The AI Toolchain — issue -022, July 28, 2026

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

Cline v4.0.12 adds free model support with limit-reset indicators in the model picker.

$ git clone --branch v4.0.12 https://github.com/cline/cline.git
# already have the repo? check out this version:
$ git checkout v4.0.12

• ›Adds support for free Cline models, labeled '(free)' in the model picker, with a dedicated error card that displays the reset time when the free usage limit is reached.

$ git clone --branch cli-v3.0.47 https://github.com/cline/cline.git
# already have the repo? check out this version:
$ git checkout cli-v3.0.47

$ git clone --branch sdk/sdk/v0.0.66 https://github.com/cline/cline.git
# already have the repo? check out this version:
$ git checkout sdk/sdk/v0.0.66

$ git clone --branch desktop-v0.0.6 https://github.com/cline/cline.git
# already have the repo? check out this version:
$ git checkout desktop-v0.0.6

The open source coding agent.

OpenCode v1.18.9 adds an opt-in V2 desktop sidecar via OPENCODE_SIDECAR_V2=1 and collapsible model provider sections in V2 settings.

$ git clone --branch v1.18.9 https://github.com/anomalyco/opencode.git
# already have the repo? check out this version:
$ git checkout v1.18.9

$ OPENCODE_SIDECAR_V2=1 opencode

• ›Adds OPENCODE_SIDECAR_V2=1 environment variable to initialize the desktop app from the V2 background daemon instead of the V1 utility sidecar, with daemon state persisted across desktop restarts.
• ›Adds collapsible model provider sections in the V2 settings page, with collapse state persisted per server and search expanding matching sections temporarily without altering saved state.

$ git clone --branch v1.18.8 https://github.com/anomalyco/opencode.git
# already have the repo? check out this version:
$ git checkout v1.18.8

Qwen Code v0.21.1 adds qwen review comment-status, GitHub channel adapter, context_search tool, retry delay settings, and sandboxed /verify lane.

$ git clone --branch v0.21.1 https://github.com/QwenLM/qwen-code.git
# already have the repo? check out this version:
$ git checkout v0.21.1

$ qwen review comment-status

{
  "retryInitialDelayMs": 2000,
  "retryMaxDelayMs": 30000
}

$ qwen --max-journal-events 5000 --max-journal-bytes 4194304

• ›Introduces qwen review comment-status subcommand to quickly triage existing inline review comments and reduce API calls during review runs.
• ›Adds retryInitialDelayMs and retryMaxDelayMs settings to configure SSE stream rate-limit retry delays for provider quota windows.
• ›Adds a retrieval-only context_search tool for external corpora, configured by administrators without automatic recall or write capabilities.
• ›Adds a sandboxed /verify deep-verification lane on the triage CI path.
• ›Exposes workspace Channel management API on the serve layer, with a Channel management page and configuration flows in the Web Shell.

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

gemini-cli v0.53.0 adds an eval coverage report command and an LLM triage orchestrator.

$ git clone --branch v0.53.0 https://github.com/google-gemini/gemini-cli.git
# already have the repo? check out this version:
$ git checkout v0.53.0

• ›Adds an eval coverage report command (feat(evals)) to measure and report evaluation coverage.
• ›Implements an LLM triage orchestrator and container build for caretaker-triage workflows.

Framework for orchestrating role-playing, autonomous AI agents. By fostering collaborative intelligence, CrewAI empowers agents to work together seamlessly, tackling complex tasks.

CrewAI 1.15.8 adds WaitTool for pausing agent execution on long-running jobs.

$ git clone --branch 1.15.8 https://github.com/crewAIInc/crewAI.git
# already have the repo? check out this version:
$ git checkout 1.15.8

• ›Adds WaitTool to pause agent execution while waiting for long-running jobs to complete.

Build resilient agents.

LangGraph 1.2.10 adds trace_policy on add_node and typed v3 stream_events return with native projections.

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

graph.add_node('my_node', my_node_fn, trace_policy=TracePolicy(...))

