The daily firehose — everything the toolchain shipped today, already filtered.
// HOW THIS ISSUE IS MADE
We read every release from the 174 tools on our watchlist at the source — GitHub and GitLab release notes, vendor release pages and changelogs, project blogs and feeds, vendor press releases, and the source code behind the tag. Bug-fix-only releases and non-product newsroom noise are dropped; what's left is summarized down to the new capability, how to try it, and any screenshots or videos the release itself published. Every entry links to the sources it was built from.
Agno v2.6.10 adds four new model providers, YouTools, DOCX generation, context-provider streaming, and a files field on RunCompleted.
└──▷ GET THIS VERSION
$ git clone --branch v2.6.10 https://github.com/agno-agi/agno.git
# already have the repo? check out this version:$ git checkout v2.6.10
└──▷ USE IT
Use You.com search inside an agent to ground responses in live web results.
python
from agno.agent import Agent
from agno.tools.you import YouTools
agent = Agent(tools=[YouTools()], show_tool_calls=True)
agent.print_response('What are the latest AI model releases this week?')
›Adds files field on the RunCompleted event, exposing generated files at run completion.
›Adds YouTools class for You.com Search API integration.
›Adds DOCX file generation support.
›Adds google-interactions provider to the model string parser.
›Adds knowledge and managers support in the agent registry.
+6 moreshow less
›Streams sub-agent events from context providers.
›Persists cancelled runs properly for agents, teams, and workflows.
LangChain 1.3.3 adds interrupt_mode and when predicate to HumanInTheLoopMiddleware and typed subagent run projection.
└──▷ GET THIS VERSION
$ git clone --branch langchain==1.3.3 https://github.com/langchain-ai/langchain.git
# already have the repo? check out this version:$ git checkout langchain==1.3.3
›Adds interrupt_mode and when predicate parameters to HumanInTheLoopMiddleware for fine-grained control over when human-in-the-loop interrupts trigger.
›Projects subagent runs onto a typed run.subagents channel, enabling structured access to subagent execution data.
Claude Code is an agentic coding tool that lives in your terminal, understands your codebase, and helps you code faster by executing routine tasks, explaining complex code, and handling git workflows - all through natural language commands.
Claude Code v2.1.161 adds OTEL label slicing, smarter agents UI, MCP connector collapsing, and independent parallel tool results.
└──▷ GET THIS VERSION
$ git clone --branch v2.1.161 https://github.com/anthropics/claude-code.git
# already have the repo? check out this version:$ git checkout v2.1.161
└──▷ TRY IT
Slice Claude Code usage metrics by team or repo in your OTEL backend without any code changes — just set resource attributes before launching.
$ OTEL_RESOURCE_ATTRIBUTES="team=security,repo=infra-scanner" claude <prompt>
›Adds OTEL_RESOURCE_ATTRIBUTES values as labels on metric datapoints, enabling usage metrics to be sliced by custom dimensions like team or repo.
›Shows done/total progress before detail text in claude agents rows when work is fanned out; peek view highlights the longest-running item.
›Collapses never-signed-in claude.ai connectors in /mcp behind a 'Show unused connectors' row to reduce noise.
›Parallel tool calls now return independent results — a failed Bash command no longer cancels sibling calls in the same batch.
Claude Code v2.1.160 adds shell-startup write prompts, safer acceptEdits mode, and a grep-satisfies-read shortcut.
└──▷ GET THIS VERSION
$ git clone --branch v2.1.160 https://github.com/anthropics/claude-code.git
# already have the repo? check out this version:$ git checkout v2.1.160
›Adds a confirmation prompt before writing to shell startup files (.zshenv, .zlogin, .bash_login) and ~/.config/git/ to prevent unintended command execution.
›In acceptEdits mode, prompts before writing build-tool config files that grant code execution (.npmrc, .yarnrc*, bunfig.toml, .bazelrc, .pre-commit-config.yaml, .devcontainer/, etc.).
›Single-file grep/egrep/fgrep commands now satisfy the read-before-edit check, eliminating the need for a separate Read step after grepping a file.
›Renames the dynamic-workflow trigger keyword from workflow to ultracode; the keyword is highlighted in violet in the prompt input.
└──▷ BREAKING ON UPGRADE
!The CLAUDE_CODE_OPUS_4_6_FAST_MODE_OVERRIDE environment variable is removed and is now a no-op.
!The dynamic-workflow trigger keyword is renamed from workflow to ultracode; the word workflow no longer triggers a dynamic workflow run.
Triton v2.69.0 adds a Rust gRPC client, Azure Managed Identity auth, GPU_DEVICE_IDS for vLLM, and gRPC request cancellation.
└──▷ GET THIS VERSION
$ git clone --branch v2.69.0 https://github.com/triton-inference-server/server.git
# already have the repo? check out this version:$ git checkout v2.69.0
›Adds GPU_DEVICE_IDS to the vLLM backend config to pin multi-GPU models to specific GPU devices.
›Adds Azure Managed Identity authentication support for Azure Storage model repositories, enabling credential-free model loading.
›Adds a Rust gRPC client library as a new supported client language.
›Supports request cancellation in the gRPC C++ client.
›Adds MemberIsNullIs[type] helpers to TritonJson for safer JSON field-presence and null checks.
+5 moreshow less
›Adds complex I/O type support for PT2 models in the PyTorch backend.
›Adds auditwheel to the Dockerfile.sdksdk_build stage and the Model Analyzer devel container for arch-specific manylinux wheel tagging.
›Triton now rejects inference requests whose parameters or forward headers use Triton-reserved keys; the Python client enforces the same rejection on the send side.
