Heads up This site is currently under heavy development.
Subscribe Get it delivered — the daily firehose, filtered to the tools you run, plus the documentation changes vendors never announce. Compare plans →

The AI Toolchain — issue -040, July 10, 2026

THE AI TOOLCHAIN NO. -040
Tail
THE DAILY RELEASE FIREHOSE
PUBLISHED JULY 10, 2026 · EVERY WEEKDAY
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   # 15 tools matched
AI & LLM Tooling
◆  AI Coding Agents

Anthropic Claude Code

Sources Release notes → v2.1.206 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 v2.1.206 adds directory suggestions for /cd, a /doctor CLAUDE.md trimmer, smarter git push remotes, and faster background agent upgrades.

└──▷ GET THIS VERSION
$ git clone --branch v2.1.206 https://github.com/anthropics/claude-code.git
# already have the repo? check out this version:
$ git checkout v2.1.206
└──▷ TRY IT
Quickly identify and slim down a bloated CLAUDE.md before committing it to the repo.
$ /doctor
Navigate to a subdirectory mid-session with autocomplete suggestions, the same way /add-dir works.
$ /cd src/
  • Adds directory path suggestions to /cd, matching the autocomplete behavior already available in /add-dir.
  • New /doctor check analyzes checked-in CLAUDE.md files and proposes trimming content Claude can derive directly from the codebase.
  • /commit-push-pr now auto-allows git push to the repo's configured remote.pushDefault (or the sole remote when only one exists), not just origin.
  • /login now supports Anthropic-operated public gateway endpoints.
  • EnterWorktree prompts for confirmation before entering a git worktree outside the project's .claude/worktrees/ directory.
+3 moreshow less
  • Background agents upgrade to a new Claude Code version immediately after an update, eliminating the slow stale-session upgrade cost on attach.
  • Agents view status column now uses full terminal width instead of truncating at 64 characters.
  • Ctrl+X in the agents view now permanently removes a completed session; deleted background jobs stay deleted across renders.
Was this useful?

Charm Crush

Sources Release notes → v0.84.0 NOTES

Glamourous agentic coding for all

Crush v0.84.0 adds session-affinity headers so supporting providers can pin cache to a replica.

└──▷ GET THIS VERSION
$ git clone --branch v0.84.0 https://github.com/charmbracelet/crush.git
# already have the repo? check out this version:
$ git checkout v0.84.0
  • Sends x-session-affinity and x-session-id headers (a PII-free session hash) to enable cache-pinning on supporting providers.
Was this useful?

GitHub Copilot CLI

Sources Release notes → v1.0.70 NOTES

Copilot CLI v1.0.70 adds GPT-5.6, sandbox session flags, /refine prompt rewriting, repo-pinned model policy, and paginated MCP resource RPCs.

└──▷ GET THIS VERSION
$ git clone --branch v1.0.70 https://github.com/github/copilot-cli.git
# already have the repo? check out this version:
$ git checkout v1.0.70
└──▷ TRY IT
Temporarily disable the sandbox for a single session (e.g. to run a privileged setup script) without overwriting your saved sandbox preference.
$ copilot -p --no-sandbox
Pin a repository's Copilot model, effort level, and MCP deny list so all contributors share the same policy automatically.
json
{
  "model": "gpt-5.6",
  "effort": "high",
  "contextTier": "full",
  "deny": {
    "mcp": ["untrusted-server"]
  }
}
Pin a marketplace plugin to a specific commit so your team always runs a known-good version.
yaml
plugins:
  - name: my-plugin
    source:
      url: https://github.com/org/my-plugin
      sha: a3f8c21d9e4b7065f1234567890abcdef1234567
  • Adds --sandbox and --no-sandbox flags to toggle the OS-level shell sandbox on or off for the current session only, without persisting the change to saved settings (composable with -p).
  • Adds sha field in plugin source configuration to pin plugins to an exact commit SHA.
  • Adds --repo and --local flags to /settings and /model commands.
  • Adds paginated session.mcp.resources RPCs — read, list, and listTemplates — for MCP server resources.
  • Lets a trusted repository pin the model, effort level, and context tier and extend URL/MCP/skill deny lists via .github/copilot/settings.json.
+14 moreshow less
  • Adds /refine command to rewrite a rough, stream-of-consciousness prompt into a clear one.
  • Adds GPT-5.6 model support.
  • Adds preToolUse hooks that exit with code 2 to deny tool calls.
  • Marks locally-spawned MCP servers running inside the sandbox in /mcp list output (e.g. connected (sandboxed)).
  • Exposes SDK APIs to manage live MCP servers in running sessions.
  • Enables web_fetch to work through mandatory HTTPS proxies.
  • Adds hide and search functionality on the Gists tab.
  • Adds a setting to show or hide timeline timestamps.
  • Enables Forge to create draft skills when it finds a clear workflow pattern.
  • Opens the plan file or research report with Ctrl+Y in any mode.
  • Long-running sessions now refresh enterprise managed settings hourly.
  • Makes markdown links and bare URLs in the timeline and tool output clickable.
  • Adds Tab key to switch the context window in /model.
  • Prefills /chronicle search so it can accept a query directly.
Was this useful?

