Backfill issue -052 — 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   # 3 tools matched

AI & LLM Tooling

◆ AI Coding Agents

Alibaba Qwen Code

v0.19.3 NOTES

Qwen Code v0.19.3 adds voice/control APIs, team memory, vision model fallback, MCP live reload, and a fleet of new CLI and web-shell capabilities.

└──▷ GET THIS VERSION

$ git clone --branch v0.19.3 https://github.com/QwenLM/qwen-code.git
# already have the repo? check out this version:
$ git checkout v0.19.3

└──▷ TRY IT

Cap how long a streaming response can sit idle before timing out — useful when running behind a strict gateway.

$ QWEN_STREAM_IDLE_TIMEOUT_MS=30000 qwen

Map a custom provider ID to the OpenAI SDK protocol in config so Qwen Code speaks the right wire format.

json

{
  "providerProtocol": "openai"
}

›Adds QWEN_STREAM_IDLE_TIMEOUT_MS environment variable to configure the stream idle timeout in the OpenAI pipeline.
›Adds providerProtocol config key to map a provider ID to an SDK protocol.
›Adds /model --vision subcommand flag to designate a fallback vision model.
›Adds /cd command support in ACP sessions.
›Adds a user-configurable keyterms file for voice dictation.

+25 moreshow less

›Adds 'voice' to ComposerToolbarAction for external visibility control of the voice toolbar button.
›Adds daemon workspace voice and control APIs to the serve layer.
›Adds a new API endpoint to query a single session's status by ID.
›Adds a git-shared team memory tier, enabling memory shared across a team via a git repository.
›Adds a bundled extension creator skill and supports @extension mention in CLI input autocomplete.
›MCP servers now reconcile live on settings change — no restart required.
›Web-shell /mcp dialog now allows browsing MCP server resources.
›Makes the telemetry sensitive span attribute limit configurable.
›Adds recordApiRequestBreakdown wiring into endLLMRequestSpan for richer LLM request telemetry (Phase 4c).
›Adds skill usage stats display in the CLI.
›Auto-generated skills now require user confirmation before being persisted.
›Vendors qwen-cua-driver with opt-in 0–1000 relative coordinate mode.
›TUI now partitions tool display by type — collapsing read/search tools and showing mutation tools individually.
›CLI now warns before foreground shell timeout expires.
›CLI footer now shows the count of scheduled tasks.
›Built-in status line preset is now enabled by default for new users.
›Registers Telegram bot command menu via the channels integration.
›Web-shell markdown tables gain Excel-style interactions and a manual toggle to enable/disable the enhanced table view.
›Web-shell code blocks now stream-highlight syntax as they arrive.
›Web-shell adds a workspace session sidebar.
›Web-shell adds a mobile responsive view for the TodoPanel.
›Web-shell adds error boundaries to prevent a render crash from white-screening the embed.
›Web-shell allows the host to override streaming loading phrases.
›Web-shell shows duration on finished thinking summaries.
›Startup text in auto mode is simplified and emoji removed.

└──▷ BREAKING ON UPGRADE

!The /remember command is decoupled from auto-extract and no longer writes to QWEN.md; existing workflows that relied on /remember updating QWEN.md will stop receiving those writes after upgrading.

Was this useful?

◆ AI Agent Frameworks

holmesgpt

0.34.0 NOTES

SRE Agent - CNCF Sandbox Project

HolmesGPT 0.34.0 adds admin API endpoints for runtime config reload, event-driven operator triggers, and JWT-signed approval tickets for Kubernetes remediation.

└──▷ GET THIS VERSION

$ git clone --branch 0.34.0 https://github.com/HolmesGPT/holmesgpt.git
# already have the repo? check out this version:
$ git checkout 0.34.0

›Adds admin API endpoints for runtime configuration reload, enabling live config changes without restarting the service.
›Introduces event-driven triggers in operator mode, allowing HolmesGPT to react to Kubernetes events automatically.
›Adds JWT-signed approval tickets to the Kubernetes Remediation MCP, with approval-based tool separation between safe and destructive actions.
›Updates AWS MCP Helm addon to native Streamable HTTP (v2.1.0).
›Adds firewall detection and user-friendly error handling for Supabase sign-in failures.

+1 moreshow less

›OAuth MCP now deletes expired token rows on 401 responses so users can re-authenticate cleanly.

└──▷ BREAKING ON UPGRADE

!The deprecated Tool._is_restricted() mechanism has been removed; any toolset code relying on it will break.

Was this useful?

◆ Local LLM Runtimes

llama.cpp

b9833 3 RELEASES · 2026-06-28 NOTES STABLE

llama.cpp b9833 adds a dedicated MiniCPM5 tool-call parser with XML tool-call support via autoparser.

└──▷ GET THIS VERSION

$ git clone --branch b9833 https://github.com/ggml-org/llama.cpp.git
# already have the repo? check out this version:
$ git checkout b9833

›Adds a dedicated MiniCPM5 tool-call parser using the openbmb-MiniCPM5-1B.jinja template, with XML tool-call handling via autoparser and strict JSON argument parsing.

2 more releases in this issue · 2026-06-28

b9831 NOTES STABLE

llama.cpp b9831 adds DFlash v2 support with per-layer sliding window attention.

└──▷ GET THIS VERSION

$ git clone --branch b9831 https://github.com/ggml-org/llama.cpp.git
# already have the repo? check out this version:
$ git checkout b9831

›Adds DFlash v2 support, including per-layer sliding window attention via layer_types configuration.

b9830 NOTES STABLE

llama.cpp b9830 exposes --offline to llama download for air-gapped and cache-check workflows.

└──▷ GET THIS VERSION

$ git clone --branch b9830 https://github.com/ggml-org/llama.cpp.git
# already have the repo? check out this version:
$ git checkout b9830

└──▷ TRY IT

Check whether a model is already cached before serving it in an air-gapped or network-restricted environment.

$ llama download --offline --model <model-name>

›Adds --offline flag support to llama download, letting scripts verify whether a model is already cached locally without making any network requests.

Was this useful?

// END OF TRANSMISSION

The AI Toolchain · curated daily · June 28, 2026
Releases & features from the open-source and commercial AI tools that practitioners actually run.

The AI Toolchain — issue -052, June 28, 2026

Alibaba Qwen Code

holmesgpt

llama.cpp