The AI Toolchain — issue -380, March 31, 2024

Aider v0.25.0 adds --openai-organization-id, chat size warnings, and .aiderignore enforcement

$ git clone --branch v0.25.0 https://github.com/Aider-AI/aider.git
# already have the repo? check out this version:
$ git checkout v0.25.0

$ aider --openai-organization-id org-xxxxxxxxxxxxxxxx

• ›Adds --openai-organization-id CLI argument to specify an OpenAI organization when invoking Aider.
• ›Enforces .aiderignore at chat-add time — vocally refuses to add matching files, preventing downstream git commit failures.
• ›Issues a warning when too much code is added to the chat, with a link to the FAQ on managing chat context.
• ›Surfaces a FAQ link when edits fail to apply, guiding users toward resolution steps.

Zed v0.127.3 adds new CLI flags, vim subword motions, git hunk revert, ZED_SELECTED_TEXT task variable, and more.

$ git clone --branch v0.127.3 https://github.com/zed-industries/zed.git
# already have the repo? check out this version:
$ git checkout v0.127.3

$ zed --new src/main.rs src/lib.rs

{
  "double_click_in_multibuffer": "open"
}

• ›Adds -n/--new and -a/--add flags to the zed CLI: --new always opens files in a new workspace; --add always adds directories to an existing workspace.
• ›Adds double_click_in_multibuffer setting: select makes multi-buffer behave like a singleton buffer; open opens the excerpt in a new tab.
• ›Adds editor::RevertSelectedHunks action (default keybinding cmd-alt-z) to revert git hunks directly from the editor.
• ›Adds ZED_SELECTED_TEXT task variable containing the contents of the current selection.
• ›Adds use_smartcase_find option for Vim mode.

Zed v0.126.2 adds local task configs, task context env vars, Vim paragraph objects, and new editor/workspace commands.

$ git clone --branch v0.126.2 https://github.com/zed-industries/zed.git
# already have the repo? check out this version:
$ git checkout v0.126.2

• ›Adds local task configurations via .zed/tasks.json, letting projects define tasks without touching global settings.
• ›Adds task contexts — additional environment variables injected by Zed at task spawn time, reflecting editor state (cursor position, file, selection, etc.).
• ›Adds menu::UseSelectedQuery command to populate the task modal query with the selected task name.
• ›Adds workspace::SaveWithoutFormatting, bound to cmd-k s, to save a file without invoking the auto formatter.
• ›Adds editor::OpenExcerptsSplit, bound to cmd-k enter, to open selected excerpts in the adjacent pane.

Zed v0.125.3 adds Terraform LSP, ESLint flat config, new Vim motions, gutter controls, and collaboration enhancements.

$ git clone --branch v0.125.3 https://github.com/zed-industries/zed.git
# already have the repo? check out this version:
$ git checkout v0.125.3

{
  "code_actions_on_format": { "source.fixAll.eslint": true },
  "lsp": {
    "eslint": {
      "settings": {
        "codeActionOnSave": { "rules": ["import/order"] }
      }
    }
  }
}

{
  "vim": {
    "use_system_clipboard": "on_yank"
  }
}

{
  "gutter": {
    "line_numbers": true,
    "code_actions": false,
    "folds": false
  }
}

• ›Adds code_actions_on_format dispatch to all language servers connected to a buffer (not just the primary), enabling "source.fixAll.eslint": true to trigger ESLint fixes on save alongside other formatters.
• ›Adds ESLint codeActionOnSave settings passthrough via the lsp config key, e.g. {"lsp": {"eslint": {"settings": {"codeActionOnSave": {"rules": ["import/order"]}}}}} to scope which rules fire on save.
• ›Adds {"vim": {"use_multiline_find": true}} setting to let f, t, F, T search across line boundaries.
• ›Adds {"vim": {"use_system_clipboard": "never" | "on_yank" | "always"}} setting to control default clipboard behavior.
• ›Adds {"gutter": {"line_numbers": true, "code_actions": true, "folds": true}} settings to control display of gutter visual elements.

CrewAI v0.19.0 adds execution metrics, input passing at kickoff, and function-calling LLM fallback for agents and crews.

$ git clone --branch v0.19.0 https://github.com/crewAIInc/crewAI.git
# already have the repo? check out this version:
$ git checkout v0.19.0

result = crew.kickoff(inputs={'domain': 'example.com'})
print(crew.usage_metrics)

agent = Agent(
    role='Analyst',
    llm=my_llm,
    function_calling_llm=my_function_llm,
    tools=[my_tool]
)

• ›Adds function_calling_llm parameter on Agent or Crew to fall back to function calling when standard tool usage fails.
• ›Adds crew.usage_metrics to retrieve execution metrics after a kickoff call, exposing tool usage statistics.
• ›Adds inputs parameter to crew.kickoff(inputs={'key': 'value'}) so runtime values can be injected at execution time.
• ›Enhances delegation capabilities for agents within a crew.

Haystack v1.25.0 adds page-based document splitting, new OpenAI embedding models, and local endpoint support via API_BASE.

$ git clone --branch v1.25.0 https://github.com/deepset-ai/haystack.git
# already have the repo? check out this version:
$ git checkout v1.25.0

preprocessor = PreProcessor(
    split_by='page',
    split_overlap=0
)
docs = preprocessor.process(raw_docs)

• ›Adds split_by='page' option to the Preprocessor so documents can be chunked by page break.
• ›Adds raise_on_failure flag to BaseConverter so large batch processes can continue past individual conversion exceptions.
• ›Adds support for OpenAI embedding models text-embedding-3-large and text-embedding-3-small.
• ›Adds API_BASE as an optional parameter to PromptNode and PromptModel, enabling RAG against any OpenAI-compatible local endpoint (e.g. LM Studio at http://localhost:1234/v1).
• ›Upgrades Transformers to 4.37.2, adding support for Phi-2 and Qwen2 models and improved quantization support.

