Backfill issue -200 — 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   # 10 tools matched

AI & LLM Tooling

◆ AI Coding Agents

Anthropic Claude Code

v2.1.23 NOTES

Claude Code is an agentic coding tool that lives in your terminal, understands your codebase, and helps you code faster by executing routine tasks, explaining complex code, and handling git workflows - all through natural language commands.

Claude Code v2.1.23 adds customizable spinner verbs and shows PR merge status and timeout durations in the UI.

└──▷ GET THIS VERSION

$ git clone --branch v2.1.23 https://github.com/anthropics/claude-code.git
# already have the repo? check out this version:
$ git checkout v2.1.23

›Adds spinnerVerbs setting to customize the verbs displayed in the activity spinner.
›Shows timeout duration alongside elapsed time for Bash commands.
›Displays a purple status indicator in the prompt footer for merged pull requests.

Was this useful?

GitHub Copilot CLI

v0.0.399 NOTES

GitHub Copilot CLI v0.0.399 adds /allow-all, /yolo, /diff, and .claude/commands/ single-file command support alongside AI-named sessions.

└──▷ GET THIS VERSION

$ git clone --branch v0.0.399 https://github.com/github/copilot-cli.git
# already have the repo? check out this version:
$ git checkout v0.0.399

└──▷ TRY IT

Quickly approve all permission prompts for an entire session without repeated confirmations — useful when running trusted, wide-ranging automation.

$ /allow-all

Drop a single Markdown file into .claude/commands/ to expose a lightweight slash command without building a full skill.

$ echo '# summarize\nSummarize the current diff in one paragraph.' > .claude/commands/summarize.md

Inspect all file changes made during the current session before committing.

$ /diff

›Adds /allow-all and /yolo slash commands to auto-approve all permissions for the duration of a session.
›Supports .claude/commands/ single-file commands as a simpler alternative to skills.
›Adds /diff command to review session changes with visual indicators and scroll acceleration.
›Adds Ctrl+X then / keyboard shortcut to invoke slash commands without losing current input.
›Extends /usage to include token consumption from sub-agents (e.g., the general-purpose agent).

+5 moreshow less

›Adds double-Esc to undo/rewind to previous session states.
›Adds LSP (Language Server Protocol) tool for code intelligence (requires staff flag).
›Sessions now receive AI-generated names derived from the first message.
›Adds Copilot option in the agent creation wizard to auto-generate agent name, description, and instructions from an initial description.
›Skills remain effective after conversation history is compacted.

Was this useful?

Block Goose

v1.22.0 NOTES

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

Goose v1.22.0 adds session forking, BERT-based command injection detection, container extensions, and an OSS skills marketplace.

└──▷ GET THIS VERSION

$ git clone --branch v1.22.0 https://github.com/block/goose.git
# already have the repo? check out this version:
$ git checkout v1.22.0

└──▷ USE IT

Cap how many turns a recipe's subagent is allowed to run, preventing runaway automation.

yaml

# In your recipe YAML
max_turns: 10

Authenticate to AWS Bedrock using a bearer token instead of standard AWS credentials.

$ export AWS_BEARER_TOKEN_BEDROCK=<your-token>
goose run --provider bedrock --model anthropic.claude-3-5-sonnet-20241022-v2:0

›Adds session forking to branch existing conversations into independent sessions.
›Introduces BERT ML-based command injection detection in tool calls for prompt injection defense.
›Supports running extensions inside containers via new Container Extensions option.
›Adds an open-source skills marketplace for discovering and adding community skills.
›Integrates OVHcloud AI as a new model provider.

+15 moreshow less

›Adds max_turns setting to recipes and subagents to cap agent execution depth.
›Supports requires_auth flag for configuring custom providers that need no authentication.
›Adds AWS_BEARER_TOKEN_BEDROCK environment variable support for AWS Bedrock authentication.
›Supports custom headers with the Anthropic provider.
›Enables external editor support (vim, helix, etc.) for composing prompts.
›Adds a built-in Goose Documentation Skill for in-session reference to Goose docs.
›Adds a Diagnostics Viewer for debugging Goose sessions.
›Makes skills available in codemode.
›Adds PR code review functionality.
›Adds Vibe MCP Apps, a new MCP apps integration.
›Adds collapsible chat navigation for a cleaner desktop UI.
›Adds new keyboard shortcuts.
›Adds reduced motion support for accessibility preferences.
›Creates a default gooseignore file automatically when one is missing.
›Adds Codex subscription support as a provider option.