All Hands AI OpenHands

Sources Release notes → cloud-1.46.0 NOTES

OpenHands: AI-Driven Development

OpenHands 1.46 adds configurable BYOR key alias patterns and enriched final archive manifests.

└──▷ GET THIS VERSION
$ git clone --branch cloud-1.46.0 https://github.com/All-Hands-AI/OpenHands.git
# already have the repo? check out this version:
$ git checkout cloud-1.46.0
  • Enables configurable key alias patterns for Bring Your Own Router (BYOR) keys in enterprise deployments.
  • Enriches final archive manifests with additional metadata and removes initial snapshots to streamline workspace archives.
Was this useful?

Alibaba Qwen Code

Sources Release notes → v0.19.9 NOTES

Qwen Code v0.19.9 adds qwen update, webhook-triggered channel tasks, PDF-to-image fallback, cursor-paged transcript replay, and multi-workspace sidebar.

└──▷ GET THIS VERSION
$ git clone --branch v0.19.9 https://github.com/QwenLM/qwen-code.git
# already have the repo? check out this version:
$ git checkout v0.19.9
└──▷ TRY IT
Keep the Qwen Code CLI up to date without leaving the terminal.
$ qwen update
  • Adds qwen update and /update commands with auto-update support, letting users self-update the CLI in place.
  • Adds create_sub_session tool for scheduled tasks, enabling isolated run mode with precondition gating.
  • Adds serve --channel channel worker settings reload, so channel configuration changes take effect without restarting the server.
  • Adds workspace-qualified core REST routes, exposing per-workspace API surfaces.
  • Adds cursor-paged transcript replay endpoint (serve) for streaming session history replay.
+22 moreshow less
  • Adds forceGlobalCacheScope config for Anthropic proxy providers, enabling cross-request prompt-cache sharing.
  • Adds parentSessionId recording and querying for sub-sessions, making sub-session lineage queryable from the daemon.
  • Adds voiceBridge support for ACP audio prompts.
  • Adds webhook-triggered channel tasks, allowing external webhooks to fire channel-side agent tasks.
  • Adds session owner index for workspace runtimes, improving multi-workspace session tracking.
  • Adds listing of persisted sessions for trusted workspaces and archived/organized sessions for non-primary workspaces.
  • Adds background memory agent timeout configuration (feat(memory)).
  • Adds Ctrl+O frozen transcript view and unified tool output rendering in the TUI.
  • Renders PDF pages as images when text extraction overflows or fails, enabling vision-model processing of dense PDFs.
  • Adds web-shell workspace management sidebar with dynamic registration (daemon multi-workspace phase 4).
  • Adds context mention customization to the web-shell.
  • Adds collapse/expand toggle to the web-shell AskUserQuestion panel.
  • Adds bottom status items and assistant turn footer slot to the web-shell.
  • Adds inline VP-mode thought expand on click with auto-hiding scrollbar.
  • Posts code-review Suggestion findings as inline PR comments.
  • Assigns an accountable reviewer to every line of a large diff in the review workflow.
  • Adds QQ bot group message handling and cron-msg-experimental support.
  • Adds session runtime status exposure from the daemon.
  • Allows long /goal conditions in the CLI.
  • Adds pure-ASCII fast path to text token estimation, improving throughput for ASCII-heavy workloads.
  • Gates browser automation MCP on an external adapter, making the integration opt-in.
  • Persists session artifacts across daemon restarts.
└──▷ BREAKING ON UPGRADE
  • !The feat(daemon) change drops isolated scheduled-task mode: sub-sessions now use parentSessionId tracking instead of the previous isolated run mode.
Was this useful?

Zed

Sources Release notes → v1.10.2 NOTES

Zed v1.10.2 adds GPT 5.6 Sol & Terra models for ChatGPT subscription users.

