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 v2.2.2 adds LLM response caching, async SQLite, Claude Skills, Tavily Extract, and automatic MCP lifecycle management.
└──▷ GET THIS VERSION
$ git clone --branch v2.2.2 https://github.com/agno-agi/agno.git
# already have the repo? check out this version:$ git checkout v2.2.2
└──▷ USE IT
Cache model responses during development to avoid redundant API calls and cut costs.
python
from agno.agent import Agent
from agno.models.openai import OpenAIChat
agent = Agent(
model=OpenAIChat(id='gpt-4o', cache_response=True),
description='A cost-efficient research assistant',
)
agent.print_response('Summarize the history of the Roman Empire')
Use SQLite as an async session store when running agents in an async application.
python
from agno.agent import Agent
from agno.storage.sqlite import AsyncSqliteDb
storage = AsyncSqliteDb(db_path='tmp/agent_sessions.db')
agent = Agent(storage=storage)
import asyncio
asyncio.run(agent.aprint_response('Hello!'))
›Adds cache_response=True on the model class to cache LLM responses, reducing costs and speeding up development and testing.
›Adds AsyncSqliteDb class for asynchronous access to SQLite databases.
›Adds summary_request_message on SessionSummaryManager to override the user instruction sent to the LLM when generating session summaries.
›Adds refresh_connection on MCPTools and MultiMCPTools to manually refresh MCP server connections.
›Enables automatic MCP connection lifecycle management: passing MCPTools or MultiMCPTools directly to an Agent or Team now handles connect and reconnect automatically per run.
+4 moreshow less
›Adds support for Claude's native Skills, enabling enhanced reasoning, code execution, and tool interactions via the Anthropic integration.
›Adds TavilyReader for Tavily-based knowledge base integration and extends TavilyTools with URL content extraction via the Tavily Extract API, with full async support.
›Adds Team Model Inheritance: member agents automatically inherit model, reasoning_model, parser_model, and output_model from their parent team when none is specified.
›Enables Slack interface output to use mrkdwn formatting by default.
LangChain 1.0.3 adds structured output retry middleware and exports the UsageMetadata type.
└──▷ GET THIS VERSION
$ git clone --branch langchain==1.0.3 https://github.com/langchain-ai/langchain.git
# already have the repo? check out this version:$ git checkout langchain==1.0.3
›Exports UsageMetadata from the langchain package, making token-usage metadata directly importable.
$ git clone --branch 1.0.2 https://github.com/langchain-ai/langgraph.git
# already have the repo? check out this version:$ git checkout 1.0.2
└──▷ USE IT
Use Overwrite on a state field when you want the latest value to always win instead of being merged by a reducer.
python
from langgraph.types import Overwrite
from typing import Annotated
from typing_extensions import TypedDict
class State(TypedDict):
messages: Annotated[list, Overwrite()] # latest assignment replaces, no reducer merging
›Adds Overwrite type to bypass reducers and directly overwrite state channel values without merging.
›Adds cursory Python 3.14 support.
›Ships Checkpointers 3.0 release.
└──▷ BREAKING ON UPGRADE
!Python 3.9 is no longer supported; the minimum supported Python version has been raised.
LangGraph prebuilt 1.0.2 adds Python 3.14 support and drops Python 3.9.
└──▷ GET THIS VERSION
$ git clone --branch prebuilt==1.0.2 https://github.com/langchain-ai/langgraph.git
# already have the repo? check out this version:$ git checkout prebuilt==1.0.2
›Adds cursory Python 3.14 support for prebuilt components.
›Un-deprecates ToolNode, restoring it as a supported API.
└──▷ BREAKING ON UPGRADE
!Python 3.9 is no longer supported; setups running on Python 3.9 will break on upgrade.
PydanticAI v1.8.0 adds experiment metadata support and honors openai_supports_tool_choice_required in OpenAI Responses models.
└──▷ GET THIS VERSION
$ git clone --branch v1.8.0 https://github.com/pydantic/pydantic-ai.git
# already have the repo? check out this version:$ git checkout v1.8.0
›Adds openai_supports_tool_choice_required model profile setting support in OpenAIResponsesModel, enabling correct tool-choice behavior for OpenAI-compatible endpoints that do or don't support required tool choice.
›Adds experiment metadata support via the new experiment metadata API.
Continue 1.28.0 adds OpenAI Responses API support, enabling GPT-5 features via the oai-adapters layer.
└──▷ GET THIS VERSION
$ git clone --branch @continuedev/[email protected] https://github.com/continuedev/continue.git
# already have the repo? check out this version:$ git checkout @continuedev/[email protected]
›Implements the OpenAI Responses API in oai-adapters, enabling GPT-5 features through the adapter layer.
›Shows a gradient border in edit mode for visual distinction when editing.
›Removes the auto-accept edits setting.
└──▷ BREAKING ON UPGRADE
!The auto-accept edits setting has been removed; any configuration relying on it will no longer be recognized after upgrade.
Continue 1.31.0 integrates the OpenAI Responses API (enabling GPT-5 features) and removes the auto-accept edits setting.
└──▷ GET THIS VERSION
$ git clone --branch @continuedev/[email protected] https://github.com/continuedev/continue.git
# already have the repo? check out this version:$ git checkout @continuedev/[email protected]
›Integrates the OpenAI Responses API to enable GPT-5 features, implemented in oai-adapters.
›Removes the auto-accept edits setting (previously a configurable toggle).
›Adds a gradient border visual indicator when in edit mode.