LangChain v0.1.12 adds Anthropic tool calling, Claude v3, MongoDB LLM cache, new vector stores, and lazy_load() across 20+ document loaders.

$ git clone --branch v0.1.12 https://github.com/langchain-ai/langchain.git
# already have the repo? check out this version:
$ git checkout v0.1.12

from langchain_anthropic import ChatAnthropic
from langchain_core.tools import tool

@tool
def get_weather(location: str) -> str:
    """Return weather for a location."""
    return f"Sunny in {location}"

llm = ChatAnthropic(model="claude-3-opus-20240229")
llm_with_tools = llm.bind_tools([get_weather])
result = llm_with_tools.invoke("What is the weather in Paris?")
print(result)

from langchain_community.document_loaders import ConfluenceLoader

loader = ConfluenceLoader(
    url="https://your-org.atlassian.net/wiki",
    username="[email protected]",
    api_key="<api_key>",
    space_key="ENG"
)
for doc in loader.lazy_load():
    print(doc.metadata["title"], len(doc.page_content))

• ›Adds tool calling support to the Anthropic integration via langchain-anthropic.
• ›Adds ElasticsearchRetriever to the Elasticsearch partner package.
• ›Adds MongoDB LLM Cache to langchain-mongodb, available at the top-level library import.
• ›Adds dangerous parameter to the requests tool to require explicit opt-in for unsafe HTTP requests.
• ›Adds TritonTensorRTLLM(verbose_client=False) parameter to the nvidia-trt integration.

LangChain v0.1.11 adds Claude 3 and multimodal support, Azure Cosmos Mongo vCore caching, You.com tool, and RAPTOR retrieval.

$ git clone --branch v0.1.11 https://github.com/langchain-ai/langchain.git
# already have the repo? check out this version:
$ git checkout v0.1.11

• ›Adds ChatAnthropic support for Claude 3 models in the anthropic partner package.
• ›Adds multimodal (image input) support to ChatAnthropic.
• ›Adds a You.com tool and async support to the You.com retriever in the community package.
• ›Adds a tools renderer for non-OpenAI agents, broadening agent compatibility.
• ›Adds session-level feedback support for LangSmith evals.

LangChain v0.1.10 adds Fireworks/Mistral function calling, PNG graph rendering, SQLDatabaseLoader, LLMLingua compression, and new partner packages.

$ git clone --branch v0.1.10 https://github.com/langchain-ai/langchain.git
# already have the repo? check out this version:
$ git checkout v0.1.10

from langchain_community.document_loaders import SQLDatabaseLoader
from langchain_community.utilities import SQLDatabase

db = SQLDatabase.from_uri('postgresql://user:pass@localhost/mydb')
loader = SQLDatabaseLoader(query='SELECT id, content FROM documents', db=db)
docs = loader.load()

• ›Adds ChatFireworks.with_structured_output for structured output support in the Fireworks partner package.
• ›Adds function calling and with_structured_output to the Mistral partner package (langchain-mistral).
• ›Adds SET allow_experimental_[engine]_index as a configurable option in vectorstores.clickhouse.
• ›Adds SQLDatabaseLoader document loader to langchain_community for loading documents directly from SQL databases.
• ›Adds BaseMessage.id field to core message types, with automatic assignment in ChatOpenAI.

Letta 0.3.6 expands the REST API with archival memory, data sources, and message UUIDs in SSE streaming.

$ git clone --branch 0.3.6 https://github.com/letta-ai/letta.git
# already have the repo? check out this version:
$ git checkout 0.3.6

$ curl -X GET 'http://localhost:8283/api/agents/{agent_id}/archival' -H 'Authorization: Bearer <api_key>'

• ›Adds archival memory endpoints to the REST API: GET, POST, and DELETE under /api/agents.
• ›Adds data sources endpoints to the REST API for managing agent data sources.
• ›Adds last_run field to the agent state model, available via the REST API.
• ›Adds persona_name and human_name fields to the Preset model.
• ›Adds memory data and tool data to the GET /api/agents list response.

Letta 0.3.5 adds REST endpoints for agents, humans, personas, LLM config, and tools listing.

$ git clone --branch 0.3.5 https://github.com/letta-ai/letta.git
# already have the repo? check out this version:
$ git checkout 0.3.5

• ›Adds GET REST API route for listing tools.
• ›Adds REST API routes to GET information for agents, humans, and personas, and stores humans and personas in the database.
• ›Returns server.server_llm_config information via REST endpoint.
• ›Moves agent_id from a query parameter to a path variable in REST API routes.
• ›Adds REST API support for creating humans and personas.

Letta 0.3.4 adds RESTClient and Admin Python clients, HTTPS support for the server, and a Dockerfile for self-hosted deployments.

$ git clone --branch 0.3.4 https://github.com/letta-ai/letta.git
# already have the repo? check out this version:
$ git checkout 0.3.4

• ›Adds RESTClient and Admin Python client classes for interacting with the Letta server programmatically.
• ›Adds basic HTTPS support to memgpt server for development environments that require it.
• ›Adds a Dockerfile for running the Letta server as a container.
• ›Allows admin users to access all API routes, with authentication re-enabled.
• ›Adds authentication support to the Chat UI.

LlamaIndex v0.10.17 adds relative/dist-based fusion scoring, Anthropic multimodal models, a finance chat llama-pack, and SQL refine templates.

$ git clone --branch v0.10.17 https://github.com/run-llama/llama_index.git
# already have the repo? check out this version:
$ git checkout v0.10.17

from llama_index.core.retrievers import QueryFusionRetriever

retriever = QueryFusionRetriever(
    retrievers=[retriever_a, retriever_b],
    mode="dist_based_score",
    num_queries=4,
)
nodes = retriever.retrieve("What is the capital of France?")