›Triton validates and rejects inference requests containing duplicate output names in both HTTP and gRPC.
›Replaces pickle with json for inter-process payloads in the vLLM backend.
llama.cpp b9481 adds support for IBM Granite multilingual embedding R2 models and a new <arch>.hidden_activation GGUF key.
└──▷ GET THIS VERSION
$ git clone --branch b9481 https://github.com/ggml-org/llama.cpp.git
# already have the repo? check out this version:$ git checkout b9481
›Adds new GGUF key <arch>.hidden_activation (internal key LLM_KV_HIDDEN_ACT) and a writer, enabling models to declare their hidden-layer activation function (e.g. SwiGLU) in the GGUF metadata.
›Adds llm_ffn_op_type_from_string() helper and a new LLM_FFN_NONE = 0 sentinel to llm_ffn_op_type, centralising hidden-activation mapping in llama-model.cpp.
›Adds support for ibm-granite/granite-embedding-97m-multilingual-r2 and ibm-granite/granite-embedding-311m-multilingual-r2 embedding models, including SwiGLU FFN, a fixed-regex GPT-4o tokenizer variant (97m), and the Gemma4 tokenizer (311m).
›Extends the GGUF converter to emit hidden_activation from config.json and adds model hashes for the Granite Embedding Multilingual R2 models.
llama.cpp b9474 adds a Thinking mode toggle with reasoning effort levels to the chat UI.
└──▷ GET THIS VERSION
$ git clone --branch b9474 https://github.com/ggml-org/llama.cpp.git
# already have the repo? check out this version:$ git checkout b9474
›Adds a 'Thinking' toggle and status icon to the chat UI, enabling chain-of-thought / reasoning mode for supported models.
›Adds a reasoning effort levels selector in the model dropdown, letting users tune how much reasoning the model applies per request.
›Adds a chat template thinking detector utility and thinking support detection to the models store, so the UI automatically surfaces thinking controls only for capable models.
›Makes the server origin configurable via an environment variable for the UI build.
llama.cpp b9468 adds a POST /v1/chat/completions/control endpoint to interrupt model reasoning mid-generation in real time.
└──▷ GET THIS VERSION
$ git clone --branch b9468 https://github.com/ggml-org/llama.cpp.git
# already have the repo? check out this version:$ git checkout b9468
›Adds reasoning_control opt-in flag on completion slots that arms the budget sampler on demand, enabling the new control endpoint to call common_sampler_reasoning_budget_force and halt thinking at any point during streaming.
›Web UI gains a 'skip' button that appears only during the active reasoning phase (tracked via isReasoning streaming state), posts the streamed completion id back to the control endpoint, and is wired through the agentic flow.
vMLX 1.5.52 reads greedy generation defaults (temperature, top-p, top-k) directly from JANG model metadata and ships Sequoia- and Tahoe-compatible DMGs.
└──▷ GET THIS VERSION
$ git clone --branch v1.5.52 https://github.com/jjang-ai/vmlx.git
# already have the repo? check out this version:$ git checkout v1.5.52
›Reads and applies model-owned greedy generation defaults (temperature 0.00, top_p 1.0, top_k off) declared in JANG bundle metadata, so sampling behaviour is driven by the model file rather than hidden app-side assumptions.
›Updates the local Step-3.7-Flash JANG metadata path so startup settings (default temperature=0.0, top_p=1.0, top_k=0) derive from model files; explicit request parameters still override model defaults.
›Builds both public DMG flavors from the same source: Sequoia-compatible macosx_14 wheels and Tahoe-native macosx_26 wheels, with Developer ID signing, notarization, stapling, and Gatekeeper acceptance.
Phoenix 17.1.0 adds a load_dataset tool for the PXI agent and enables LLM-evaluator authoring directly in the PXI agent interface.
└──▷ GET THIS VERSION
$ git clone --branch arize-phoenix-v17.1.0 https://github.com/Arize-ai/phoenix.git
# already have the repo? check out this version:$ git checkout arize-phoenix-v17.1.0
›Adds load_dataset tool to the PXI agent in the Playground, enabling dataset loading directly from the agent interface.
›Enables LLM-evaluator authoring within the PXI agent, allowing practitioners to create and configure LLM-based evaluators without leaving the agent workflow.
›Adds skill loading display to surface skill status during PXI agent interactions.
›Adds warning colors and a search-off icon to the UI for clearer visual feedback.
Phoenix v17.0.0 adds admin-managed system settings for assistant enablement and trace recording policy.
└──▷ GET THIS VERSION
$ git clone --branch arize-phoenix-v17.0.0 https://github.com/Arize-ai/phoenix.git
# already have the repo? check out this version:$ git checkout arize-phoenix-v17.0.0
›Adds server-side system settings for admin-managed assistant enablement and trace recording policy control.
└──▷ BREAKING ON UPGRADE
!The addition of system settings with admin-managed assistant enablement and trace recording policy introduces breaking changes — see MIGRATION.md for upgrade steps.
Phoenix v16.6.0 adds playground model switching, web search toggle in agent chat, and code-evaluator authoring.
└──▷ GET THIS VERSION
$ git clone --branch arize-phoenix-v16.6.0 https://github.com/Arize-ai/phoenix.git
# already have the repo? check out this version:$ git checkout arize-phoenix-v16.6.0
›Adds code-evaluator authoring in PXI, enabling practitioners to write and manage custom code-based evaluators directly in the UI.
›Adds a model switching tool to the PXI playground, allowing users to swap models mid-session without leaving the interface.
›Adds a web search globe toggle to the agent chat interface, letting users enable or disable live web search during agent conversations.
THE AI TOOLCHAIN