└──▷ GET THIS VERSION
$ git clone --branch v1.10.2 https://github.com/zed-industries/zed.git
# already have the repo? check out this version:
$ git checkout v1.10.2
  • Adds GPT 5.6 Sol and GPT 5.6 Terra model options for ChatGPT subscription users in the agent panel (GPT 5.6 Luna is not yet available, pending OpenAI third-party access).
Was this useful?
◆  AI Agent Frameworks

LangChain

Sources Release notes → langchain==1.3.13 NOTES

LangChain 1.3.13 adds a meta extra with langchain-meta support in init_chat_model and explicit OpenAI prompt caching.

└──▷ GET THIS VERSION
$ git clone --branch langchain==1.3.13 https://github.com/langchain-ai/langchain.git
# already have the repo? check out this version:
$ git checkout langchain==1.3.13
└──▷ TRY IT
Initialize a Meta-hosted model using the unified init_chat_model factory after installing the new meta extra.
$ pip install 'langchain[meta]'
  • Adds meta extra and integrates langchain-meta into init_chat_model, enabling Meta model initialization through the unified chat model factory.
  • Adds explicit prompt caching support for OpenAI models in the langchain-openai integration.
Was this useful?

LangChain LangGraph

Sources Release notes → cli==0.4.31 NOTES

Build resilient agents.

LangGraph CLI now supports prebuild images for langgraph deploy.

└──▷ GET THIS VERSION
$ git clone --branch cli==0.4.31 https://github.com/langchain-ai/langgraph.git
# already have the repo? check out this version:
$ git checkout cli==0.4.31
  • Supports prebuild images for langgraph deploy, enabling faster deployments by skipping the image build step.
Was this useful?

PydanticAI

Sources Release notes → v2.8.0 NOTES

PydanticAI v2.8.0 lets to_cli() accept a model override and bumps the bundled chat UI to 2.0.0.

└──▷ GET THIS VERSION
$ git clone --branch v2.8.0 https://github.com/pydantic/pydantic-ai.git
# already have the repo? check out this version:
$ git checkout v2.8.0
└──▷ USE IT
Run a model-agnostic agent from the CLI by supplying the model at invocation time.
python
agent.to_cli(model='openai:gpt-4o')
  • Adds model parameter to to_cli() so agents defined without a model can have one supplied at CLI invocation time.
  • Bumps bundled chat UI to 2.0.0 and targets sdk_version=7 in Agent.to_web().
Was this useful?
◆  Local LLM Runtimes

SGLang

Sources Release notes → v0.5.15 NOTES

SGLang v0.5.15 adds native Exa web search, decode context parallelism, new FlashKDA kernels, and expands model support to Hunyuan 3, HRM-Text, and more.

└──▷ GET THIS VERSION
$ git clone --branch v0.5.15 https://github.com/sgl-project/sglang.git
# already have the repo? check out this version:
$ git checkout v0.5.15
  • Adds flashinfer_trtllm_routed MoE runner with FlashInfer all-to-all for routed Mixture-of-Experts inference.
  • Adds opt-in non-paged indexer for DeepSeek-V4 long-context prefill, now enabled by default for large prefill chunks, with >5% end-to-end throughput gain.
  • Enables Breakable CUDA Graph as the default capture path, reducing per-step kernel-launch overhead; full CUDA Graph for the prefill phase lands as experimental.
  • Adds web_search built-in tool backed by Exa for native web search support.
  • Adds FlashKDA prefill backend for safe-gate KDA linear attention, plus ReplaySSM buffered output-only decode for linear attention.
+10 moreshow less
  • Adds Spec V2 as the default speculative decoding path with CUDA-graphable DSA draft-extend, dropped D2H/H2D syncs, and fused metadata ops for +11% end-to-end TPS.
  • Adds IndexShare MTP, reusing indexer top-k across draft steps for up to 1.9x lower draft-step cost at long context.
  • Adds TopK V2 fusing top-k selection with the page-table transform, supporting runtime k up to 2048.
  • Enables FlashMLA sparse prefill by default for DeepSeek-V4, providing >10% throughput gain on long context.
  • Enables decode context parallelism for MLA models including DeepSeek V3 and Kimi K2 series.
  • Adds shape-specialized JIT router GEMM and CuteDSL BF16 GEMM support for Blackwell (SM120).
  • Extends FlashInfer autotune coverage to draft-model graphs.
  • Adds new model support: Hunyuan 3 (Hy3), Hierarchical Reasoning Model (HRM-Text), NVIDIA LocateAnything-3B, Baidu Unlimited-OCR, JoyEcho multi-shot A/V, and Qwen3.6 NVFP4.
  • Adds speculative decoding support on XPU.
  • Bumps transformers to 5.12.1 and upgrades tvm-ffi, sgl-deep-gemm, and tilelang.
Was this useful?

llama.cpp

Sources Release notes → b9957 4 RELEASES · 2026-07-10 NOTES STABLE

llama.cpp b9957 improves server tool-use with a new tools_io abstraction and enhanced edit tool.