• ›Adds relative_score and dist_based_score scoring modes to QueryFusionRetriever in llama-index-core.
• ›Adds support for a refine template in BaseSQLTableQueryEngine via llama-index-core.
• ›Adds support for Anthropic multimodal models haiku and sonnet in llama-index-multi-modal-llms-anthropic.
• ›Adds new llama-index-packs-finchat llama-pack for hierarchical agents combined with finance chat workflows.
• ›Inherits metadata to summaries in DocumentSummaryIndex in llama-index-core.

AutoGen v0.2.20 adds image generation, Azure AI Search support, streaming replies, and a composable actor platform.

$ git clone --branch v0.2.20 https://github.com/microsoft/autogen.git
# already have the repo? check out this version:
$ git checkout v0.2.20

{
  "model": "gpt-4",
  "api_type": "azure",
  "api_key": "<your-key>",
  "base_url": "<your-azure-endpoint>",
  "extra_body": {
    "dataSources": [
      {
        "type": "AzureCognitiveSearch",
        "parameters": {
          "endpoint": "<search-endpoint>",
          "key": "<search-key>",
          "indexName": "<index-name>"
        }
      }
    ]
  }
}

• ›Adds extra_body field to LLMConfig dataclass to enable Azure AI Search support in AutoGen Studio.
• ›New ImageGenerationCapability contrib feature (2.0) lets agents generate images as part of conversations.
• ›New Composable Actor Platform (CAP) sample app enables distributed, actor-based AutoGen agent deployments.
• ›AutoGen Studio gains upload/download of Skills and Workflows, streaming agent replies, and agent message summarization.
• ›Nested chat now supports different senders, enabling more flexible multi-agent conversation topologies.

AutoGen v0.2.18 adds callable messages, a fine-tuning tool for conversable agents, and a Docker-based command-line code executor.

$ git clone --branch v0.2.18 https://github.com/microsoft/autogen.git
# already have the repo? check out this version:
$ git checkout v0.2.18

• ›Adds LocalCommandLineCodeExecutor support for accepting a path object for work_dir, in addition to strings.
• ›Implements a Docker-based CommandLineCodeExecutor for sandboxed, containerized code execution.
• ›Supports callable messages, allowing user-defined message functions to control what agents send to one another.
• ›Adds a fine-tuning tool (samples/tools/finetuning) for training custom models on conversable agents.

AutoGen v0.2.17 adds customizable speaker selection for group chats and tightens nested chat registration.

$ git clone --branch v0.2.17 https://github.com/microsoft/autogen.git
# already have the repo? check out this version:
$ git checkout v0.2.17

• ›Allows users to pass a customized speaker selection method into group chat, enabling fully programmable agent turn-ordering beyond the built-in strategies.
• ›Removes the default trigger value for register_nested_chats, requiring callers to supply an explicit trigger and making nested chat configuration unambiguous.
• ›Raises errors when incompatible arguments are used together with a code executor, surfacing misconfiguration at startup instead of silently misbehaving.
• ›Adjusts message processing order to ensure proper combination of agent capabilities across multi-agent pipelines.

AutoGen v0.2.16 adds register_nested_chats, a Docker-based Jupyter executor, and expanded hook and function-removal APIs.

$ git clone --branch v0.2.16 https://github.com/microsoft/autogen.git
# already have the repo? check out this version:
$ git checkout v0.2.16

• ›Adds register_nested_chats method to simplify composing nested chats, letting agents use other multi-agent conversations as inner monologue before replying.
• ›Adds support for removing function calls in ConversableAgent.
• ›Hook methods updated to accept a sender argument, enabling per-sender logic in hook callbacks.
• ›Introduces a Docker-based Jupyter executor for sandboxed, container-isolated code execution.
• ›FSM-based group chat with user-specified agent transitions now documented via an official blog post.

Semantic Kernel 1.6.3 exposes Agent Thread Messages and clones KernelFunctions on plugin insertion.

$ git clone --branch dotnet-1.6.3 https://github.com/microsoft/semantic-kernel.git
# already have the repo? check out this version:
$ git checkout dotnet-1.6.3

• ›Exposes Agent Thread Messages, allowing callers to retrieve messages from an agent thread directly.
• ›Creates a clone of KernelFunction when it is added to a KernelPlugin, preventing unintended shared-state mutations across plugin registrations.
• ›Adds a HuggingFace image-to-text Windows Forms sample demonstrating the image-to-text capability.
• ›Uses stateful tokenizer/encoding instances across token-count examples for consistent tokenization behavior.

Semantic Kernel Python gains Handlebars and Jinja2 prompt templating with loops, variables, and static function execution.

$ git clone --branch python-0.9.3b1 https://github.com/microsoft/semantic-kernel.git
# already have the repo? check out this version:
$ git checkout python-0.9.3b1

• ›Adds Handlebars prompt template support, enabling loops, variables, and static function execution in prompts — matching the existing dotnet Handlebars implementation for cross-language compatibility.
• ›Adds Jinja2 prompt template support, enabling loops, variables, and static function execution in Python-native prompt workflows.

Semantic Kernel Python gains the Function Calling Stepwise Planner for agentic, tool-driven reasoning loops.

$ git clone --branch python-0.9.2b1 https://github.com/microsoft/semantic-kernel.git
# already have the repo? check out this version:
$ git checkout python-0.9.2b1

• ›Introduces FunctionCallingStepwisePlanner to the Python SDK, enabling LLM-driven stepwise planning via function/tool calling.

Semantic Kernel .NET 1.6.1 adds Agent file-ID reference handling and respects live Kernel plugin changes during auto-invocation.

