Backfill issue -037 — The AI Toolchain

EDITIONS tail grep head diff uniq

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.

VIEW

ISSUE VIEW full issue

Do you prefer this view?

$ tct list   # 8 tools matched

AI & LLM Tooling

◆ AI Coding Agents

Cline

cli-v3.0.40 NOTES

Autonomous coding agent as an SDK, IDE extension, or CLI assistant.

Cline CLI v3.0.40 adds manual API key entry for OAuth providers and clearer ClinePass usage-limit errors.

└──▷ GET THIS VERSION

$ git clone --branch cli-v3.0.40 https://github.com/cline/cline.git
# already have the repo? check out this version:
$ git checkout cli-v3.0.40

›Adds manual API key escape hatch for OAuth providers, letting you enter a key by hand from settings instead of going through OAuth.
›Surfaces ClinePass usage-limit errors clearly when the limit is reached.

Was this useful?

Block Goose

v1.42.0 NOTES

an open source, extensible AI agent that goes beyond code suggestions - install, execute, edit, and test with any LLM

Goose v1.0.42 adds Ollama cloud provider with dynamic model discovery, per-message token/cost stats, and project-grouped chat sessions.

└──▷ GET THIS VERSION

$ git clone --branch v1.42.0 https://github.com/block/goose.git
# already have the repo? check out this version:
$ git checkout v1.42.0

›Adds OllamaCloudProvider with dynamic model discovery and automatic context limit detection.
›Displays per-message usage stats in the UI — tokens, cost, time-to-first-token, and tokens/sec — with derived session totals.
›Groups chat sessions by project in the navigation panel for easier multi-project workflows.
›Shows project names in the directory switcher dropdown.
›Publishes provider bindings MVP to goose-sdk as a Python wheel, enabling Python-based provider integrations.

+1 moreshow less

›Adds AskAI bot hint suggestions for follow-up prompts.

Was this useful?

SST OpenCode

v1.17.19 NOTES

The open source coding agent.

OpenCode v1.17.19 adds per-prompt model selection, middle-click tab sessions, and a redesigned review panel with persistent file browsing.

└──▷ GET THIS VERSION

$ git clone --branch v1.17.19 https://github.com/sst/opencode.git
# already have the repo? check out this version:
$ git checkout v1.17.19

›Adds per-prompt model selection in the composer, letting you switch models on a message-by-message basis.
›Adds middle-click to open sessions in a new tab in the Desktop app.
›Adds a temporary setting to toggle between the old and new interface.
›Updates the review panel with persistent file browsing, improved file tabs, and easier open-in-app actions.
›Redesigns attachment cards and file comment chips in the new interface.

+1 moreshow less

›Restyled the Edit Project modal to match the new interface.

Was this useful?

◆ AI Agent Frameworks

OpenClaw

v2026.7.1 NOTES

Your own personal AI assistant. Any OS.

OpenClaw v2026.7.1 adds new AI providers, native macOS/mobile chat, Telegram Codex pairing, ClawRouter, and expanded CLI workflows.

└──▷ GET THIS VERSION

$ git clone --branch v2026.7.1 https://github.com/openclaw/openclaw.git
# already have the repo? check out this version:
$ git checkout v2026.7.1

└──▷ TRY IT

Attach to an existing running OpenClaw session from a second terminal to monitor or continue work without starting a new session.

$ openclaw attach

Pair a Telegram bot with Codex and steer an active run mid-execution.

$ /login
/steer
/tell <instruction>

›Adds Featherless, Claude Sonnet 5, Mythos 5, Meta Muse Spark 1.1, GPT-5.6, and ClawRouter as new model providers and routing options.
›Adds openclaw attach CLI subcommand plus node context-path support, soft-resume on prompt metadata changes, and actionable device-approval recovery guidance.
›Adds ClawRouter bundled provider plugin with credential-scoped dynamic model discovery, OpenAI-compatible and native Anthropic/Gemini transports, and managed budget reporting.
›Adds Meta muse-spark-1.1 Responses API support with streaming, tool calls, encrypted reasoning replay, and standalone npm distribution as @openclaw/meta-provider.
›Adds exit-triggered cron schedules and detached session-targeted cron runs.

