Backfill issue -168 — 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   # 9 tools matched

AI & LLM Tooling

◆ AI Coding Agents

Cline

v3.69.0 NOTES

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

Cline v3.69.0 adds User-Agent headers to backend requests, SDK docs in-app, and improved Windows PowerShell hooks support.

└──▷ GET THIS VERSION

$ git clone --branch v3.69.0 https://github.com/cline/cline.git
# already have the repo? check out this version:
$ git checkout v3.69.0

›Adds User-Agent header to all requests sent to the Cline backend.
›Surfaces Cline SDK documentation directly on the Cline page.
›Improves hooks support for Windows PowerShell environments.

Was this useful?

GitHub Copilot CLI

v0.0.421 NOTES

GitHub Copilot CLI v0.0.421 adds repo-level config, --plugin-dir flag, MCP Elicitations, and COPILOT_CLI=1 env var for Git hooks.

└──▷ GET THIS VERSION

$ git clone --branch v0.0.421 https://github.com/github/copilot-cli.git
# already have the repo? check out this version:
$ git checkout v0.0.421

└──▷ TRY IT

Load and test a locally developed plugin without publishing it to a marketplace.

$ copilot --plugin-dir ./my-local-plugin

Share marketplace sources and launch messages across a team by committing a repo-level config file.

json

{
  "marketplaces": ["https://example.com/my-marketplace"],
  "launchMessage": "Welcome to the team repo!"
}

›Adds --plugin-dir flag to load a plugin from a local directory instead of a marketplace.
›Adds COPILOT_CLI=1 environment variable so Git hooks can detect Copilot CLI subprocesses and skip interactive prompts.
›Supports repo-level configuration via .github/copilot/config.json for shared project settings such as marketplaces and launch messages.
›Reads extraKnownMarketplaces from .claude/settings.json (project-level) for Claude plugin compatibility.
›Adds structured form input for the ask_user tool using MCP Elicitations (experimental).

+8 moreshow less

›ACP clients can now configure reasoning effort via session config options.
›AUTO theme reads the terminal's ANSI color palette directly so colors match the user's terminal theme.
›Clickable PR reference displayed next to branch name in the status bar.
›Mouse text selection is automatically copied to the Linux primary selection buffer (middle-click to paste).
›Links in the terminal are now clickable and open in the default browser.
›Markdown tables render with proper column widths, word wrap, and Unicode borders that adapt to terminal width.
›Autopilot permission dialog now appears on first prompt submission instead of on mode switch.
›Uses consistent ~/.copilot/pkg path for auto-update instead of XDG_STATE_HOME.

Was this useful?

SST OpenCode

v1.2.16 NOTES

The open source coding agent.

OpenCode v1.2.16 adds remote workspace support, pending tool-call counts in TUI, compact UI, and a Go provider list command.

└──▷ GET THIS VERSION

$ git clone --branch v1.2.16 https://github.com/sst/opencode.git
# already have the repo? check out this version:
$ git checkout v1.2.16

›Adds basic remote workspace support, including username/password authentication when connecting to a remote server.
›Adds workspace_id to the session table and WorkspaceContext to core.
›Shows pending tool call count in the TUI instead of a generic 'Running…' message.
›Adds arrow indicator for active tool execution in the TUI.
›Adds keybindings to navigate between child sessions.

+12 moreshow less

›Enables markdown rendering by default via OpenTUI upgrade to v0.1.86.
›Adds Go provider list command.
›Exposes OPENCODE_PID environment variable on shutdown and kills orphaned MCP child processes.
›Adds compact UI mode to the desktop app.
›Adds recent projects section to the desktop command palette.
›Adds auto-accept all permissions mode to the desktop app.
›Adds Turkish locale support for app and UI packages.
›Adds Warp to the desktop open menu.
›Adds comprehensive animation system with multiple easing functions and transition utilities to the desktop app.
›Enables faster session switching via windowed rendering and staged timeline in the desktop app.
›Adds zen mode feature to the SDK.
›Opens the desktop app in PowerShell instead of Command Prompt on Windows.

Was this useful?

Alibaba Qwen Code

v0.11.1 NOTES

Qwen Code v0.11.1 adds AGENTS.md context support, Ctrl+Y retry shortcut, and improved auth dialog UX.

└──▷ GET THIS VERSION

$ git clone --branch v0.11.1 https://github.com/QwenLM/qwen-code.git
# already have the repo? check out this version:
$ git checkout v0.11.1

└──▷ TRY IT

Retry a failed API request mid-session without re-entering your prompt — useful when hitting transient rate limits or network errors.