$ git clone --branch dotnet-1.6.1 https://github.com/microsoft/semantic-kernel.git
# already have the repo? check out this version:
$ git checkout dotnet-1.6.1

• ›Adds File ID Reference Handling to the experimental Agent package, allowing user messages to be associated with file IDs in OpenAI Storage.
• ›Kernel plugin changes made during function auto-invocation are now respected at runtime, enabling dynamic plugin registration mid-call.
• ›Marks the Experimental.Orchestration.Flow package as alpha status.
• ›Adds a home automation example demonstrating dependency injection (DI) patterns in a real application.

Semantic Kernel Python 0.9.1b1 adds auto tool calling for OpenAI/AzureOpenAI with configurable invocation limits.

$ git clone --branch python-0.9.1b1 https://github.com/microsoft/semantic-kernel.git
# already have the repo? check out this version:
$ git checkout python-0.9.1b1

from semantic_kernel.connectors.ai.open_ai import OpenAIChatPromptExecutionSettings

settings = OpenAIChatPromptExecutionSettings(
    auto_invoke_kernel_functions=True,
    max_auto_invoke_attempts=5
)
result = await kernel.invoke_prompt(prompt, settings=settings)

• ›Adds auto tool calling for AzureOpenAI/OpenAI models, enabled via auto_invoke_kernel_functions=True and max_auto_invoke_attempts=<max_attempts_int> in PromptExecutionSettings; disabled by default.
• ›Adds function_name and plugin_name parameters to invoke_prompt for more precise prompt-level kernel function targeting.

Semantic Kernel Python hits 0.9.0 beta with reworked Kernel Arguments, Function Result, prompt templating, and complex type support for method functions.

$ git clone --branch python-0.9.0.beta1 https://github.com/microsoft/semantic-kernel.git
# already have the repo? check out this version:
$ git checkout python-0.9.0.beta1

• ›Splits KernelFunction into KernelFunctionFromPrompt and KernelFunctionFromMethod classes for cleaner separation of prompt-based and code-based function definitions.
• ›Adds complex type support for method functions, enabling richer input/output signatures beyond primitive types.
• ›Introduces major rework of Kernel Arguments, Function Result, and prompt templating engine.
• ›Removes memory tied directly to the Kernel object, decoupling memory management from the core kernel.
• ›Rebuilds the exceptions structure into a more Pythonic hierarchy.

Jan v0.4.8 adds vision (image input), model import, and an experimental Hugging Face to GGUF converter

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

• ›Adds vision capability ('Jan can see') enabling image input to supported models, including drag-and-drop image upload.
• ›Adds an import model feature so users can bring local models into Jan directly from the UI.
• ›Adds an experimental one-step conversion of Hugging Face models to GGUF format.
• ›Prompts users to download updates manually when automatic update is not possible.

KoboldCpp v1.60.1 adds native local image generation via stable-diffusion.cpp with A1111-compatible txt2img and flexible context sizing.

$ git clone --branch v1.60.1 https://github.com/LostRuins/koboldcpp.git
# already have the repo? check out this version:
$ git checkout v1.60.1

$ koboldcpp.exe --model my7b_q4.gguf --sdconfig my_sd15_model.safetensors --lowvram

$ koboldcpp.exe --model my7b_q4.gguf --contextsize 24576

$ koboldcpp.exe --model my7b_q4.gguf --remotetunnel

• ›Adds --sdconfig CLI flag to load a compatible SD1.5 or SDXL .safetensors fp16 image model, enabling native local image generation via stable-diffusion.cpp with an A1111-compatible txt2img endpoint usable from Kobold Lite, SillyTavern, and other compatible frontends.
• ›Adds --remotetunnel flag for Cloudflare tunnel support on macOS (amd64 only; M1 not supported).
• ›Enables LCM sampler support and allows loading GGUF-format SD models.
• ›Supports running KoboldCpp in text-gen-only, image-gen-only, or hybrid modes via launcher config, with a 'Compress Weights (quant)' option to reduce image model VRAM usage.
• ›Integrates AllTalk TTS support in Kobold Lite.

LocalAI v2.10.0 adds Intel GPU support, Elevenlabs/OpenAI TTS compatibility, Bitsandbytes quantization, and vLLM streaming.

$ git clone --branch v2.10.0 https://github.com/mudler/LocalAI.git
# already have the repo? check out this version:
$ git checkout v2.10.0

trust_remote_code: true

• ›Now honors trust_remote_code in model YAML config files for the AutoGPTQ and transformers backends, controlling whether remote code execution is allowed during model loading.
• ›Adds Bitsandbytes quantization support for the transformer backend, enabling lower-memory model loading on CUDA hardware.
• ›Adds compatibility layers for Elevenlabs and OpenAI TTS, allowing LocalAI to serve as a drop-in endpoint for both TTS APIs.
• ›Adds diffusers/transformers support for Intel GPUs, enabling image generation and transformer-based inference on Intel GPU hardware.
• ›Enables stream: true for the vLLM backend via AsyncLLMEngine, supporting true token streaming from vLLM-served models.

Adds AQLM quantization support, StreamingLLM for llama.cpp backends, and a new User Description field in the UI.

$ git clone --branch snapshot-2024-03-17 https://github.com/oobabooga/textgen.git
# already have the repo? check out this version:
$ git checkout snapshot-2024-03-17

• ›Adds a new 'User description' field in the UI for setting user personality/biography context.
• ›Adds experimental AQLM quantization support for loading AQLM-quantized models.
• ›Adds StreamingLLM support for the llamacpp and llamacpp_HF backends, enabling efficient long-context streaming inference.
• ›Bumps AutoAWQ to 0.2.3 on Linux, enabling use of newer AWQ model variants on that platform.

Adds cache_q4 KV-cache quantization for ExLlamaV2 and introduces an update wizard in textgen WebUI.