└──▷ BREAKING ON UPGRADE
!The auto-accept edits setting has been removed; any configuration or workflow relying on it will no longer function after upgrading.
Continue 1.4.0 adds OpenAI Responses API support (enabling GPT-5), a session preview panel, and surfaces fetch errors to the agent.
└──▷ GET THIS VERSION
$ git clone --branch @continuedev/[email protected] https://github.com/continuedev/continue.git
# already have the repo? check out this version:$ git checkout @continuedev/[email protected]
└──▷ TRY IT
Browse and preview past sessions before resuming one, useful when you have many saved contexts.
$ cn ls
›Integrates the OpenAI Responses API, enabling GPT-5 features and parity support via oai-adapters.
›Adds a preview panel to the session selector, visible when running cn ls or /resume.
›Surfaces fetch errors back to the agent so it can react to network failures during tool calls.
›Adds a gradient border visual indicator when the UI enters edit mode.
└──▷ BREAKING ON UPGRADE
!The auto-accept edits setting is removed; any configuration relying on it will no longer apply after upgrading.
Auto-open a file when pasting its path into the project panel.
json
{
"open_file_on_paste": true
}
›Adds --reuse (-r) CLI flag to replace the workspace in an existing window instead of opening a new one.
›Adds status_bar.line_endings_button setting (set to true to enable) to display a line ending indicator in the status bar.
›Adds open_file_on_paste setting to configure auto-opening of a file on paste in the project panel.
›Adds title_bar.window_controls_position setting on Linux to customize window control button placement ('left' for macOS style or 'right' for Windows style).
›Adds editor::CollapseAllDiffHunks action, mirroring the existing editor::ExpandAllDiffHunks.
+20 moreshow less
›Adds workspace::NewFileSplit action to programmatically split the editor in a specified direction.
›Adds vim::HelixSubstitute and vim::HelixSubstituteNoYank actions for better Helix substitution behavior.
›Adds keyword.control syntax property for separately highlighting control flow keywords (if, else, return, etc.) in JS/TS/TSX, Rust, C, and C++.
›Adds zed: reveal log in file manager action to the command palette.
›Adds Claude Haiku 4.5 to the agent panel.
›Adds git diff view, toolbar, and destructive-action prompt for stash entries.
›Adds Vim filename autocomplete for commands: write, edit, split, vsplit, tabedit, tabnew.
›Adds Markdown Preview default Vim keybindings gt (next tab) and gT (previous tab).
›Adds Vue language server version 3 support.
›Adds comment language injections for built-in languages, enabling TODO and similar note highlighting when the comment extension is installed.
›Registers .rules, .cursorrules, .windsurfrules, and .clinerules as Markdown files.
›Adds basic ICO image format support.
›Adds searchable theme and icon theme dropdowns in the Settings UI.
›Adds a validation warning banner in the Settings UI when the active settings file is broken or invalid.
›Adds window controls to the settings window on Linux.
›Adds more bindings for the Emacs keymap.
›Adds the ability to hide the git blame popover by pressing Escape.
›Adds the ability to run search actions while a modal is active.
›Improves startup time on systems with slow process spawning.
›Improves editor font rendering on LoDPI displays.
$ git clone --branch v0.12.0-preview.0 https://github.com/google-gemini/gemini-cli.git
# already have the repo? check out this version:$ git checkout v0.12.0-preview.0
└──▷ TRY IT
Call the new executeCommand endpoint to list extensions programmatically via the API.
Get up and running with Kimi-K2.6, GLM-5.2, MiniMax, DeepSeek, gpt-oss, Qwen, Gemma and other models.
Ollama v0.12.7 adds Qwen3-VL and MiniMax-M2 models, file uploads in the app, adjustable thinking levels, and OpenAI-compatible embedding format support.
└──▷ GET THIS VERSION
$ git clone --branch v0.12.7 https://github.com/ollama/ollama.git
# already have the repo? check out this version:$ git checkout v0.12.7
└──▷ TRY IT
Request base64-encoded embeddings via the OpenAI-compatible endpoint — useful when piping embeddings into downstream systems that expect a specific format.
$ curl http://localhost:11434/v1/embeddings -H 'Content-Type: application/json' -d '{"model": "nomic-embed-text", "input": "Detect lateral movement in auth logs", "encoding_format": "base64"}'
›Adds Qwen3-VL multimodal model support in sizes from 2B to 235B parameters.
›Adds MiniMax-M2, a 230B parameter model optimized for coding and agentic workflows, available on Ollama's cloud.
›New app UI: attach one or multiple files when prompting a model.
›New app UI: adjust thinking levels for gpt-oss models to tune response depth.
›The OpenAI-compatible /v1/embeddings endpoint now supports the encoding_format parameter.
+2 moreshow less
›Extends tool-call parsing to handle responses that don't conform to the standard {"name": name, "arguments": args} format.
›New API documentation site launched at https://docs.ollama.com/api.
Phoenix 12.9.0 adds AWS IAM auth for database configuration and metadata display for experiment run annotations.
└──▷ GET THIS VERSION
$ git clone --branch arize-phoenix-v12.9.0 https://github.com/Arize-ai/phoenix.git
# already have the repo? check out this version:$ git checkout arize-phoenix-v12.9.0
›Enables AWS IAM authentication for database configuration.
›Displays metadata for experiment run annotations in the UI.
›Merges dataset label application with new label creation into a unified workflow.
›Moves reference output to the bottom pane of the experiment compare slideover.
THE AI TOOLCHAIN