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.
$ git clone --branch v1.0.17 https://github.com/github/copilot-cli.git
# already have the repo? check out this version:$ git checkout v1.0.17
›Supports HTTPS redirect URIs in MCP OAuth flows via a self-signed certificate fallback, enabling compatibility with providers that require HTTPS (such as Slack).
›Adds built-in skills to the CLI, starting with a guide for customizing Copilot cloud agent's environment.
›/resume session picker now loads significantly faster when working with large session histories.
›New createAgentSessionRuntime() and AgentSessionRuntime SDK API: closure-based runtime that recreates cwd-bound services and session config on every session switch, used consistently across startup, /new, /resume, /fork, and import.
›New defineTool() helper: create standalone custom tool definitions with full TypeScript parameter type inference, eliminating manual casts.
›Label timestamps in /tree: toggle timestamps on session tree entries with Shift+T, with smart date formatting and preservation through branching.
›Unified structured diagnostics: arg parsing, service creation, session option resolution, and resource loading now return info/warning/error diagnostics instead of logging or exiting, letting the app layer control presentation and exit behavior.
›Error diagnostics now reported for missing explicit CLI resource paths (-e, --skill, --prompt-template, --theme).
└──▷ BREAKING ON UPGRADE
!Extension post-transition events session_switch and session_fork are removed; use session_start with event.reason ("startup" | "reload" | "new" | "resume" | "fork") and event.previousSessionFile (set for "new", "resume", "fork").
!Session-replacement methods (newSession(), switchSession(), fork(), importFromJsonl()) are removed from AgentSession; use AgentSessionRuntime instead.
!session_directory is removed from extension and settings APIs.
!Unknown single-dash CLI flags (e.g. -s) now produce an error instead of being silently ignored.
›Adds save field in tasks.json to configure whether edited buffers are saved before running a task (previously always saved; now off by default).
›Adds controls for flexible or fixed width on Terminal and Agent panels, accessible from both the settings window and the status bar button right-click menu.
└──▷ BREAKING ON UPGRADE
!Edited buffers are no longer saved automatically before running a task by default; existing workflows that relied on pre-task auto-save must now set the save field in tasks.json explicitly.
Hermes v2026.4.3 adds pluggable memory backends, credential pool rotation, Camoufox stealth browser, secret exfiltration blocking, and ACP/MCP editor integration.
└──▷ GET THIS VERSION
$ git clone --branch v2026.4.3 https://github.com/NousResearch/hermes-agent.git
# already have the repo? check out this version:$ git checkout v2026.4.3
└──▷ TRY IT
Install the Camoufox stealth browser backend so the agent can browse without bot detection.
$ hermes tools
Configure multiple API keys for the same provider so Hermes rotates them automatically under load or on 401 failures.
Maintain a persistent session across multiple API server requests so the agent retains context between calls.
$ curl -X POST https://localhost:8080/v1/chat \
-H 'X-Hermes-Session-Id: my-session-42' \
-H 'Content-Type: application/json' \
-d '{"message": "continue the analysis"}'
›Adds pluggable memory provider interface — third-party backends (Honcho, vector stores, custom DBs) implement a provider ABC and register via the plugin system.
›Adds same-provider credential pools with automatic least_used rotation and 401-triggered failover across multiple API keys via credential_pool config.
›Adds Camoufox anti-detection browser backend for stealth browsing with persistent sessions, VNC URL discovery, and configurable SSRF bypass; auto-install via hermes tools.
›Adds inline diff previews for file write and patch operations in the tool activity feed.
›Adds API server session continuity via X-Hermes-Session-Id headers and real-time tool progress streaming for Open WebUI integration.
+11 moreshow less
›Adds ACP support for client-provided MCP servers — VS Code, Zed, and JetBrains editor MCP servers are picked up as additional agent tools.
›Adds secret exfiltration blocking: browser URLs and LLM responses are scanned for secret patterns, blocking URL-encoded, base64, and prompt-injection exfiltration attempts.
›Expands credential directory protection to .docker, .azure, and .config/gh; redacts execute_code sandbox output.