$ git clone --branch snapshot-2024-03-10 https://github.com/oobabooga/textgen.git
# already have the repo? check out this version:
$ git checkout snapshot-2024-03-10

• ›Adds cache_q4 option for ExLlamaV2 to enable 4-bit KV-cache quantization, reducing VRAM usage during inference.
• ›Introduces an update wizard to guide users through upgrading the WebUI.
• ›Extensions requirements are no longer installed by default, reducing setup overhead for users who don't need them.

Adds cubic sampling with a curve parameter and Gemma model compatibility via transformers 4.38.1.

$ git clone --branch snapshot-2024-03-03 https://github.com/oobabooga/textgen.git
# already have the repo? check out this version:
$ git checkout snapshot-2024-03-03

• ›Adds cubic sampling with a configurable curve parameter as a new sampling method for text generation.
• ›Supports Google Gemma models via a bump to transformers 4.38.1.

vLLM v0.3.3 adds StarCoder2, Orion model support, 2/3/8-bit GPTQ, Marlin Int4 kernels, and experimental structured output for the OpenAI server.

$ git clone --branch v0.3.3 https://github.com/vllm-project/vllm.git
# already have the repo? check out this version:
$ git checkout v0.3.3

• ›Adds experimental structured output (JSON, Regex) via guided decoding for the OpenAI-compatible API server.
• ›Adds logit bias support for the OpenAI API.
• ›Adds log probabilities for Chat Completions in the OpenAI-compatible API.
• ›Adds cache_config information to Prometheus metrics.
• ›Supports 2/3/8-bit GPTQ quantization models.

Ollama v0.1.29 adds AMD GPU acceleration and experimental Safetensors model import via Modelfile

$ git clone --branch v0.1.29 https://github.com/ollama/ollama.git
# already have the repo? check out this version:
$ git checkout v0.1.29

$ FROM /path/to/safetensors-model-dir

# Save as a Modelfile, then:
ollama create my-model -f Modelfile

• ›Adds experimental Safetensors model import via the FROM <directory with safetensors model> directive in a Modelfile.
• ›Shows documentation for supported environment variables when running ollama <command> -h.
• ›Adds AMD GPU acceleration support (preview) on Windows, Linux, and Docker, covering AMD Radeon RX, Radeon PRO, and Instinct families including MI300X, 7900 XTX, W7900, and more.

Ollama v0.1.28 adds StarCoder2 and DolphinCoder models with improved LLaVA 1.6 support.

$ git clone --branch v0.1.28 https://github.com/ollama/ollama.git
# already have the repo? check out this version:
$ git checkout v0.1.28

$ ollama run starcoder2:7b

$ ollama run dolphincoder

• ›Adds starcoder2 model (3B, 7B, and 15B parameter variants), a transparently trained open code LLM.
• ›Adds dolphincoder model, a code-focused chat model based on StarCoder2 15B.
• ›Improves support for llava 1.6 models.
• ›Vision models such as llava now respond better to text-only prompts.

Triton v2.43.0 adds Windows Python backend support and switches OpenTelemetry tracing to Batch Span Processor with configurable batching parameters.

$ git clone --branch v2.43.0 https://github.com/triton-inference-server/server.git
# already have the repo? check out this version:
$ git checkout v2.43.0

• ›OpenTelemetry trace mode switches to Batch Span Processor, enabling bulk export of completed spans with size-based batching controlled by bsp_max_export_batch_size and bsp_max_queue_size, and time-based batching controlled by bsp_schedule_delay.
• ›Adds base Python backend functionality for Windows, enabling Python model serving on Windows deployments.
• ›Removes Wait/Read(avg) and Overhead metrics for gRPC from the Trace Summary Tool to eliminate inaccurate readings.

Phoenix Evals now uses the Messages API for Claude on Amazon Bedrock.

$ git clone --branch arize-phoenix-evals-v0.6.0 https://github.com/Arize-ai/phoenix.git
# already have the repo? check out this version:
$ git checkout arize-phoenix-evals-v0.6.0

• ›Updates the Bedrock integration to use the Messages API when calling Claude models, enabling compatibility with the newer Claude API surface.

Arize Phoenix v3.17.0 adds Mistral AI as an eval model and a response_format argument to MistralAIModel.

$ git clone --branch arize-phoenix-v3.17.0 https://github.com/Arize-ai/phoenix.git
# already have the repo? check out this version:
$ git checkout arize-phoenix-v3.17.0

from phoenix.evals import MistralAIModel

model = MistralAIModel(
    model="mistral-large-latest",
    response_format={"type": "json_object"}
)

• ›Adds response_format argument to MistralAIModel to control output formatting when using Mistral for evals.
• ›Adds Mistral AI as a supported eval model via MistralAIModel in the evals module.

Arize Phoenix Evals v0.5.0 adds response_format argument to MistralAIModel.

$ git clone --branch arize-phoenix-evals-v0.5.0 https://github.com/Arize-ai/phoenix.git
# already have the repo? check out this version:
$ git checkout arize-phoenix-evals-v0.5.0

from phoenix.evals import MistralAIModel

model = MistralAIModel(response_format={"type": "json_object"})

• ›Adds response_format argument to MistralAIModel to control the output format of Mistral AI responses.

Arize Phoenix Evals v0.4.0 adds Mistral as a supported evaluation model.

$ git clone --branch arize-phoenix-evals-v0.4.0 https://github.com/Arize-ai/phoenix.git
# already have the repo? check out this version:
$ git checkout arize-phoenix-evals-v0.4.0

• ›Adds Mistral as a supported model for running evals via the arize-phoenix-evals library.

Arize Phoenix 3.16.0 adds a delete project UI.

$ git clone --branch arize-phoenix-v3.16.0 https://github.com/Arize-ai/phoenix.git
# already have the repo? check out this version:
$ git checkout arize-phoenix-v3.16.0

• ›Adds a UI action to delete projects directly from the interface.