• ›Exposes trace_policy parameter on add_node, letting you control tracing behavior per node when building graphs.
• ›Types the v3 stream_events return value and adds native projections, enabling strongly-typed streaming event handling.

PydanticAI v2.19.0 adds headers and retry_after fields to ModelHTTPError for richer HTTP error inspection.

$ git clone --branch v2.19.0 https://github.com/pydantic/pydantic-ai.git
# already have the repo? check out this version:
$ git checkout v2.19.0

import asyncio
from pydantic_ai.exceptions import ModelHTTPError

try:
    result = await agent.run('summarize this')
except ModelHTTPError as e:
    wait = e.retry_after  # seconds until the provider allows retry
    if wait:
        await asyncio.sleep(wait)
    # inspect raw response headers if needed
    print(e.headers)

• ›Adds headers and retry_after attributes to ModelHTTPError, populated from all provider SDKs, enabling programmatic inspection of rate-limit and retry signals from HTTP errors.

llama.cpp b10164 adds chunked SSD matrix multiplication on CUDA/HIP/MUSA to accelerate Mamba-2 prefill.

$ git clone --branch b10164 https://github.com/ggml-org/llama.cpp.git
# already have the repo? check out this version:
$ git checkout b10164

• ›Adds chunked SSD (State Space Duality) matmul CUDA kernel to accelerate Mamba-2 prefill, with support across CUDA, HIP, and MUSA backends.

$ git clone --branch b10159 https://github.com/ggml-org/llama.cpp.git
# already have the repo? check out this version:
$ git checkout b10159

Phoenix v19.10.0 adds experiment evaluation metrics charts, token sub-type breakdowns, and drag-to-reorder chart menus.

$ git clone --branch arize-phoenix-v19.10.0 https://github.com/Arize-ai/phoenix.git
# already have the repo? check out this version:
$ git checkout arize-phoenix-v19.10.0

• ›Adds experiment evaluation metrics charts to the experiments UI for visualizing eval results over runs.
• ›Breaks down token charts by sub-type (e.g. prompt, completion) and surfaces cost and token breakdown bars inside chart tooltips.
• ›Enables drag-to-reorder selected charts via the chart menu for custom dashboard layouts.
• ›Collapses experiment table rows to a single line for a more compact experiments view.
• ›Adds Monty provider with shared runtime to the sandbox environment.

$ git clone --branch arize-phoenix-v19.9.0 https://github.com/Arize-ai/phoenix.git
# already have the repo? check out this version:
$ git checkout arize-phoenix-v19.9.0

LanceDB python-v0.36.0 adds elastic dataloaders, OpenTelemetry metrics, WatsonxReranker, FTS tokenization, Tencent COS/GooseFS support, and remote branch diff/merge APIs.

$ git clone --branch python-v0.36.0 https://github.com/lancedb/lancedb.git
# already have the repo? check out this version:
$ git checkout python-v0.36.0

import lancedb
from lancedb.rerankers import WatsonxReranker

db = lancedb.connect('~/.lancedb')
tbl = db.open_table('my_table')
reranker = WatsonxReranker()
results = tbl.search('network intrusion detection', query_type='hybrid').rerank(reranker=reranker).limit(5).to_pandas()

• ›Adds get_lsm_write_spec method to read the installed LSM write spec from a table.
• ›Exposes Lance metrics via OpenTelemetry in Python and Node for observability into query and ingestion performance.
• ›Adds remote branch diff and merge client APIs for version-controlled table workflows.
• ›Adds WatsonxReranker reranker component for IBM Watsonx-backed reranking in retrieval pipelines.
• ›Adds Tencent COS and GooseFS object store support via new feature flags.

$ git clone --branch v0.33.0 https://github.com/lancedb/lancedb.git
# already have the repo? check out this version:
$ git checkout v0.33.0

from lancedb.rerankers import WatsonxReranker
import lancedb

db = lancedb.connect("./my_db")
table = db.open_table("my_table")
reranker = WatsonxReranker()
results = (
    table.search("neural network", query_type="hybrid")
    .rerank(reranker=reranker)
    .limit(5)
    .to_list()
)