Backfill issue -188 — 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 Agent Frameworks

AutoGPT

autogpt-platform-beta-v0.6.47 NOTES

AutoGPT Platform adds ClamAV file scanning, video editing blocks, Claude Opus 4.6 + extended thinking, and AI SDK streaming integration.

└──▷ GET THIS VERSION

$ git clone --branch autogpt-platform-beta-v0.6.47 https://github.com/Significant-Gravitas/AutoGPT.git
# already have the repo? check out this version:
$ git checkout autogpt-platform-beta-v0.6.47

›Adds ClamAV scanning for local file paths, enabling antivirus checks within agent workflows.
›Adds video editing blocks for use in agent pipelines.
›Adds a Text Encode block for encoding text within agent workflows.
›Adds Claude Opus 4.6 model support.
›Enables extended thinking mode for Claude models.

+4 moreshow less

›Implements AI SDK integration with custom streaming response handling.
›Adds a ready-to-test prompt displayed after agent creation.
›Adds a progress indicator shown during agent generation.
›Improves the Linear Search Block.

Was this useful?

LangChain

langchain-core==1.2.10 NOTES

langchain-core 1.2.10 adds ContextOverflowError, token counting for tool schemas, and new model-profile fields.

└──▷ GET THIS VERSION

$ git clone --branch langchain-core==1.2.10 https://github.com/langchain-ai/langchain.git
# already have the repo? check out this version:
$ git checkout langchain-core==1.2.10

›Adds ContextOverflowError exception class, raised automatically by Anthropic and OpenAI integrations when a request exceeds the model's context window.
›Adds text_inputs and text_outputs fields to model profiles, expanding the model-profile specification.
›Extends count_tokens_approximately to include tokens from tool schemas in its count, giving more accurate estimates when tools are attached to a model call.

Was this useful?

LlamaIndex

v0.14.14 NOTES

LlamaIndex v0.14.14 adds TokenBudgetHandler, MCP discovery, Chonkie node parser, adaptive thinking in Bedrock, and more new integrations.

└──▷ GET THIS VERSION

$ git clone --branch v0.14.14 https://github.com/run-llama/llama_index.git
# already have the repo? check out this version:
$ git checkout v0.14.14

└──▷ USE IT

Pre-bind fixed parameters to an MCP tool so callers never need to supply them explicitly.

python

from llama_index.tools.mcp import McpToolSpec

tool_spec = McpToolSpec(
    server_url='http://localhost:8080',
    partial_params={'environment': 'production', 'tenant_id': 'acme'}
)
tools = tool_spec.to_tool_list()

›Adds TokenBudgetHandler to llama-index-core callbacks for LLM cost governance.
›Adds partial_params support to McpToolSpec in llama-index-tools-mcp, allowing default parameter values to be pre-bound to MCP tool calls.
›New llama-index-tools-mcp-discovery integration package for MCP server discovery.
›New llama-index-node-parser-chonkie integration package adding Chonkie as a node parser.
›Adds custom base_url support to the Cohere LLM integration (llama-index-llms-cohere).

+11 moreshow less

›Adds support for adaptive thinking in llama-index-llms-bedrock-converse.
›Adds support for Claude Opus 4.6 in llama-index-llms-bedrock-converse and llama-index-llms-anthropic.
›Adds support for gpt-5.2-chat model in llama-index-llms-openai.
›Adds new reasoning types in llama-index-llms-openai.
›Adds OpenAI-like server mode for VllmServer in llama-index-llms-vllm.
›Adds event and memory record deletion methods to llama-index-memory-bedrock-agentcore.
›Adds Sharepoint page support events to llama-index-readers-microsoft-sharepoint.
›Adds new solar-pro3 model support to llama-index-llms-upstage.
›New llama-index-tools-moss integration package adding Moss search engine as a tool.
›Adds LangChain 1.x support across llama-index-core, llama-index-llms-langchain, and llama-index-readers-obsidian.
›Makes transformers an optional dependency in llama-index-llms-openai-like and llama-index-llms-openrouter.

Was this useful?

OpenAI Agents SDK

v0.8.3 NOTES

Realtime agents SDK gains model_version param for turn detection control.

└──▷ GET THIS VERSION

$ git clone --branch v0.8.3 https://github.com/openai/openai-agents-python.git
# already have the repo? check out this version:
$ git checkout v0.8.3

›Adds model_version parameter to turn detection configuration in the realtime agents SDK, allowing selection of the turn-detection model version.