Phoenix 3.15.0 adds experimental persistent span storage to launch_app() and project archiving/deletion.

$ git clone --branch arize-phoenix-v3.15.0 https://github.com/Arize-ai/phoenix.git
# already have the repo? check out this version:
$ git checkout arize-phoenix-v3.15.0

• ›Extends launch_app() with experimental span storage configurable via environment variables for storage path and storage type enums.
• ›Adds project archiving and deletion capabilities.

Phoenix v3.14.0 adds experimental append-only text file span storage.

$ git clone --branch arize-phoenix-v3.14.0 https://github.com/Arize-ai/phoenix.git
# already have the repo? check out this version:
$ git checkout arize-phoenix-v3.14.0

• ›Introduces experimental span storage backed by append-only text files.

Arize Phoenix v3.13.0 adds Python 3.12 support.

$ git clone --branch arize-phoenix-v3.13.0 https://github.com/Arize-ai/phoenix.git
# already have the repo? check out this version:
$ git checkout arize-phoenix-v3.13.0

• ›Adds support for Python 3.12.

Arize Phoenix Evals v0.3.0 adds Python 3.12 support.

$ git clone --branch arize-phoenix-evals-v0.3.0 https://github.com/Arize-ai/phoenix.git
# already have the repo? check out this version:
$ git checkout arize-phoenix-evals-v0.3.0

• ›Adds Python 3.12 support to the phoenix-evals library.

Arize Phoenix 3.12.0 enables dynamic project switching in the UI.

$ git clone --branch arize-phoenix-v3.12.0 https://github.com/Arize-ai/phoenix.git
# already have the repo? check out this version:
$ git checkout arize-phoenix-v3.12.0

• ›Enables dynamic project switching, allowing users to change the active project without restarting.

Phoenix v3.11.0 adds a context manager to pause tracing and embeds project inside GraphQL spans.

$ git clone --branch arize-phoenix-v3.11.0 https://github.com/Arize-ai/phoenix.git
# already have the repo? check out this version:
$ git checkout arize-phoenix-v3.11.0

• ›Adds a context manager to temporarily pause tracing, allowing selective suppression of trace collection within a code block.
• ›Embeds project as a private attribute inside GraphQL spans, making project context available at the span level.

Arize Phoenix 3.10.0 adds PHOENIX_PROJECT_NAME env var and smarter default project display.

$ git clone --branch arize-phoenix-v3.10.0 https://github.com/Arize-ai/phoenix.git
# already have the repo? check out this version:
$ git checkout arize-phoenix-v3.10.0

$ PHOENIX_PROJECT_NAME=my-rag-pipeline python app.py

• ›Adds support for the PHOENIX_PROJECT_NAME environment variable to specify the active project without code changes.
• ›UI now displays the first non-empty project by default instead of a blank project view.

Arize Phoenix 3.9.0 adds copy-to-clipboard for prompt templates in the UI.

$ git clone --branch arize-phoenix-v3.9.0 https://github.com/Arize-ai/phoenix.git
# already have the repo? check out this version:
$ git checkout arize-phoenix-v3.9.0

• ›Adds copy-to-clipboard functionality for prompt templates and related content in the UI.

Phoenix evals graduates from experimental: install standalone via pip install arize-phoenix-evals or bundled with pip install arize-phoenix[evals].

$ git clone --branch arize-phoenix-v3.8.0 https://github.com/Arize-ai/phoenix.git
# already have the repo? check out this version:
$ git checkout arize-phoenix-v3.8.0

• ›Promotes phoenix.evals out of phoenix.experimental.evals — now installable as a standalone package (pip install arize-phoenix-evals) or bundled (pip install arize-phoenix[evals]).
• ›Adds trace count to GraphQL project queries.

Phoenix Evals v0.2.0 updates AnthropicModel to use the messages API.

$ git clone --branch arize-phoenix-evals-v0.2.0 https://github.com/Arize-ai/phoenix.git
# already have the repo? check out this version:
$ git checkout arize-phoenix-evals-v0.2.0

• ›Updates AnthropicModel to use the messages API, replacing the legacy completions interface.

Arize Phoenix v3.7.0 adds project listing and project node interface capabilities.

$ git clone --branch arize-phoenix-v3.7.0 https://github.com/Arize-ai/phoenix.git
# already have the repo? check out this version:
$ git checkout arize-phoenix-v3.7.0

• ›Adds project listing support, enabling users to enumerate and navigate projects.
• ›Adds a project node interface for interacting with individual project nodes.

Phoenix v3.6.0 adds project-name-scoped span storage and querying plus an auto-expanding side nav.

$ git clone --branch arize-phoenix-v3.6.0 https://github.com/Arize-ai/phoenix.git
# already have the repo? check out this version:
$ git checkout arize-phoenix-v3.6.0

• ›Enables storing and querying spans by project name, allowing traces to be scoped and retrieved per project.
• ›Side navigation panel now auto-expands on hover in the UI.

Phoenix v3.5.0 adds metadata columns to spans/traces tables, a new project-aware side nav, and removes the token processing module from phoenix.evals.

$ git clone --branch arize-phoenix-v3.5.0 https://github.com/Arize-ai/phoenix.git
# already have the repo? check out this version:
$ git checkout arize-phoenix-v3.5.0

• ›Removes the token processing module from phoenix.evals, streamlining the evals package surface.
• ›Adds metadata columns to the spans and traces table in the UI for richer trace inspection.
• ›New side navigation with projects view, making multi-project workflows accessible from the main UI.

Langfuse v2.15.0 adds time-based cursor parameters to GET /traces, /scores, and /observations for incremental sync.

