connections for encrypted object-storage communication.
• ›Import tasks now support waiting for data index completion before returning.

Milvus v2.3.13 adds TLS for MinIO, new observability metrics, entity-stats tracking, and expanded RESTful APIs for Partition and Index operations.

$ git clone --branch v2.3.13 https://github.com/milvus-io/milvus.git
# already have the repo? check out this version:
$ git checkout v2.3.13

• ›Adds TLS support for MinIO connections, enabling encrypted object-storage communication.
• ›Adds new metrics for QueryCoord current-target checkpoint lag, improving replication observability.
• ›Adds new metrics for entities statistics, enabling per-collection entity count monitoring.
• ›Expands RESTful interfaces with Partition and Index operation endpoints (see the RESTful API reference for v2.3.x).
• ›Adds validation checks for field data type legality on ingestion.