+21 moreshow less

›Adds native macOS session browser with model and thinking pickers, slash commands, transcript export, and context usage display.
›Adds conversational Crestodian onboarding agent loop across CLI, web install, and macOS app with AI-guided provider setup, masked credential prompts, and exact operation approval binding.
›Adds iOS/Android offline pre-painted session and transcript caches, with Apple Watch full voice turns and iOS Gateway TTS reply playback with on-device fallback.
›Adds Gateway-backed session groups, unread state, rename, fork, archive, and delete controls across web, iOS, and Android.
›Adds Telegram Codex pairing via private /login, active-run steering with /steer and /tell, and multi-lane Telegram progress summaries.
›Adds native iMessage polls and Signal target aliases.
›Adds Android in-chat agent selector to switch active agents from the live chat screen.
›Adds Gateway host status panel in Control UI Settings showing host, network address, OS, runtime, uptime, CPU, memory, and disk details.
›Adds Logbook work-journal bundled plugin that turns paired-node screen snapshots into a private timeline, daily standups, and timeline-grounded Q&A in a Control UI tab.
›Adds operator-scoped tts.speak RPC returning configured-provider speech as inline whole-clip audio for remote clients.
›Adds auto-discovery of Ollama inference nodes.
›Adds Cursor Agent as an autoreview engine.
›Adds doctor --fix policy repair previews showing review-required Gateway bind and denied node-command changes without applying them.
›Expands doctor diagnostics to surface auth-profile, workspace, device-pairing, channel-plugin, memory-provider, systemd exhaustion, and Windows LAN firewall findings.
›Adds Control UI GitHub hover/focus cards showing issue and PR state, title, author, activity, comments, and change statistics.
›Adds generated session titles via utility-model routing with configurable per-agent utilityModel for lower-cost title generation.
›Adds Control UI reasoning-effort slider and compact context ring showing context-window progress, token counts, and active model.
›Adds Android syntax highlighting for fenced Kotlin, Swift, TypeScript, JavaScript, Python, Bash, and JSON code blocks.
›Slack integration now uses native assistant thread status and rotating loading messages by default; lifecycle reaction updates require messages.statusReactions.enabled: true.
›Adds Swedish mobile localization.
›Makes GPT-5.6 the default model for new setups with /think ultra for Sol and Terra and max for Luna.

└──▷ BREAKING ON UPGRADE

!Slack lifecycle reaction updates now require messages.statusReactions.enabled: true; previously they were emitted by default.

Was this useful?

holmesgpt

0.36.0 NOTES

SRE Agent - CNCF Sandbox Project

HolmesGPT 0.36.0 adds a /api/info endpoint, in-app TLS, Helm secret sugar, and multi-MCP-server tool-name isolation.

└──▷ GET THIS VERSION

$ git clone --branch 0.36.0 https://github.com/HolmesGPT/holmesgpt.git
# already have the repo? check out this version:
$ git checkout 0.36.0

└──▷ TRY IT

Check the running HolmesGPT instance's operational state — model, version, and configuration — without digging through logs.

$ curl -s https://<holmesgpt-host>/api/info | jq .

Mount an existing Kubernetes secret containing API keys into the HolmesGPT pod via Helm without manually listing each env var.

yaml

extraEnvVarsSecrets:
  - my-llm-api-keys

›Adds GET /api/info endpoint for operational visibility into the running HolmesGPT instance.
›Adds extraEnvVarsSecrets and envRef Helm values as shorthand for injecting secrets and configuring modelList without manual env var wiring.
›Enables in-app HTTPS/TLS support for the HolmesGPT API, allowing TLS termination without an external proxy.
›Supports multiple MCP servers simultaneously without tool-name collisions — identical tool names across servers are now disambiguated rather than overridden.
›Adds a standalone manifest for provisioning additional Azure MCP accounts alongside existing ones.

+5 moreshow less

›Marks multi-instance-capable data sources in the data source catalog so operators can distinguish sources that support parallel deployments.
›Consolidates Datadog and Elasticsearch/OpenSearch entries into unified catalog cards in the data source catalog.
›Routes call_aws JSON-argument commands through the no-JSON path, cutting approximately 87 seconds of retries on AWS MCP calls.
›GitHub MCP personal-access-token (PAT) path now uses native HTTP mode for more reliable authentication.
›Bounds conversation and tool-call claiming to worker capacity, preventing oversubscription under load.

└──▷ BREAKING ON UPGRADE

!KRR resource recommendation features have been removed from HolmesGPT.

Was this useful?

◆ Local LLM Runtimes

llama.cpp

b9993 4 RELEASES · 2026-07-13 NOTES STABLE

llama.cpp b9993 adds Tencent Hunyuan 3 (hy_v3) architecture support with MTP speculative decoding.