$ git clone --branch v2.15.0 https://github.com/langfuse/langfuse.git
# already have the repo? check out this version:
$ git checkout v2.15.0

$ curl -X GET 'https://<your-langfuse-host>/api/public/traces?fromTimestamp=2024-05-01T00:00:00Z' \
  -H 'Authorization: Basic <base64-encoded-credentials>'

$ curl -X GET 'https://<your-langfuse-host>/api/public/observations?fromStartTime=2024-05-01T00:00:00Z' \
  -H 'Authorization: Basic <base64-encoded-credentials>'

• ›Adds fromTimestamp query parameter to GET /traces and GET /scores endpoints to enable time-based cursor pagination for incremental sync into external systems.
• ›Adds fromStartTime query parameter to GET /observations endpoint for the same time-based cursor sync pattern.

Langfuse v2.14.2 adds a project dropdown in the nav and optional Turnstile captcha on sign-in.

$ git clone --branch v2.14.2 https://github.com/langfuse/langfuse.git
# already have the repo? check out this version:
$ git checkout v2.14.2

• ›Adds optional Cloudflare Turnstile captcha challenge on the sign-in page to reduce bot and credential-stuffing risk.
• ›Adds a project dropdown to the navigation bar and reorganises settings menu items.

Langfuse v2.14.0 shows human-readable display names in filters while preserving IDs in URLs.

$ git clone --branch v2.14.0 https://github.com/langfuse/langfuse.git
# already have the repo? check out this version:
$ git checkout v2.14.0

• ›Filter UI now shows human-readable display names instead of raw IDs, while URLs continue to use IDs for stable linking.

Dataset items API now returns results ordered by creation date descending.

$ git clone --branch v2.13.1 https://github.com/langfuse/langfuse.git
# already have the repo? check out this version:
$ git checkout v2.13.1

• ›Dataset items returned by the API are now ordered by createdAt descending, so the most recently created items appear first.

Langfuse v2.13.0 adds pagination to dataset tables and optional metadata on dataset runs.

$ git clone --branch v2.13.0 https://github.com/langfuse/langfuse.git
# already have the repo? check out this version:
$ git checkout v2.13.0

• ›Adds optional metadata field to dataset runs, enabling structured context to be attached to each run.
• ›Adds pagination to the datasets tables in the UI, improving navigation of large dataset collections.

Langfuse v2.12.0 adds a prompt playground and column visibility counts to data tables.

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

• ›Adds a prompt playground UI for interactively testing and iterating on prompts.
• ›Adds column visibility count to tables, showing how many columns are currently displayed.

Adds per-provider SSO account-linking toggles via new environment variables for Google and GitHub auth.

$ git clone --branch v2.11.2 https://github.com/langfuse/langfuse.git
# already have the repo? check out this version:
$ git checkout v2.11.2

$ AUTH_GOOGLE_ALLOW_ACCOUNT_LINKING=false
AUTH_GITHUB_ALLOW_ACCOUNT_LINKING=false

• ›Adds AUTH_GOOGLE_ALLOW_ACCOUNT_LINKING and AUTH_GITHUB_ALLOW_ACCOUNT_LINKING environment variables to control whether SSO logins via Google or GitHub are allowed to link to existing accounts.
• ›Project settings UI now always displays the hostname.

Langfuse v2.11.1 adds score source tracking, a level column in the traces table, and extended prompts table.

$ git clone --branch v2.11.1 https://github.com/langfuse/langfuse.git
# already have the repo? check out this version:
$ git checkout v2.11.1

• ›Adds source field to scores, allowing the origin of each score to be recorded and distinguished.
• ›Adds level column to the traces table for at-a-glance severity or log-level visibility across traces.
• ›Extends the prompts table with additional columns for richer prompt management views.
• ›Adds support for the Claude Haiku model in the model catalog.

Langfuse v2.11.0 extends ChatML message rendering to tools and functions in the UI.

$ git clone --branch v2.11.0 https://github.com/langfuse/langfuse.git
# already have the repo? check out this version:
$ git checkout v2.11.0

• ›Extends ChatML message rendering in the UI to cover additional JSON structures for tools and functions.

Adds AUTH_AZURE_ALLOW_ACCOUNT_LINKING env var to trust Azure AD emails for SSO account linking.

$ git clone --branch v2.10.10 https://github.com/langfuse/langfuse.git
# already have the repo? check out this version:
$ git checkout v2.10.10

• ›Adds AUTH_AZURE_ALLOW_ACCOUNT_LINKING environment variable to enable trusting Azure AD email addresses for SSO account linking.

Langfuse v2.10.9 adds a copyable LlamaIndex code snippet in the API popup.

$ git clone --branch v2.10.9 https://github.com/langfuse/langfuse.git
# already have the repo? check out this version:
$ git checkout v2.10.9

• ›Adds a copyable LlamaIndex integration snippet inside the API popup, making it faster to wire up LlamaIndex projects.

Langfuse v2.10.7 adds cost tracking to dataset run table and Claude 3 model support.

$ git clone --branch v2.10.7 https://github.com/langfuse/langfuse.git
# already have the repo? check out this version:
$ git checkout v2.10.7

• ›Adds cost column to the dataset run table, surfacing per-run cost data alongside existing metrics.
• ›Adds support for Claude 3 models in cost and usage tracking.

Langfuse v2.10.0 adds latency tables for traces/spans/generations and extends CSV/JSON export for generations.

$ git clone --branch v2.10.0 https://github.com/langfuse/langfuse.git
# already have the repo? check out this version:
$ git checkout v2.10.0

• ›Extends generations CSV and JSON export with additional fields.
• ›Adds latency tables for traces, spans, and generations with tooltip support.
• ›Extends the traces table with additional columns.
• ›Extends the sessions table with additional columns.
• ›Enables renaming of datasets.

