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 v2.8.5 adds AgentOSTools for platform observability and Moonshot thinking-mode toggle with file/video input support.
└──▷ GET THIS VERSION
$ git clone --branch v2.8.5 https://github.com/agno-agi/agno.git
# already have the repo? check out this version:$ git checkout v2.8.5
└──▷ USE IT
Enable Moonshot extended reasoning for a complex analysis task.
python
from agno.models.moonshot import Moonshot
model = Moonshot(use_thinking=True)
agent = Agent(model=model)
agent.print_response('Analyze the security implications of this architecture')
›Adds AgentOSTools class, a read-only platform operations toolkit that reports on AgentOS usage, latency, failures, schedules, evals, components, and pending approvals.
›Adds use_thinking parameter to the Moonshot integration to toggle thinking mode.
›Adds file and video input support to the Moonshot integration.
›Adds latency and error stats grouped by agent, team, workflow, or endpoint — plus tool and model call stats — to Traces, implemented for PostgresDb and SqliteDb.
└──▷ BREAKING ON UPGRADE
!The Moonshot integration default model is changed to kimi-k3; any setup relying on the previous default model will now use kimi-k3 without an explicit override.
OpenAI Agents SDK v0.19.0 adds Programmatic Tool Calling, a @tool decorator alias, and a Vercel cloud bucket mount strategy.
└──▷ GET THIS VERSION
$ git clone --branch v0.19.0 https://github.com/openai/openai-agents-python.git
# already have the repo? check out this version:$ git checkout v0.19.0
└──▷ USE IT
Decorate a plain function as an agent tool using the new short-form @tool alias from the public agents.decorators module.
python
from agents.decorators import tool
@tool
def get_weather(city: str) -> str:
"""Return current weather for the given city."""
return fetch_weather_api(city)
›Adds agents.tool.ProgrammaticToolCallingTool class, enabling supported OpenAI Responses models to generate JavaScript to coordinate eligible tools, with per-tool allowed_callers, structured function-tool outputs, and integration with Runner streaming, guardrails, approvals, sessions, and RunState.
›Adds the public agents.decorators module and a shorter @tool alias alongside existing function and guardrail decorators.
›Extends function tools to support async callable objects in addition to plain async functions.
›Adds VercelCloudBucketMountStrategy for sandbox sessions, excluding bucket contents from workspace persistence.
›SDK configuration now consistently accepts either typed settings objects or plain dictionaries across agents, runs, models, sessions, sandboxes, and voice pipelines, with validation for unknown settings.
HolmesGPT 0.37.0 adds JSON log format for scrapers, Langfuse OpenTelemetry tracing, and AZURE_AD_TOKEN Entra ID auth support.
└──▷ GET THIS VERSION
$ git clone --branch 0.37.0 https://github.com/HolmesGPT/holmesgpt.git
# already have the repo? check out this version:$ git checkout 0.37.0
›Supports pre-acquired AZURE_AD_TOKEN environment variable for Entra ID authentication with Azure AD, enabling token-based auth flows without interactive login.
›Adds JSON logging format support for log scrapers, enabling structured log ingestion pipelines.
›Adds Langfuse support in OpenTelemetry tracing for LLM observability and trace analytics.
Autonomous coding agent as an SDK, IDE extension, or CLI assistant.
Cline desktop v0.0.5 adds Cline API key onboarding, browser sign-in cancellation, and a major performance overhaul.
└──▷ GET THIS VERSION
$ git clone --branch desktop-v0.0.5 https://github.com/cline/cline.git
# already have the repo? check out this version:$ git checkout desktop-v0.0.5
›Adds a Cline API key path to onboarding so users can authenticate without a browser sign-in flow.
›Enables cancellation of a pending browser sign-in instead of waiting indefinitely.
›Streaming responses now coalesce updates instead of re-rendering the entire chat per token, unlocking smooth real-time output at scale.
›Animated background renders at a locked 60fps (up from ~10fps) and typing in the composer drops from 245 slow keystrokes to 3, making the app usable for longer sessions.
›Native folder picker and command execution no longer block the UI while the sidecar writes session logs or discovers the editor.
+1 moreshow less
›MCP server cards are now consistent across marketplace views, with a single uninstall action and setup guidance shown on installed servers.
An MCP that lets AI tools securely connect to your infrastructure, write IaaS code, debug issues, and assist during incidents - without risking production stability. Built for security teams to approve and infrastructure teams to experience like magic.
Runner identity now derives from hostname; pack retention protects versions active runners still advertise.
└──▷ GET THIS VERSION
$ git clone --branch v0.35.0 https://github.com/AndrewDryga/emisar.git
# already have the repo? check out this version:$ git checkout v0.35.0
›A runner with no explicit runner.id now presents its hostname as its identity, so a rebooted host reconnects as the same runner and a replaced ephemeral host auto-enrolls as a new one — no generated identity file required.