└──▷ GET THIS VERSION
$ git clone --branch b9957 https://github.com/ggml-org/llama.cpp.git
# already have the repo? check out this version:
$ git checkout b9957
  • Adds tools_io and tools_io_basic abstraction layer to the server, giving tool-use I/O a dedicated interface.
  • Improves the server edit tool with enhanced functionality.
  • Removes apply_diff from the server tool set.
└──▷ BREAKING ON UPGRADE
  • !The apply_diff tool has been removed from the server; any workflow relying on it will break on upgrade.
3 more releases in this issue · 2026-07-10
b9951 NOTES STABLE

llama.cpp b9951 adds an initial ExecuTorch (ET) backend with hardware-accelerated kernels for inference on ET-SoC silicon.

└──▷ GET THIS VERSION
$ git clone --branch b9951 https://github.com/ggml-org/llama.cpp.git
# already have the repo? check out this version:
$ git checkout b9951
└──▷ TRY IT
Build llama.cpp with the ET backend enabled and run the llama-server against an ET-SoC device.
$ cmake -B build -DGGML_CUDA=OFF -DGGML_ET=ON -DLLAMA_CURL=OFF -DGGML_CCACHE=ON
cmake --build build --config Release -j $(nproc)
./build/bin/llama-server --model Qwen3-0.6B-Q8_0.gguf --alias Qwen3-0.6B-Q8_0 -fa 0 --ctx-size 1024 --no-warmup --host 127.0.0.1 --port 8080
Build with ET emulator support to develop and validate ET backend kernels without physical ET-SoC hardware.
$ cmake -B build -DGGML_ET=ON -DGGML_ET_SYSEMU=ON
cmake --build build --config Release -j $(nproc)
./build/bin/test-backend-ops
  • Adds new GGML_ET backend (ExecuTorch) enabled at build time with -DGGML_ET=ON, supporting MUL_MAT, ROPE, RMS_NORM, GLU, SOFT_MAX, GET_ROWS, SET_ROWS, MUL_MAT_ID, FlashAttention, and many more operators offloaded to ET-SoC hardware.
  • Adds -DGGML_ET_SYSEMU=ON compile-time flag to enable ET-SoC emulator (sysemu) support for development and testing without physical hardware.
  • Supports Q4_0, Q4_K, Q8_0, F16, and F32 quantization formats in the ET backend, including matrix-engine (TensorFMA) accelerated paths for Q4_0 and Q8_0 MUL_MAT.
  • Supports kernel fusion in the ET backend, initially fusing RMS_NORM + MUL into a single accelerated pass.
  • Adds FlashAttention support (F32 and F16 via matrix engine) in the ET backend, enabling full offload of attention for models such as LLaMA 3.1 and Gemma 4.
+6 moreshow less
  • Adds MROPE (multi-rope) operator support in the ET backend, broadening compatible model architectures.
  • Adds uberkernel support in the ET backend, batching multiple operator kernels into a single dispatch for reduced launch overhead.
  • Adds WKV*, SSM_CONV, SSM_SCAN, GATED_DELTA_NET, and GROUP_NORM operators to the ET backend, enabling inference for Mamba/RWKV-style and hybrid SSM architectures.
  • Embeds ET kernels directly into the build via CMake (with filesystem fallback), eliminating the need for runtime kernel import hacks.
  • Adds DeepSeek-V2-Lite RoPE test coverage to test-backend-ops.
  • Supports Gemma 4 model family via the ET backend.
b9949 NOTES STABLE

llama.cpp b9949 adds cluster-parallel Flash Attention decode for Adreno GPUs via OpenCL.

└──▷ GET THIS VERSION
$ git clone --branch b9949 https://github.com/ggml-org/llama.cpp.git
# already have the repo? check out this version:
$ git checkout b9949
  • Adds cluster-parallel decode Flash Attention for Adreno GPUs via the OpenCL backend, improving inference throughput on Qualcomm Adreno hardware.
b9947 NOTES STABLE

llama.cpp b9947 adds --output option to the CLI.

└──▷ GET THIS VERSION
$ git clone --branch b9947 https://github.com/ggml-org/llama.cpp.git
# already have the repo? check out this version:
$ git checkout b9947
└──▷ TRY IT
Redirect model inference output to a file instead of stdout for downstream processing or logging.
$ llama-cli --output result.txt -m model.gguf -p "Summarize this document:"
  • Adds --output flag to the llama-cli command for directing output to a specified destination.
Was this useful?
Other / