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.
Override the output schema on a per-call basis so one agent instance can return different structured shapes for different tasks.
python
from agno.agent import Agent
from pydantic import BaseModel
class SummaryOutput(BaseModel):
summary: str
key_points: list[str]
agent = Agent(model=...)
result = agent.run("Summarise this document", output_schema=SummaryOutput)
Control Gemini reasoning depth by passing thinking_level when running a Gemini-backed agent.
python
from agno.models.google import Gemini
from agno.agent import Agent
agent = Agent(model=Gemini(thinking_level="high"))
agent.run("Explain the proof of Fermat's Last Theorem")
›Adds optimize_memories and aoptimize_memories methods to MemoryManager for summarizing a user's memories outside of agent runs (beta).
›Adds output_schema override support to run() and arun() on both Agent and Team, as well as AgentOS API endpoints, enabling per-call schema control at runtime.
›Adds api_key support for AWS Bedrock authentication.
›Adds thinking_level parameter support to Gemini.
›Introduces Context Compression (beta): compresses tool call results in a running agent context to stay within context windows and avoid rate limits.
+3 moreshow less
›Adds Gemini File Search support, including document store create/list/get/delete, direct file upload with custom chunking configuration and metadata, document management with metadata filtering, citation extraction helpers, and full async/await support. See cookbooks: cookbook/models/google/gemini/file_search_basic.py, cookbook/models/google/gemini/file_search_advanced.py, cookbook/models/google/gemini/file_search_rag_pipeline.py.
›Extends AWS Bedrock Claude compatibility with native Claude, adding support for thinking models and caching.
›Extends VertexAI Claude compatibility with native Claude, adding support for thinking models and caching.
Continue 1.32.0 adds headless CLI mode, Gemini thought signatures, and Opus 4-5 model support.
└──▷ 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]
›Removes TTY requirement for the CLI, enabling headless/non-interactive operation in CI and automation pipelines.
›Adds support for Gemini thought signature in model responses.
›Adds Opus 4-5 model updates to the adapter layer.
›Adds Klavis AI Strata to the MCP Cookbook integrations.
›Allows renaming a rule before it is created in the UI.
Qwen Code v0.3.0-nightly adds stream-json I/O format for headless/programmatic use and internationalization support.
└──▷ GET THIS VERSION
$ git clone --branch v0.3.0-nightly.20251127.c9af7481 https://github.com/QwenLM/qwen-code.git
# already have the repo? check out this version:$ git checkout v0.3.0-nightly.20251127.c9af7481
└──▷ TRY IT
Pipe structured commands into Qwen Code headlessly and receive machine-parseable streaming JSON output for CI/automation pipelines.
›Adds stream-json as a supported value for --input-format and --output-format options, enabling programmatic/headless use of Qwen Code via streaming JSON I/O.
›Adds internationalization (i18n) support for both UI text and LLM output.
An open-source AI agent that brings the power of Gemini directly into your terminal.
gemini-cli v0.20.0-nightly adds Databricks auth, hooks for tool/LLM events, MCP server instructions, and policy-driven model fallback.
└──▷ GET THIS VERSION
$ git clone --branch v0.20.0-nightly.20251127.5bed97064 https://github.com/google-gemini/gemini-cli.git
# already have the repo? check out this version:$ git checkout v0.20.0-nightly.20251127.5bed97064
›Adds Databricks authentication support and a custom header option for connecting gemini-cli to Databricks-hosted models.
›Adds hook integration for tool execution, enabling scripts or handlers to fire before/after the CLI invokes any tool.
›Adds hook integration for LLM request/response events, allowing interception or logging of raw model traffic.
›Adds support for MCP server instructions via a config option, letting operators inject server-level guidance into the context.
›Injects GoogleCredentialProvider headers into MCP client connections, enabling authenticated MCP server access.
+3 moreshow less
›Implements a policy-driven model fallback mechanism so the CLI can automatically switch models when a primary model is unavailable.
›Adds a configurable inactivity timeout for shell commands, preventing runaway shell tool calls from hanging indefinitely.
›Adds a consent flag to the Link command for non-interactive linking flows.
Get up and running with Kimi-K2.6, GLM-5.2, MiniMax, DeepSeek, gpt-oss, Qwen, Gemma and other models.
Ollama v0.13.1 adds two new Mistral models, tool calling and thinking support for cogito-v2.1, and switches nomic-embed-text to Ollama's native engine.
└──▷ GET THIS VERSION
$ git clone --branch v0.13.1 https://github.com/ollama/ollama.git
# already have the repo? check out this version:$ git checkout v0.13.1
└──▷ TRY IT
Run the new edge-optimized Ministral-3 model locally for lightweight inference tasks.
$ ollama run ministral-3
Use cogito-v2.1 with tool calling enabled for agentic workflows.
$ ollama run cogito-v2.1
›Adds Ministral-3, an edge-optimized model designed to run on a wide range of hardware.
›Adds Mistral-Large-3, a multimodal mixture-of-experts model for production-grade and enterprise workloads.
›Enables tool calling support for the cogito-v2.1 model.
›Adds thinking and tool parsing for cogito-v2.1.
›Switches nomic-embed-text to use Ollama's native engine by default.
THE AI TOOLCHAIN