$ # When a request fails in the Qwen Code CLI, press:
Ctrl+Y

›Adds Ctrl+Y keyboard shortcut in the CLI to retry failed requests without retyping.
›Supports AGENTS.md as a default context file, recognized automatically alongside existing context files.
›Improves the auth dialog UX with a clearer three-option layout for easier authentication setup.

Was this useful?

Google gemini-cli

v0.32.0 NOTES

An open-source AI agent that brings the power of Gemini directly into your terminal.

gemini-cli v0.32.0 adds shell autocompletion, plan-mode editor support, a Gemma router, billing overage flow, and richer OTEL telemetry.

└──▷ GET THIS VERSION

$ git clone --branch v0.32.0 https://github.com/google-gemini/gemini-cli.git
# already have the repo? check out this version:
$ git checkout v0.32.0

›Adds interactive shell autocompletion inside the CLI REPL.
›Supports opening and modifying the plan in an external editor during plan approval.
›Adapts the planning workflow based on task complexity, encouraging multi-select with option descriptions.
›Adds a temporary flag to disable workspace policies.
›Hides the workspace policy update dialog and auto-accepts by default.

+17 moreshow less

›Enables model steering in workspace context.
›Introduces an experimental Gemma Router using a LiteRT-LM shim as a composite model classifier strategy.
›Implements a task tracker foundation and service for tracking agent tasks.
›Introduces Kind.Agent classification for sub-agent tools, with contiguous parallel admission.
›Enables the generalist agent.
›Improves A2A streaming reassembly and task continuity.
›Improves A2A content extraction.
›Loads extensions in parallel for faster startup.
›Adds support for a policy engine in extensions.
›Instruments OTEL traces with more attributes.
›Implements G1 AI credits overage flow with billing telemetry.
›Adds low/full CLI error verbosity mode for a cleaner UI.
›Adds a Ctrl+O hint to truncated content messages in the UI.
›Renames the grep_search tool's include parameter to include_pattern.
›Plumbs the maxAttempts setting through config args.
›Maps HTTP 499 to retryable quota errors.
›Increases fetch timeout.

└──▷ BREAKING ON UPGRADE

!The grep_search tool's include parameter is renamed to include_pattern.

Was this useful?

◆ Local LLM Runtimes

KoboldCpp

v1.109.2 NOTES

KoboldCpp v1.109.2 adds native music generation via Ace Step and Qwen3-TTS voice cloning with new CLI flags.

└──▷ GET THIS VERSION

$ git clone --branch v1.109.2 https://github.com/LostRuins/koboldcpp.git
# already have the repo? check out this version:
$ git checkout v1.109.2

└──▷ TRY IT

Generate music natively using the Ace Step backend with a 6 GB VRAM budget.

$ koboldcpp --musicllm ace-step-lm.gguf --musicdiffusion ace-step-diffusion.gguf --musicembeddings ace-step-embedder.gguf --musicvae ace-step-vae.gguf

›Adds music generation via Ace Step with CLI flags --musicllm, --musicdiffusion, --musicembeddings, --musicvae, and --musiclowvram; runs in as little as 4 GB of VRAM.
›Adds a music generation UI at http://localhost:5001/musicui when music model files are loaded.
›Adds Qwen3-TTS support for high-quality voice cloning; specify a directory of .mp3 or .wav voice samples with --ttsdir to enable voice selection during TTS narration.
›Adds support for importing SillyTavern JSONL exports.
›Adds ComfyUI auth token support.

+2 moreshow less

›--admindir now indexes subdirectories up to 1 level deep.
›SmartCache is now automatically enabled for RNN/hybrid models (e.g., Qwen 3.5) unless fast forwarding is disabled.

Was this useful?

llama.cpp

b8191 NOTES

llama.cpp b8191 adds an optimized Q4_1 matrix-multiplication OpenCL kernel for Adreno GPUs.

└──▷ GET THIS VERSION

$ git clone --branch b8191 https://github.com/ggml-org/llama.cpp.git
# already have the repo? check out this version:
$ git checkout b8191

›Adds an optimized q4_1 matrix-multiplication (mm) OpenCL kernel targeting Qualcomm Adreno GPUs, accelerating Q4_1 quantized inference on mobile/edge hardware via ggml_cl_mul_mat_q4_1_f32_adreno.

Was this useful?

◆ AI Agent Frameworks

OpenClaw

v2026.3.2 NOTES

Your own personal AI assistant. Any OS.

OpenClaw v2026.3.2 adds PDF analysis, Ollama memory embeddings, config validation CLI, 64-target SecretRef coverage, and 10+ new plugin runtime APIs.

└──▷ GET THIS VERSION

$ git clone --branch v2026.3.2 https://github.com/openclaw/openclaw.git
# already have the repo? check out this version:
$ git checkout v2026.3.2

└──▷ TRY IT

Validate your OpenClaw config file and get machine-readable errors before restarting the gateway — useful in CI or pre-deploy hooks.

$ openclaw config validate --json

Use a locally-running Ollama model for memory embeddings instead of a cloud provider, keeping all data on-premise.

$ openclaw config set memorySearch.provider ollama && openclaw config set memorySearch.fallback ollama

›Adds first-class pdf tool with native Anthropic and Google PDF provider support, extraction fallback for non-native models, and configurable defaults (agents.defaults.pdfModel, pdfMaxBytesMb, pdfMaxPages).
›Adds openclaw config validate (with --json flag) to validate config files before gateway startup, with detailed invalid-key paths in startup errors.
›Adds memorySearch.provider = "ollama" and memorySearch.fallback = "ollama" support for local Ollama-powered memory embeddings.
›Expands SecretRef support across 64 credential targets, including runtime collectors and openclaw secrets planning/apply/audit flows; unresolved refs now fail fast on active surfaces.
›Adds inline file attachment support for sessions_spawn (subagent runtime only) with base64/utf8 encoding, transcript redaction, and configurable limits via tools.sessions_spawn.attachments.

+15 moreshow less

›Adds api.runtime.stt.transcribeAudioFile(...) so plugins can transcribe local audio files through OpenClaw's configured audio providers.
›Exposes channelRuntime on ChannelGatewayContext so external channel plugins access shared runtime helpers without internal imports.
›Adds runtime.events.onAgentEvent and runtime.events.onSessionTranscriptUpdate for extension-side event subscriptions, with isolated listener failure handling.
›Exposes runtime.system.requestHeartbeatNow(...) so plugins can wake targeted sessions immediately after enqueueing system events.
›Adds internal hook events message:transcribed and message:preprocessed, plus isGroup/groupId context on outbound message:sent for group-conversation correlation.
›Includes sessionKey in session_start/session_end hook events so plugins can correlate lifecycle callbacks with routing identity.
›Adds optional tools.media.audio.echoTranscript and echoFormat to send a pre-agent transcript confirmation message back to the originating chat.
›Adds optional disableAudioPreflight on Telegram group/topic config to skip mention-detection preflight transcription for inbound voice notes.
›Adds sendPayload support across direct-text-media, Discord, Slack, WhatsApp, Zalo, and Zalouser adapters with multi-media iteration and chunk-aware text fallback.
›Defaults channels.telegram.streaming to partial (from off) so new Telegram setups get live preview streaming out of the box.
›Adds cli.banner.taglineMode (random | default | off) to control startup tagline behavior.
›Adds first-class MiniMax-M2.5-highspeed model support across provider catalogs and onboarding flows, with legacy MiniMax-M2.5-Lightning compatibility.
›Adds PDF file output and rendering quality controls (fileQuality, fileScale, fileMaxWidth) for generated diff artifacts.
›Rebuilds Zalo Personal plugin (@openclaw/zalouser) to use native zca-js in-process, removing dependency on external CLI binaries.
›Adds sendMessageDraft for Telegram private DM preview streaming with separated reasoning/answer preview lanes.

└──▷ BREAKING ON UPGRADE

!Onboarding now defaults tools.profile to messaging for new local installs; new setups no longer start with broad coding/system tools unless explicitly configured.
!ACP dispatch now defaults to enabled; set acp.dispatch.enabled=false to disable if you need to pause ACP turn routing while keeping /acp controls.
!Plugin SDK removed api.registerHttpHandler(...); plugins must use api.registerHttpRoute({ path, auth, match, handler }) and registerPluginHttpRoute(...) for dynamic webhook lifecycles.
!Zalo Personal plugin (@openclaw/zalouser) no longer depends on external zca-compatible CLI binaries (openzca, zca-cli); run openclaw channels login --channel zalouser after upgrade to refresh sessions.

Was this useful?

PydanticAI

v1.65.0 NOTES

PydanticAI v1.65.0 adds provider-uploaded file support via UploadedFile and the gemini-3.1-flash-lite-preview model.

└──▷ GET THIS VERSION

$ git clone --branch v1.65.0 https://github.com/pydantic/pydantic-ai.git
# already have the repo? check out this version:
$ git checkout v1.65.0

›Adds UploadedFile object to support files uploaded directly to providers, enabling agents to work with provider-hosted file references.
›Adds gemini-3.1-flash-lite-preview as a supported model identifier.

Was this useful?

// END OF TRANSMISSION

The AI Toolchain · curated daily · March 3, 2026
Releases & features from the open-source and commercial AI tools that practitioners actually run.

The AI Toolchain — issue -168, March 3, 2026