›Adds developer role support for GPT-5 and Codex models.
›Adds Anthropic long-context tier 429 handling — automatically reduces context to 200k on tier limit hits.
›Adds auto-detection of models from server probe during custom endpoint setup.
›Adds skill-aware slash commands — gateway dynamically registers installed skills as slash commands with paginated /commands list.
›Makes config.yaml the single source of truth for endpoint URLs, eliminating conflicts with environment variables.
›Adds Honcho full integration parity as the reference memory provider plugin with profile-scoped host/peer resolution.
›Adds token usage persistence for non-CLI sessions.
›Adds DM thread sessions seeded with parent transcript to preserve context.
LangGraph prebuilt 1.0.9 exposes richer execution information at runtime for agent observability.
└──▷ GET THIS VERSION
$ git clone --branch prebuilt==1.0.9 https://github.com/langchain-ai/langgraph.git
# already have the repo? check out this version:$ git checkout prebuilt==1.0.9
›Enhances the runtime with additional execution information, giving agents and tools access to more context about the current run.
PydanticAI v1.77.0 adds a local WebFetch tool, deferred tool loading, a ThreadExecutor capability, and smart Anthropic/Bedrock instruction caching.
└──▷ GET THIS VERSION
$ git clone --branch v1.77.0 https://github.com/pydantic/pydantic-ai.git
# already have the repo? check out this version:$ git checkout v1.77.0
└──▷ USE IT
Defer tool loading so tools are only resolved at call time, enabling dynamic tool search before execution.
python
from pydantic_ai import Agent
from pydantic_ai.tools import Tool
def my_tool_fn(ctx, query: str) -> str:
return f'result for {query}'
tool = Tool(my_tool_fn, defer_loading=True)
agent = Agent('openai:gpt-4o', tools=[tool])
Run an agent in a thread executor to avoid blocking the event loop when integrating with sync-heavy workloads.
python
import asyncio
from pydantic_ai import Agent
agent = Agent('anthropic:claude-sonnet-4-5')
async def main():
async with agent.using_thread_executor():
result = await agent.run('Summarize this document.')
print(result.output)
asyncio.run(main())
›Adds defer_loading parameter to tools and toolsets, enabling lazy/deferred tool loading to support tool search workflows.
›Adds Agent.using_thread_executor() method and a ThreadExecutor capability for running agents in thread executors.
›Adds a local WebFetch tool that activates automatically when a provider lacks built-in web-fetch support, extending WebFetch capability to more providers.
›Adds smart instruction caching for Anthropic and Bedrock providers — automatically inserts a cache boundary at the static/dynamic instruction split.
›Adds support for server_message_id in VercelAIEventStream.
›Adds --useswa flag to optionally enable Sliding Window Attention for Gemma 4, reducing VRAM usage significantly.
›Adds --jinja flag support for Gemma 4 to enforce correct chat completions format and avoid bad outputs from wrong templates.
›Adds --jinjatools flag to enable Jinja-based tool calling for broader model compatibility; falls back to universal tool calling when not set.
›Adds --jinja-kwargs / --chat-template-kwargs flags (matching llama.cpp syntax) to pass Jinja chat template kwargs, e.g. --chat-template-kwargs '{"enable_thinking":false}'.
›Adds --quantkv 3 option (also selectable in the GUI launcher) to enable BF16 KV cache type.
+16 moreshow less
›Adds --autoswap flag that, in router mode, swaps loaded features (Text/Images/Music) on and off per request type to save VRAM when running multi-feature configs.
›Adds --sdmaingpu flag allowing image generation models to be independently placed on any GPU.
›Adds support for credentials supplied via environment variables KCPP_ADMINPASSWORD and KCPP_PASSWORD at launch.
›Adds basic /v1/responses and /v1/messages compatibility API endpoints.
›Adds encapsulate_thinking request field (set to false to disable) controlling whether detected thinking content is sent via reasoning_content in chat completions.