Was this useful?

SST OpenCode

v1.1.42 2 RELEASES · 2026-01-29 NOTES STABLE

The open source coding agent.

OpenCode v1.1.42 adds sequential numbering for forked session titles and cache token stats in CLI output.

└──▷ GET THIS VERSION

$ git clone --branch v1.1.42 https://github.com/sst/opencode.git
# already have the repo? check out this version:
$ git checkout v1.1.42

›Adds sequential numbering for forked session titles to distinguish multiple forks of the same session.
›Includes cache tokens in CLI statistics output.

1 more release in this issue · 2026-01-29

v1.1.41 NOTES STABLE

OpenCode v1.1.41 adds ctx.abort to the grep tool and AbortSignal support to Ripgrep.files() and GlobTool.

└──▷ GET THIS VERSION

$ git clone --branch v1.1.41 https://github.com/sst/opencode.git
# already have the repo? check out this version:
$ git checkout v1.1.41

›Adds ctx.abort to the grep tool, enabling cooperative cancellation of grep operations.
›Adds AbortSignal support to Ripgrep.files() and GlobTool, allowing callers to cancel file-search operations mid-flight.
›Supports Copilot model variants with maxTokens set, enabling use of previously broken model configurations.

Was this useful?

Earendil Works Pi

v0.50.3 2 RELEASES · 2026-01-29 NOTES STABLE

AI agent toolkit: unified LLM API, agent loop, TUI, coding agent CLI

Pi v0.50.3 adds Kimi For Coding (Moonshot AI) as a new provider via KIMI_API_KEY.

└──▷ GET THIS VERSION

$ git clone --branch v0.50.3 https://github.com/earendil-works/pi.git
# already have the repo? check out this version:
$ git checkout v0.50.3

›Adds Kimi For Coding provider, enabling access to Moonshot AI's Anthropic-compatible coding API via the KIMI_API_KEY environment variable.

1 more release in this issue · 2026-01-29

v0.50.2 NOTES STABLE

Pi v0.50.2 adds Hugging Face provider, extended prompt caching, /files command, shell keybindings, and RPC get_commands.

└──▷ GET THIS VERSION

$ git clone --branch v0.50.2 https://github.com/earendil-works/pi.git
# already have the repo? check out this version:
$ git checkout v0.50.2

└──▷ TRY IT

Use a Hugging Face model in place of another provider without changing your workflow.

$ HF_TOKEN=hf_xxxxxxxxxxxx pi

Enable 24-hour prompt cache retention for OpenAI to reduce token costs on long, repeated context.

$ PI_CACHE_RETENTION=long pi

Audit every file the AI has touched in the current session before committing changes.

$ /files

›Adds Hugging Face provider via OpenAI-compatible Inference Router — set HF_TOKEN to access HF models.
›New PI_CACHE_RETENTION=long env var enables 1-hour prompt caching for Anthropic and 24-hour for OpenAI, vs. short in-memory defaults.
›New /files command lists all file operations (read, write, edit) performed in the current session.
›Adds shell-style keybindings: alt+b/alt+f for word navigation and ctrl+d for forward character delete.
›New autocompleteMaxVisible setting (3–20 items, default 5) controls autocomplete dropdown height via /settings or settings.json.

+1 moreshow less

›New get_commands RPC method lets headless clients programmatically list available commands.

Was this useful?

Alibaba Qwen Code

v0.9.0-preview.0 3 RELEASES · 2026-01-29 NOTES STABLE

Qwen Code v0.9.0 adds multi-modal input, concurrent batch CLI, extra_body for OpenAI providers, LSP support, and a Zed extension.

