Backfill issue -109 — 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 Agent Frameworks

OpenAI Agents SDK

v0.15.1 NOTES

OpenAI Agents SDK v0.15.1 exposes WebSocket keepalive options for the Responses API connection.

└──▷ GET THIS VERSION

$ git clone --branch v0.15.1 https://github.com/openai/openai-agents-python.git
# already have the repo? check out this version:
$ git checkout v0.15.1

›Exposes WebSocket keepalive options for the Responses API, giving callers control over connection liveness behavior.

Was this useful?

OpenClaw

v2026.5.2 NOTES

Your own personal AI assistant. Any OS.

OpenClaw v2026.5.2 adds git-based plugin installs, ClawHub search, proxy validation, Google Meet captions, and a structured heartbeat tool.

└──▷ GET THIS VERSION

$ git clone --branch v2026.5.2 https://github.com/openclaw/openclaw.git
# already have the repo? check out this version:
$ git checkout v2026.5.2

└──▷ TRY IT

Verify proxy configuration and destination reachability before rolling out proxy-routed OpenClaw commands in a restricted network.

$ openclaw proxy validate

Force a Gateway restart immediately and wait up to 60 seconds for it to come back, useful in automated deployment scripts.

$ openclaw gateway restart --force --wait 60s

›Adds openclaw gateway restart --force and --wait <duration> flags, with active task run ID logging before restart deferral and explicit forced-restart reporting.
›Adds first-class git: plugin installs with ref checkout, commit metadata, scanner/staging, and plugins update support for recorded git sources.
›Adds ClawHub plugin search plus Crestodian plugin list/search/install/uninstall operations with approval and audit coverage.
›Adds openclaw proxy validate so operators can verify effective proxy configuration, reachability, and allow/deny behavior before deploying proxy-routed commands.
›Includes package dependency install state in openclaw plugins list --json so scripts can detect missing plugin dependencies without runtime-loading plugins.

+21 moreshow less

›On the beta update channel, plugin updates now try @beta first and fall back to default/latest when no plugin beta release exists.
›Adds Grok 4.3 to the bundled xAI catalog and makes it the default xAI chat model.
›Adds googlemeet end-active-conference command for closing managed Google Meet spaces after a call.
›Adds googlemeet test-listen command and matching google_meet test_listen action so transcribe-mode joins wait for real caption or transcript movement before reporting listen-first health.
›Adds live caption health monitoring for Chrome transcribe mode in Google Meet, including caption observer state, transcript counters, last caption text, and recent transcript lines in status and doctor output.
›Adds SDK-facing tools.invoke RPC with shared HTTP policy, typed approval/refusal results, and SDK helper support.
›Adds extraBody/extra_body passthrough for OpenAI-compatible TTS endpoints, enabling custom fields such as lang in /audio/speech requests.
›Adds structured heartbeat_respond tool for tool-capable heartbeat runs so agents can record quiet outcomes or explicit notification text.
›Adds agents.defaults.skipOptionalBootstrapFiles config key for skipping selected optional workspace files during bootstrap without disabling required workspace setup.
›Adds UTC quarter-hour token buckets for the Usage Mosaic in Control UI with reuse for hour filtering.
›Adds opt-in channels.bluebubbles.replyContextApiFallback for fetching original messages from the BlueBubbles HTTP API on reply-context cache misses, with SSRF protection and concurrent-request coalescing.
›Adds accessGroup:<name> reusable message-channel access groups for Discord channel-audience DM authorization so allowlists can reference named groups across channel auth paths.
›Supports explicit WhatsApp Channel/Newsletter @newsletter outbound message targets with channel session metadata.
›Allows $include directives in Gateway config to read files from operator-approved OPENCLAW_INCLUDE_ROOTS directories.
›Adds a platform-level tool descriptor planner and caches plugin tool descriptors from api.registerTool(...) so repeated prompt-time planning skips plugin runtime loading.
›Adds a fast path for canonical absolute POSIX containment checks in hot filesystem walkers, avoiding repeated path.resolve and path.relative work.
›Externalizes ACPX behind @openclaw/acpx and diagnostics OpenTelemetry behind @openclaw/diagnostics-otel, keeping heavier runtime stacks out of core until installed.
›Keeps active Discord buttons, selects, and forms working across Gateway restarts until they expire, preserving multi-step interactions during upgrades.
›Publishes a safe default Slack App Home tab view on app_home_opened and tracks bot-participated threads across restarts so threaded conversations continue auto-replying after Gateway restarts.
›Moves macOS menu bar companion recent session context rows into a Context submenu while keeping usage and cost details root-level for compactness.
›Loads bundled plugins from the extensions/* pnpm workspace tree in source checkouts so plugin-local dependencies and edits are used directly.

└──▷ BREAKING ON UPGRADE

!The split subagent/ACP thread-spawn toggles are replaced by threadBindings.spawnSessions; thread-bound spawns now default on. Run openclaw doctor --fix to migrate the legacy keys.

Was this useful?

◆ Local LLM Runtimes

llama.cpp

b9000 NOTES

llama.cpp b9000 adds HMX-accelerated flash attention for Hexagon HTP prefill, with multi-threaded Q/O phases and FP16 softmax.

└──▷ GET THIS VERSION

$ git clone --branch b9000 https://github.com/ggml-org/llama.cpp.git
# already have the repo? check out this version:
$ git checkout b9000

›Adds HMX-accelerated flash attention (op_hmx_flash_attn_ext) for the Qualcomm Hexagon HTP backend, enabling hardware-accelerated prefill across GQA, ALiBi, softcap, and masked attention shapes.
›Enables multi-threaded Q-load and O-store phases via fa_phase_q_load and fa_phase_o_store worker-pool helpers, parallelising F32↔F16 conversion across row ranges on HVX threads.
›Adds experimental FP16 softmax path (FA_EXP_F16, controlled via CMake host-side option) using hvx_exp2_hf directly — ~22 ALU ops per 64 lanes versus ~44 for the F32 round-trip path.
›Relaxes the HMX matmul pipeline gate to cover k > n shapes (e.g. FFN_down layers), selecting pipeline vs. sequential layout based on actual chunker output rather than a static m>=128 && n>=256 threshold.
›Adds HMX flash-attention test coverage in test-backend-ops over the full {mask, ALiBi, logit_softcap} cross-product for prefill — head dims 64/128, GQA 4×4, kv=512/nb=64 and kv=113/nb=32 non-aligned cases, F16 KV.

Was this useful?

// END OF TRANSMISSION

The AI Toolchain · curated daily · May 2, 2026
Releases & features from the open-source and commercial AI tools that practitioners actually run.

The AI Toolchain — issue -109, May 2, 2026

OpenAI Agents SDK

OpenClaw

llama.cpp