›Adds Qwen3 TTS CustomVoice and VoiceDesign support, enabling narration with voice instructions in square brackets at the start of a TTS prompt (e.g. [A depressed woman is crying] I want to go home!).
›Adds config overwriting for admin mode: two config files (base and target) can now be specified on admin API reload and KoboldCpp will merge them before switching.
›Adds planner mode in Music Gen that uses the main LLM to generate better lyrics, toggled in the MusicUI advanced settings.
›Adds API key support for Music Gen.
›Adds ESRGAN passthrough for image gen, enabling upscale-only mode via img2img with denoise 0.0 and 1 step.
›Image gen now returns metadata alongside generated images.
›Doubles the logical batch size (while physical batch size is unchanged) when using pipeline parallel, improving throughput on multi-GPU setups.
›Adds a popular community models section in the help button menu, driven by .kcppt template files.
›Supports Gemma 4 models including vision, with AutoGuess non-thinking template applied by default.
›TTS embedded Music UI now supports both music and TTS generation across two tabs.
›Increases the max vision image limit and the GUI launcher max context size slider limit.
└──▷ BREAKING ON UPGRADE
!Detected thinking content is now sent via reasoning_content instead of content in chat completions API responses. Set encapsulate_thinking to false in your request to restore the previous behavior.
llama.cpp server gains --clear-idle to reclaim VRAM from idle KV cache slots on new task arrival.
└──▷ GET THIS VERSION
$ git clone --branch b8658 https://github.com/ggml-org/llama.cpp.git
# already have the repo? check out this version:$ git checkout b8658
└──▷ TRY IT
Run the inference server with idle-slot KV cache clearing enabled to reclaim VRAM between requests in a shared-GPU environment.
$ llama-server --model <model.gguf> --clear-idle
›Adds --clear-idle flag to llama-server to automatically free VRAM used by idle KV cache slots (via LLAMA_KV_KEEP_ONLY_ACTIVE) when a new task arrives, reducing GPU memory pressure in multi-user deployments. Opt out with --no-kv-clear-idle.
llama.cpp b8648 adds ZenDNN MUL_MAT_ID op acceleration for Mixture-of-Experts models.
└──▷ GET THIS VERSION
$ git clone --branch b8648 https://github.com/ggml-org/llama.cpp.git
# already have the repo? check out this version:$ git checkout b8648
›Adds MUL_MAT_ID op acceleration in the ggml-zendnn backend for Mixture-of-Experts (MoE) models, falling back to the CPU backend when total experts exceed 32.
›Updates the ZenDNN library reference to ZenDNN-2026-WW13.
›Adds W4A16 compressed tensors quantization support for CPU backends.
›Adds CompressedTensor W4A8 quantization support for Intel XPU.
›Adds AWQ Marlin support for ROCm backends.
›Adds full Gemma 4 architecture support including MoE, multimodal, reasoning, and tool-use; requires transformers>=5.5.0 (recommended image: vllm/vllm-openai:gemma4).
›Adds zero-bubble async scheduling with speculative decoding overlap, improving throughput.
›Adds general CPU KV cache offloading for V1 with pluggable CachePolicy and block-level preemption.
›Adds full CUDA graph capture for Vision Transformer (ViT) encoders.
›Adds piecewise CUDA graphs for pipeline parallelism in Model Runner V2.
›Adds configurable acceptance rate for Model Runner V2 spec decode.
›Adds DBO (Dual-Batch Overlap) microbatch optimization generalized to all model architectures.
›Adds NVIDIA B300/GB300 (SM 10.3) support with allreduce fusion enabled by default and tuned all-reduce communicator.
›Adds Triton autotuning disk cache enabled by default.
Phoenix client 2.3.0 adds a utility to convert ATIF data into trace trajectories.
└──▷ GET THIS VERSION
$ git clone --branch arize-phoenix-client-v2.3.0 https://github.com/Arize-ai/phoenix.git
# already have the repo? check out this version:$ git checkout arize-phoenix-client-v2.3.0
›Adds an ATIF-to-trace-trajectory conversion utility for transforming ATIF data into trace trajectories.
THE AI TOOLCHAIN