›Pack retention no longer sweeps a version that a connected runner still advertises, preventing dispatch from failing with pack_untrusted during quiet weeks of inactivity.
›The hosted admin runner bootstraps from a version-pinned release download with retries instead of querying the GitHub API, preventing anonymous rate limits from blocking control-plane VM startup.
›Every high or critical pack action now requires either a successful behavior case or a machine-readable risk exception (requires_cluster, requires_external_service, …), enforced by CI on every change.
›The snmp pack's OSPF adjacency fixture now builds its network namespace on AppArmor-confined hosts.
+1 moreshow less
›Behavior matrix reporting now completes in less than half the previous wall-clock time via pre-pulled images, slow-suite sharding across CI rows, and concurrency caps on heavyweight services.
llama.cpp b10142 adds multimodal vision support for MiniMax-M3, including sparse attention and GPU-accelerated inference.
└──▷ GET THIS VERSION
$ git clone --branch b10142 https://github.com/ggml-org/llama.cpp.git
# already have the repo? check out this version:$ git checkout b10142
›Adds vision support for MiniMax-M3 via mtmd (multimodal) framework, including mmproj and clip graph construction for the vision tower.
›Implements MiniMax-M3 sparse attention (MSA) with CUDA-native indexer op and flash attention for sparse layers, with dense fallback for multi-sequence --kv-unified mode.
›Sets default KV cache type to F32 for MiniMax-M3 and allocates in-cache k_idx only for architectures that opt in, reducing compute buffer from ~6.8 GiB to ~4.2 GiB at ub2048/62k context.
›Supports prompt caching and multi-stream inference (-np N when kv_unified=false) for MiniMax-M3, with context shift explicitly disallowed for this architecture.
›Adds MiniMax-M3 end-of-sequence token to the llama vocabulary.
+1 moreshow less
›Changes MiniMax-M3 image resize padding to 'none' and resize algorithm to Bicubic (Pillow) during conversion.
└──▷ BREAKING ON UPGRADE
!All MiniMax-M3 GGUFs generated before this release must be regenerated due to the renaming of MSA index tensors to the indexer naming convention.
vLLM v0.26.0 adds the Inkling model family, head_dtype fp32 lm_head, per-KV-cache-group attention backends, KV offloading metrics, and major API/frontend expansions.
└──▷ GET THIS VERSION
$ git clone --branch v0.26.0 https://github.com/vllm-project/vllm.git
# already have the repo? check out this version:$ git checkout v0.26.0
└──▷ TRY IT
Filter completions against a banned-word list using the new bad_words field on the OpenAI-compatible completions endpoint.
›Adds head_dtype config to enable fp32 lm_head for generation models, extended to the LoRA path and given a ROCm torch.mm fast path.
›Adds kv_cache_dtype field inside speculative_config to set a separate KV-cache dtype for speculative decoding.
›Adds kv_cache_dtype_skip_layers support for MLA quantization.
›Adds blocks_per_chunk config key for heterogeneous KV groups in KV offloading.
›Adds bad_words parameter to /v1/completions for OpenAI-compatible completions.
+46 moreshow less
›Adds logprob_token_ids field exposed on Python OpenAI endpoints.
›Adds include_reasoning parameter for non-Harmony models on OpenAI-compatible endpoints.
›Adds num_cache_creation_tokens field populated on Messages responses.
›Adds /abort_requests endpoint on the RLHF dev API router.
›Adds continue_final_message handling with renderer sentinel in the Rust frontend.
›Adds endpoint plugins framework for extending API routing.
›Introduces the Inkling model family with full support: base modeling, piecewise CUDA graph, Hopper FA4 relative attention, MTP=1 speculative decoding, LoRA, and ModelOpt NVFP4 quantization.
›Adds support for selecting a different attention backend per KV-cache group, enabling better hybrid-model support.
›Makes sliding-window an explicit backend capability for per-backend declaration.
›Adds partial prefix-cache hit for hybrid models and selective hybrid cache retention.
›Reports prefix-cache-reused blocks in full report mode.
›Adds KV offloading metrics: basic offloading gauges, CPU cache read/write split gauges, and tiering-lookup-delay sync/async histograms.
›Adds object-store secondary KV tier with workload identity support and DP-replica-aware tiering.
›Adds runtime draft weight update for speculative decoding.
›Adds hybrid (SWA + full attention) DFlash drafters for speculative decoding.
›Adds nvfp4_per_token online MoE quantization.
›Adds Humming w[2-7]a[4,8] weight-only inference with compressed-tensors.
›Adds int4 quantization for the emulation MoE backend.
›Adds INT2 XPU weight-only quant linear.
›Adds FlashInfer MoE LoRA for BF16 models.
›Adds LoRA support for tower/connector in LlavaNextVideo.
›Adds opt-in persistence and reuse of the memory-profiling result across boots.
›Adds Rust frontend support for multimodal video and audio inputs.
›Adds Seed-OSS tool parser to the Rust frontend.
›Adds native vllm-bench port in the Rust frontend.
›Adds Deepstream video decoding backend.
›Adds overlap of preprocessing and computation for pooling models in offline inference.
›Adds new models: BertForMaskedLM, RobertaForTokenClassification, XLMRobertaForTokenClassification, LongCat-Flash-Lite n-gram embedding, Cosmos3 Edge Reasoner, Cosmos3-Super, and TranslateGemma-12b-it.
›Migrates Olmo/Olmo2, MistralLarge3, and HunyuanVL to the Transformers modeling backend (Transformers 5.13.0).
›Adds automatic fallback to ViT data parallelism when TP is unavailable for multimodal models.
›Adds NIXL pipeline-parallel prefill in push mode for PD disaggregation.
›Adds Decode Context Parallel (DCP) hybrid attention support and DCP + Eagle for Tokenspeed MLA backends.
›Adds native macOS arm64 CPU wheel builds.
›Adds s390x NUMA topology support for CPU inference.
›Adds POWER VSX math function optimization and IBM Power docker builds using prebuilt wheels.
›Adds CuTeDSL compilation progress bar for UX feedback during compilation.
›Adds human-readable integer formatting for more CLI arguments.
›Adds expanded GPU profiler config scope and annotations.
›Adds worker exit-code logging when a process dies unexpectedly.
›Replaces diskcache with a non-pickle alternative to eliminate pickle deserialization risks.
›Adds resource-bounds validation to derender endpoints and sanitizes server file paths from validation error responses.
›Adds grammar compilation failure handling that avoids crashing the engine.
›Adds timeout guard on lm-format-enforcer regex compilation.
›Adds STATEFUL trainer-send abstractions for RLHF.
›Adds DSpark speculative decoding for DSv4 on AMD ROCm and XPU.
›Adds DFlash speculative decoding for GDN models on CPU.
└──▷ BREAKING ON UPGRADE
!Models TeleChat, Persimmon, and Fuyu have been removed and are no longer supported.
vMLX 1.6.18 adds a process-wide SSD cache budget with LRU eviction and detailed cache reporting.
└──▷ GET THIS VERSION
$ git clone --branch v1.6.18 https://github.com/jjang-ai/vmlx.git
# already have the repo? check out this version:$ git checkout v1.6.18
›Adds a process-wide SSD cache budget with coordinated LRU eviction and off-request-path publication, plus reporting of attempted prefixes, applied reuse, uncached suffixes, reconstruction, fallback prefill, and aggregate disk usage.
›Validates and atomically pins the exact ordered cache chain before reuse, preventing recycled paged-cache block IDs from receiving stale prefix credit.
›Accepts MiniMax tool arguments emitted as request-schema XML children without exposing native tool markup as visible answer text.
›Retains media across Responses tool continuations and permits local attachment playback in the Electron app.
›Derives fresh-session sampling, parser, output, template, and native MTP settings from the selected model bundle without converting inherited values into hidden saved overrides.
+1 moreshow less
›Aligns Auto reasoning with each model bundle's native policy unless a request supplies an explicit supported thinking budget, keeping reasoning-marker aliases out of visible content.
└──▷ BREAKING ON UPGRADE
!This release requires JANG 2.5.34; earlier JANG versions are no longer supported.
Phoenix v19.8.0 adds Claude Opus 5 to the curated agent model list and the Playground.
└──▷ GET THIS VERSION
$ git clone --branch arize-phoenix-v19.8.0 https://github.com/Arize-ai/phoenix.git
# already have the repo? check out this version:$ git checkout arize-phoenix-v19.8.0
›Adds Claude Opus 5 to the curated model list for agent evaluations.
›Registers Claude Opus 5 in the Playground for interactive prompt testing.
Phoenix 19.7.0 adds span detail downloads, a root-span DSL predicate, and richer span inspection tables in the UI.
└──▷ GET THIS VERSION
$ git clone --branch arize-phoenix-v19.7.0 https://github.com/Arize-ai/phoenix.git
# already have the repo? check out this version:$ git checkout arize-phoenix-v19.7.0
›Adds parent_span is None root-span predicate to the span filter DSL, enabling queries scoped to only root spans.
›Adds span detail downloads from the span detail view.
›Adds evaluator trace isolation and experiment metadata support to the pytest plugin.
›Adds collapse/expand row controls in the tracing tables UI.
›Adds a searchable table view for span attributes in the UI.
+3 moreshow less
›Adds span annotations and notes rendered as tables in the span detail UI.
›Makes the retention policy project list manageable from the UI.
›Adds an expandable search button (expands from icon) in the UI.
THE AI TOOLCHAIN