LanceDB python-v0.6.5 adds async query support, reranking, HuggingFace dataset writing, and a to_batches API

$ git clone --branch python-v0.6.5 https://github.com/lancedb/lancedb.git
# already have the repo? check out this version:
$ git checkout python-v0.6.5

from datasets import load_dataset
ds = load_dataset("squad")
table = db.create_table("squad", data=ds)

• ›Adds to_batches API for streaming query results as Arrow record batches.
• ›Adds reranking support for vector and full-text search (FTS) queries in the Python API.
• ›Adds query support to the Python async API, including a refactored query interface.
• ›Supports writing HuggingFace Dataset and DatasetDict objects directly to LanceDB tables.
• ›Adds order_by field support for full-text search (FTS) queries.

LanceDB python-v0.6.4 expands the async API with index creation, time travel, update, list_indices, and index_stats.

$ git clone --branch python-v0.6.4 https://github.com/lancedb/lancedb.git
# already have the repo? check out this version:
$ git checkout python-v0.6.4

• ›Adds create_index to the async Python API, enabling non-blocking index builds.
• ›Adds list_indices to the async Python API for querying available indices asynchronously.
• ›Adds index_stats to the Python API for retrieving statistics about a specific index.
• ›Adds update to the async Python API, enabling asynchronous record updates.
• ›Adds time travel operations to the async Python API, allowing point-in-time dataset queries asynchronously.

LanceDB v0.4.13 expands the async Python API with index creation, time travel, update, list_indices, and index_stats.

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

• ›Adds create_index to the async Python API, enabling non-blocking index builds.
• ›Adds list_indices to the async Python API for querying available indexes asynchronously.
• ›Adds index_stats to the Python API for inspecting index statistics.
• ›Adds update to the async Python API for asynchronous record updates.
• ›Adds time-travel operations (version rollback/query) to the async Python API.

LanceDB Cloud queries now support a configurable timeout parameter.

$ git clone --branch python-v0.6.3 https://github.com/lancedb/lancedb.git
# already have the repo? check out this version:
$ git checkout python-v0.6.3

• ›Adds configurable timeout for LanceDB Cloud queries, allowing callers to control how long a query waits before failing.

LanceDB v0.4.12 adds column management APIs, scalar index creation, remote table support in Rust, and paginated table listing.

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

• ›Adds add_columns, alter_columns, and drop_columns APIs for in-place schema and data manipulation on tables.
• ›Adds create scalar index to the SDK, enabling scalar (non-vector) index creation from client code.
• ›Adds page_token and limit parameters to the native table_names function for paginated table listing.
• ›Adds initial remote table implementation for the Rust SDK, enabling Rust clients to operate against remote LanceDB tables.
• ›Changes arrow from a direct dependency to a peer dependency in the TypeScript/Node.js package, giving callers control over the Arrow version.

LanceDB v0.6.2 adds async create_table/add, scalar index creation, model_names() for OpenAI embeddings, and API URL override.

$ git clone --branch python-v0.6.2 https://github.com/lancedb/lancedb.git
# already have the repo? check out this version:
$ git checkout python-v0.6.2

from lancedb.embeddings import get_registry

openai = get_registry().get('openai').create()
print(openai.model_names())

• ›Adds model_names() method to the OpenAI embedding function to list available models programmatically.
• ›Adds create_scalar_index to the Python SDK, enabling scalar index creation directly from the client.
• ›Adds page_token and limit parameters to the native table_names function for paginated table listing.
• ›Allows users to override the API URL, enabling custom or self-hosted LanceDB remote endpoints.
• ›Ports create_table to the async Python API and the remote Rust API.

Milvus v2.3.11 adds TLS for Kafka, varchar autoID in bulk insert, and collection-level flush rate limiting.

$ git clone --branch v2.3.11 https://github.com/milvus-io/milvus.git
# already have the repo? check out this version:
$ git checkout v2.3.11

• ›Adds TLS support for Kafka connections, securing message-bus traffic without requiring a separate proxy.
• ›Adds support for varchar autoID in bulk insert operations, enabling string-typed primary keys to be auto-generated at ingest time.
• ›Adds collection-level rate limiting for flush operations, giving operators fine-grained control over flush throughput per collection.
• ›Adds a load memory factor for enhanced memory estimation during loading, reducing over-provisioning risk.
• ›Enables Milvus containers to run as a non-root user, improving deployment security posture.

Qdrant v1.8.0 adds Scroll API ordering by payload, datetime index, collection-exists API, minimum-match filters, and 16x faster sparse vector search.

$ git clone --branch v1.8.0 https://github.com/qdrant/qdrant.git
# already have the repo? check out this version:
$ git checkout v1.8.0

$ curl -X POST 'http://localhost:6333/collections/events/points/scroll' \
  -H 'Content-Type: application/json' \
  -d '{
    "limit": 50,
    "order_by": {
      "key": "timestamp"
    }
  }'

$ curl -X PUT 'http://localhost:6333/collections/events/index' \
  -H 'Content-Type: application/json' \
  -d '{
    "field_name": "created_at",
    "field_schema": "datetime"
  }'

$ curl -X POST 'http://localhost:6333/collections/products/points/search' \
  -H 'Content-Type: application/json' \
  -d '{
    "vector": [0.1, 0.2, 0.3],
    "limit": 10,
    "filter": {
      "min_should": {
        "conditions": [
          {"key": "category", "match": {"value": "electronics"}},
          {"key": "in_stock", "match": {"value": true}},
          {"key": "rating", "range": {"gte": 4.0}}
        ],
        "min_count": 2
      }
    }
  }'

• ›Adds order_by support to the Scroll API, enabling results to be ordered by payload field values.
• ›Adds a datetime payload index type for efficient filtering over date-time ranges.
• ›Adds an API endpoint to check whether a collection exists.
• ›Adds min_should (minimum number of conditions to match) support in payload filters.
• ›Improves the set_payload API to support modifying nested fields.