Was this useful?

Microsoft Semantic Kernel

python-1.39.4 NOTES

Semantic Kernel Python 1.39.4 refines filtering behavior.

└──▷ GET THIS VERSION

$ git clone --branch python-1.39.4 https://github.com/microsoft/semantic-kernel.git
# already have the repo? check out this version:
$ git checkout python-1.39.4

›Refines filtering behavior in the Python SDK.

Was this useful?

camel-ai

v0.2.86 NOTES

camel-ai v0.2.86 adds Kimi K2.5 and Claude Opus 4.6 support, prompt caching, browser visual pixel mode, ChatAgent skills, Jina reranking, and file upload/download tooling.

└──▷ GET THIS VERSION

$ git clone --branch v0.2.86 https://github.com/camel-ai/camel.git
# already have the repo? check out this version:
$ git checkout v0.2.86

›Adds support for kimi k2.5 as a model backend.
›Adds support for claude-opus-4-6 as a model backend.
›Adds prompt caching support for compatible model backends.
›Adds browser visual pixel mode to hybrid_browser_toolkit, enabling pixel-level visual interaction with web pages.
›Adds file upload and download tool for agent workflows.

+4 moreshow less

›Adds ChatAgent skills system, including a skill creator with scripts and reference documentation.
›Adds Jina reranking integration for retrieval pipelines.
›Adds typed parameter definitions and supported format declarations to EarthScienceToolkit, and expands supported formats in MarkItDownLoader.
›Unifies cleanup handling across all runtimes for consistent teardown behavior.

Was this useful?

◆ AI Coding Agents

SST OpenCode

v1.1.55 2 RELEASES · 2026-02-10 NOTES STABLE

The open source coding agent.

OpenCode v1.1.55 shows a helpful message when the free usage limit is exceeded.

└──▷ GET THIS VERSION

$ git clone --branch v1.1.55 https://github.com/sst/opencode.git
# already have the repo? check out this version:
$ git checkout v1.1.55

›Displays a helpful message when the free usage limit is exceeded, guiding users on next steps.

1 more release in this issue · 2026-02-10

v1.1.54 NOTES STABLE

OpenCode v1.1.54 adds --fork flag, skill discovery via well-known RFC, Wayland support, Cmd+[/] history nav, and drag-and-drop file mentions.

└──▷ GET THIS VERSION

$ git clone --branch v1.1.54 https://github.com/sst/opencode.git
# already have the repo? check out this version:
$ git checkout v1.1.54

└──▷ USE IT

Enable model autocomplete in opencode.json by referencing the models.dev schema.

json

{
  "$schema": "https://models.dev/opencode.json"
}

›Adds Claude Code-style --fork flag to the TUI to duplicate a session before continuing, preserving the original.
›Adds skill discovery from URLs via well-known RFC for automatic capability detection.
›Adds native Wayland toggle on Linux desktop.
›Adds Cmd+[ / Cmd+] keybinds in the desktop app for session history navigation.
›Adds drag-and-drop support for @mentioning files in the desktop app.

+17 moreshow less

›Adds models.dev schema reference for model autocomplete in opencode.json.
›Adds directory parameter to the plugin client for multi-project support.
›Adds native clipboard image paste support in the desktop app.
›Adds new ContextOverflowError type for detecting context-limit conditions.
›Enables thinking for all reasoning models on Alibaba Cloud (DashScope).
›Adds specific system prompt for the Trinity model.
›Adds session.error event publishing for invalid model selection.
›Adds session last updated time display in the command palette search.
›Adds macOS support for displaying only installed editors and adds Sublime Text as a supported editor.
›Adds sidecar server URL persistence via isSidecar prop in the desktop application.
›Adds keyboard accelerators to desktop menu bar items.
›Parallelize skill downloads for faster loading in the TUI.
›Adds toast notification when a session is missing on prompt submit.
›Maximizes the main desktop window by default on launch.
›Adds loading window and restructures Rust desktop code.
›Adds Windows File Explorer icon for the session header.
›Adds default clipboard copy affordance to the TextField component.

Was this useful?

Google gemini-cli

v0.28.0 NOTES

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

gemini-cli v0.28.0 adds /prompt-suggest, background shell commands, auto theme switching, and policy deny messages.