└──▷ GET THIS VERSION

$ git clone --branch v0.9.0-preview.0 https://github.com/QwenLM/qwen-code.git
# already have the repo? check out this version:
$ git checkout v0.9.0-preview.0

›Adds extra_body support for OpenAI-compatible providers, enabling pass-through of provider-specific request fields.
›Adds concurrent runner for batch CLI execution, enabling parallel processing of multiple CLI tasks.
›Adds multi-modal input support (image, PDF, audio) across all content generators.
›Adds parentToolCallId and subagentType fields for ACP subagent tracking.
›Adds experimental LSP (Language Server Protocol) support for code intelligence.

+4 moreshow less

›Adds Zed extension for Qwen Code agent server, enabling Zed IDE integration.
›Adds model selection support through ACP in the VS Code IDE companion.
›Improves retry logic for 429 and 5xx error handling in the core.
›Clarifies output formats for non-interactive mode.

2 more releases in this issue · 2026-01-29

v0.8.2-preview.2 NOTES STABLE

Qwen Code v0.8.2-preview.2 adds multi-modal input, concurrent batch execution, LSP support, Zed extension, and extra_body for OpenAI-compatible providers.

└──▷ GET THIS VERSION

$ git clone --branch v0.8.2-preview.2 https://github.com/QwenLM/qwen-code.git
# already have the repo? check out this version:
$ git checkout v0.8.2-preview.2

›Adds extra_body support for OpenAI-compatible providers, enabling pass-through of provider-specific request fields.
›Adds a concurrent runner for batch CLI execution, enabling parallelized non-interactive workloads.
›Adds multi-modal input support (image, PDF, audio) across all content generators.
›Adds experimental LSP (Language Server Protocol) support for code intelligence.
›Adds a Zed extension for the Qwen Code agent server, bringing IDE companion support to Zed.

+4 moreshow less

›Adds parentToolCallId and subagentType fields for ACP subagent tracking.
›Clarifies output formats for non-interactive mode.
›Supports model selection through ACP in the VS Code IDE companion.
›Improves retry logic for 429 and 5xx error handling.

v0.8.2-preview.1 NOTES STABLE

Qwen Code v0.8.2-preview.1 adds subagent tracking fields, improved error messages for skills, and smarter 429/5xx retry logic.

└──▷ GET THIS VERSION

$ git clone --branch v0.8.2-preview.1 https://github.com/QwenLM/qwen-code.git
# already have the repo? check out this version:
$ git checkout v0.8.2-preview.1

›Adds parentToolCallId and subagentType fields to ACP subagent tracking, enabling identification and lineage of subagent calls.
›Improves error messaging when a skill is incorrectly invoked as a tool in core, making misconfiguration easier to diagnose.
›Improves retry logic for 429 (rate-limit) and 5xx (server error) responses in core, reducing failed requests under load.

Was this useful?

◆ AI Agent Frameworks

PydanticAI

v1.50.0 2 RELEASES · 2026-01-29 NOTES STABLE

PydanticAI v1.50.0 exposes usage limits and model settings to CLI users and adds OpenAI raw text annotation access.

└──▷ GET THIS VERSION

$ git clone --branch v1.50.0 https://github.com/pydantic/pydantic-ai.git
# already have the repo? check out this version:
$ git checkout v1.50.0

›Adds usage_limits and model_settings parameters accessible to users running agents with to_cli(), enabling runtime control of limits and settings from the command line.
›Adds a setting to include OpenAI raw text annotations in TextPart.provider_details, surfacing provider-level annotation data to callers.

1 more release in this issue · 2026-01-29

v1.49.0 NOTES STABLE

PydanticAI v1.49.0 adds BedrockEmbeddingModel for Nova/Cohere/Titan and parallel tool calls in DBOSAgent.

└──▷ GET THIS VERSION

$ git clone --branch v1.49.0 https://github.com/pydantic/pydantic-ai.git
# already have the repo? check out this version:
$ git checkout v1.49.0

└──▷ USE IT

Generate embeddings via AWS Bedrock's Nova, Cohere, or Titan endpoints using the new BedrockEmbeddingModel.

python

from pydantic_ai.models.bedrock import BedrockEmbeddingModel

model = BedrockEmbeddingModel('amazon.nova-lite-v1')
result = await model.embed(['Hello, world!'])

›Adds BedrockEmbeddingModel class supporting AWS Bedrock embedding endpoints for Nova, Cohere, and Titan models.
›Enables parallel tool call execution in DBOSAgent.
›Updates Vercel AI SDK type definitions to match AI SDK v6.

Was this useful?

◆ Local LLM Runtimes

llama.cpp

b7876 2 RELEASES · 2026-01-29 NOTES STABLE

llama.cpp b7876 enables Hexagon HTP offloading on Windows on Snapdragon (WoS) devices.

└──▷ GET THIS VERSION

$ git clone --branch b7876 https://github.com/ggml-org/llama.cpp.git
# already have the repo? check out this version:
$ git checkout b7876

›Enables GPU offloading to the Qualcomm Hexagon HTP (via the ggml-backend-dl dynamic backend) on Windows on Snapdragon (WoS) devices, alongside existing Android support.
›Adds run-bench.ps1 and run-tool.ps1 PowerShell scripts to simplify benchmarking and tool execution in the WoS/Hexagon build environment.
›Renames the backend documentation directory from docs/backend/hexagon to docs/backends/snapdragon, consolidating Android and Windows build guides into a unified README.

└──▷ BREAKING ON UPGRADE

!The backend documentation directory is renamed from docs/backend/hexagon to docs/backends/snapdragon; any scripts or CI referencing the old path will break.

1 more release in this issue · 2026-01-29

b7870 NOTES STABLE

llama-batched-bench gains the -kvu flag for additional benchmarking control.

└──▷ GET THIS VERSION

$ git clone --branch b7870 https://github.com/ggml-org/llama.cpp.git
# already have the repo? check out this version:
$ git checkout b7870

›Adds -kvu argument to llama-batched-bench for expanded benchmark configuration.

Was this useful?

vLLM

v0.15.0 NOTES

vLLM v0.15.0 adds Kimi-K2.5/Molmo2/Step3vl models, Mamba prefix caching, async scheduling with PP, FIPS 140-3 hashing, and new API fields.

└──▷ GET THIS VERSION

$ git clone --branch v0.15.0 https://github.com/vllm-project/vllm.git
# already have the repo? check out this version:
$ git checkout v0.15.0

└──▷ TRY IT

Enable Mamba prefix caching to get ~2x speedup on hybrid/Mamba models by caching Mamba states across requests.

$ vllm serve <model> --enable-prefix-caching --mamba-cache-mode align

Run pipeline-parallel serving with async scheduling enabled, previously a blocked combination.

$ vllm serve <model> --pipeline-parallel-size 4 --async-scheduling

Restrict TLS cipher suites on a vLLM API server for hardened enterprise deployments.

$ vllm serve <model> --ssl-keyfile key.pem --ssl-certfile cert.pem --ssl-ciphers 'TLS_AES_256_GCM_SHA384:TLS_CHACHA20_POLY1305_SHA256'

›Enables --async-scheduling with pipeline parallelism, previously unsupported.
›Adds Mamba/hybrid model prefix caching via --enable-prefix-caching --mamba-cache-mode align, achieving ~2x speedup by caching Mamba states directly.
›Adds --ssl-ciphers CLI argument for TLS cipher suite control.
›Adds skip_special_tokens configuration to the OpenAI-compatible API.
›Adds prompt_cache_key support to the Responses API.

+26 moreshow less

›Adds include_stop_str_in_output tuning parameter to the Responses API.
›Adds data_1/data_2 and queries/documents flexible input formats to the score endpoint.
›Adds new render endpoints for prompt preprocessing.
›Adds avg_logprob and compression_ratio fields to Whisper API verbose_json segment output.
›Adds FIPS 140-3 compliant hash option for enterprise/government deployments.
›Auto-detects api_server_count based on dp_size to simplify multi-process server configuration.
›Adds support for new model architectures: Kimi-K2.5, Molmo2, Step3vl 10B, Step1, GLM-Lite, and Eagle2.5-8B VLM.
›Extends LoRA support to Nemotron-H, InternVL2, and MiniMax M2 models.
›Adds EAGLE3 speculative decoding for Pixtral/LlavaForConditionalGeneration and Qwen3 VL MoE.
›Adds BGE-M3 sparse embeddings and ColBERT embeddings support.
›Adds VLM support to Model Runner V2.
›Adds inplace LoRA loading for improved memory efficiency.
›Adds torch.compile inductor AOT artifact support.
›Adds torch.compile support for Whisper.
›Adds MXFP4 W4A16 quantization support for compressed-tensors MoE models.
›Adds quantization support for non-gated MoE with Marlin, NVFP4 CUTLASS, FP8, INT8, and compressed-tensors.
›Adds Intel Quantization Toolkit integration.
›Adds FP8 KV cache per-tensor and per-attention-head quantization via llmcompressor.
›Adds session-based streaming input accepting async generators producing StreamingInput objects for interactive workloads like ASR.
›Sets FlashInfer MLA as the default MLA backend on Blackwell, with TRTLLM as default prefill.
›Adds MoRI EP high-performance all2all backend for Expert Parallel on AMD ROCm.
›Adds Flash Attention Triton backend support for AMD RDNA3/RDNA4 consumer GPUs.
›Adds TPU pipeline parallelism support.
›Adds NUMA-aware acceleration for TP/DP inference on ARM CPU.
›Adds wheel variant auto-detection during install.
›Adds custom profiler URI scheme support.

└──▷ BREAKING ON UPGRADE

!The vllm:time_per_output_token_seconds metric has been removed — use vllm:inter_token_latency_seconds instead.
!Several deprecated environment variables have been removed.
!DeepSpeedFp8 quantization has been removed.
!RTN quantization has been removed.
!HQQ quantization is deprecated.

Was this useful?

Other / Uncategorized

◆ MCP TOOLING

Composio

v0.11.0 NOTES

Composio v0.11.0 adds Mastra v1 support, async webhook verification, redirect_url in connected accounts, and platform-specific file tool modifiers.

└──▷ GET THIS VERSION

$ git clone --branch v0.11.0 https://github.com/ComposioHQ/composio.git
# already have the repo? check out this version:
$ git checkout v0.11.0

└──▷ USE IT

Await the now-async webhook verification method after upgrading to avoid silent failures in trigger handlers.

typescript

const isValid = await composio.triggers.verifyWebhook(request);

Pass a redirect_url when initiating a connected account so users are sent back to your app after OAuth completes.

typescript

const connection = await composio.connectedAccounts.initiate({ appName: 'github', redirectUrl: 'https://yourapp.com/callback' });

›Adds redirect_url support in connectedAccounts for the TypeScript core SDK, enabling post-auth redirects in OAuth flows.
›Introduces composio.triggers.verifyWebhook as an async method in the TypeScript SDK, removing the node:crypto dependency from triggers for broader runtime compatibility.
›Adds a platform-specific file tool modifier in the TypeScript core SDK, allowing file tools to be adapted per deployment target.
›Adds support for Mastra v1 in the TypeScript SDK, including e2e tests with Tool Router.
›Optimizes @composio/client bandwidth usage in the CLI and introduces debug metrics.

+3 moreshow less

›Defaults tool selection to 'important tools' in the TypeScript core SDK when no explicit tool list is provided.
›Adds secrets detection workflow for pull requests to the CI pipeline.
›Adds an experimental assistive prompt to the client.

└──▷ BREAKING ON UPGRADE

!composio.triggers.verifyWebhook is now async in the TypeScript SDK — callers must await it or their webhook verification will silently break.
!Mastra v1 support is introduced as a breaking change in the TypeScript SDK — existing Mastra integrations targeting earlier versions may require updates.

Was this useful?

// END OF TRANSMISSION

The AI Toolchain · curated daily · January 29, 2026
Releases & features from the open-source and commercial AI tools that practitioners actually run.

The AI Toolchain — issue -200, January 29, 2026