└──▷ GET THIS VERSION

$ git clone --branch b9993 https://github.com/ggml-org/llama.cpp.git
# already have the repo? check out this version:
$ git checkout b9993

›Adds hy_v3 (HunyuanV3 / HYV3ForCausalLM) architecture support: a MoE decoder with per-head Q/K RMSNorm, sigmoid router with expert selection bias, an always-active ungated shared expert, and leading dense blocks (first_k_dense_replace), enabling MTP speculative decoding for Tencent Hunyuan 3 models in GGUF format.

3 more releases in this issue · 2026-07-13

b9990 NOTES STABLE

llama.cpp b9990 adds Minimax2 EAGLE3 speculative decoding support.

└──▷ GET THIS VERSION

$ git clone --branch b9990 https://github.com/ggml-org/llama.cpp.git
# already have the repo? check out this version:
$ git checkout b9990

›Adds EAGLE3 speculative decoding support for Minimax2 models.

b9987 NOTES STABLE

llama.cpp b9987 adds gguf_get_tensor_ne accessor for retrieving tensor shapes from GGUF files.

└──▷ GET THIS VERSION

$ git clone --branch b9987 https://github.com/ggml-org/llama.cpp.git
# already have the repo? check out this version:
$ git checkout b9987

›Adds gguf_get_tensor_ne to the GGUF API, returning tensor shape as const int64_t * for programmatic inspection of model tensor dimensions.

b9985 NOTES STABLE

llama.cpp b9985 adds fused top-k MoE kernel for SYCL backends, controllable via GGML_SYCL_ENABLE_FUSION.

└──▷ GET THIS VERSION

$ git clone --branch b9985 https://github.com/ggml-org/llama.cpp.git
# already have the repo? check out this version:
$ git checkout b9985

└──▷ TRY IT

Enable fused top-k MoE kernel on a SYCL-backed build to accelerate MoE model inference on Intel GPUs.

$ GGML_SYCL_ENABLE_FUSION=1 ./llama-cli -m <model.gguf> -p '<prompt>'

›Adds GGML_SYCL_ENABLE_FUSION environment variable to enable fused top-k Mixture-of-Experts (MoE) kernel dispatch on SYCL backends, printed at startup alongside other env vars.

Was this useful?

vMLX

v1.6.7 NOTES

vMLX - JANGTQ Uber Compressed MLX Models - L2 Disk Cache (survives restart) + L1 Paged (super fast ttft) + Hybrid SSM Scheduler + Cont Batching + etc!

vMLX 1.6.7 enables paged KV cache by default for major text model families and turns reasoning on by default for all reasoning-capable models.

└──▷ GET THIS VERSION

$ git clone --branch v1.6.7 https://github.com/jjang-ai/vmlx.git
# already have the repo? check out this version:
$ git checkout v1.6.7

›Enables paged KV cache ON by default for autodetected text model families (Qwen3.5/3.6, Zaya, Nemotron-H, LFM2.5, Laguna, Hy3, Step-3.7) when continuous batching and prefix cache are active; --use-paged-cache / --no-paged-cache flags always take precedence.
›Turns reasoning ON by default for every reasoning-capable model family, with max_thinking_tokens honored on both streaming and non-streaming paths.
›Reconciles the Electron panel cache toggle, the --use-paged-cache launch flag, and the engine's effective cache policy so all three reflect the same state for every model family.

Was this useful?

Other / Uncategorized

◆ AI OBSERVABILITY

Arize Phoenix

arize-phoenix-v17.29.0 2 RELEASES · 2026-07-13 NOTES STABLE

Phoenix v17.29.0 moves PXI tracing server-side and adds a remote export setup command.

└──▷ GET THIS VERSION

$ git clone --branch arize-phoenix-v17.29.0 https://github.com/Arize-ai/phoenix.git
# already have the repo? check out this version:
$ git checkout arize-phoenix-v17.29.0

›Adds remote export setup command to pxi for configuring remote export.
›Moves PXI tracing to server-side execution for agent workflows.

1 more release in this issue · 2026-07-13

arize-phoenix-v17.28.0 NOTES STABLE

Phoenix 17.28.0 adds session annotation editing and collapsible playground sections in the UI.

└──▷ GET THIS VERSION

$ git clone --branch arize-phoenix-v17.28.0 https://github.com/Arize-ai/phoenix.git
# already have the repo? check out this version:
$ git checkout arize-phoenix-v17.28.0

›Enables editing of session annotations directly within the session details view.
›Adds collapsible playground sections via the TitledPanel component for a cleaner workspace layout.
›Modernizes core Tabs with an animated selection indicator and scroll snapping.

Was this useful?

// END OF TRANSMISSION

The AI Toolchain · curated daily · July 13, 2026
Releases & features from the open-source and commercial AI tools that practitioners actually run.

The AI Toolchain — issue -037, July 13, 2026