└──▷ GET THIS VERSION

$ git clone --branch v0.28.0 https://github.com/google-gemini/gemini-cli.git
# already have the repo? check out this version:
$ git checkout v0.28.0

└──▷ USE IT

Restrict a shared environment to only load approved extensions, reducing supply-chain risk from arbitrary extension installs.

json

# In settings.json
{
  "allowedExtensions": ["my-org/approved-ext-a", "my-org/approved-ext-b"]
}

Get prompt ideas mid-session when you're unsure how to phrase a security analysis request.

$ /prompt-suggest

›Adds /prompt-suggest slash command to surface prompt suggestions interactively.
›Supports background shell command execution.
›Implements automatic theme switching based on terminal background color.
›Adds allowedExtensions setting to restrict which extensions can be loaded.
›Displays user identity (auth method, email, tier) on startup.

+19 moreshow less

›Adds user identity info (auth method, email, tier) to the /stats command output.
›Adds support for custom deny messages in policy rules.
›Adds GOOGLE_GENAI_API_VERSION environment variable to override the Gemini API version.
›Enables overriding CODE_ASSIST_API_VERSION with an environment variable.
›Enables activity logging for non-interactive mode and evals.
›Adds a setting to disable loop detection.
›Implements dynamic policy registration for subagents.
›Implements tool name aliasing for backward compatibility.
›Adds session resume support for ACP.
›Shows hooksConfig.enabled status in the settings dialog.
›Adds support for custom themes in extensions.
›Adds support for Positron IDE.
›Adds multimodal tool response support to token calculation.
›Preserves end-of-line style in files when making edits.
›Updates undo/redo keybindings to Cmd+Z / Alt+Z and Shift+Cmd+Z / Shift+Alt+Z.
›Aligns hooks enable/disable controls with skills and improves tab completion.
›Provides actionable error messages when admin-disabled features are accessed.
›Implements interactive and non-interactive consent flows for OAuth.
›Includes auth method in /bug reports.

└──▷ BREAKING ON UPGRADE

!The legacy tools.enableHooks setting is removed; hooks configuration must use the new hooksConfig.enabled setting.

Was this useful?

◆ Local LLM Runtimes

llama.cpp

b7990 2 RELEASES · 2026-02-10 NOTES STABLE

llama.cpp b7990 adds support for the Qwen3.5 model series with FULL_ATTENTION_INTERVAL metadata.

└──▷ GET THIS VERSION

$ git clone --branch b7990 https://github.com/ggml-org/llama.cpp.git
# already have the repo? check out this version:
$ git checkout b7990

›Adds FULL_ATTENTION_INTERVAL metadata key to support the Qwen3.5 series hybrid attention architecture.
›Supports loading and running Qwen3.5 series models.

1 more release in this issue · 2026-02-10

b7983 NOTES STABLE

llama.cpp b7983 adds quantized MoE matrix multiplication for the CANN backend, supporting Q4_0 and Q8_0 weight formats.

└──▷ GET THIS VERSION

$ git clone --branch b7983 https://github.com/ggml-org/llama.cpp.git
# already have the repo? check out this version:
$ git checkout b7983

›Adds ggml_cann_mul_mat_id_quant to the CANN backend, enabling quantized matrix multiplication for Mixture of Experts (MoE) models with Q4_0 and Q8_0 weight formats, automatic F16 type conversion, and both per-expert and broadcast input modes.

Was this useful?

Other / Uncategorized

◆ MCP TOOLING

Composio

v0.11.1 NOTES

Composio v0.11.1 adds public URL support, type-safe get_tools(), typed webhook schemas, and a CLI env-var toolkit version override.

└──▷ GET THIS VERSION

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

›Adds env-var toolkit version override in the TypeScript CLI, letting operators pin or swap toolkit versions without code changes.
›Adds typed schema for connection-expired webhook events in the core, so handlers can pattern-match on composio.* event types with full type safety.
›Adds type-safe generic get_tools() in the Python SDK that infers return types based on the given provider.
›Enables public URLs in the Python SDK (feat(py): enable public URLs).
›Loosens the V3 webhook schema to accept any composio.* event type, broadening the range of subscribable webhook events.

Was this useful?

// END OF TRANSMISSION

The AI Toolchain · curated daily · February 10, 2026
Releases & features from the open-source and commercial AI tools that practitioners actually run.

The AI Toolchain — issue -188, February 10, 2026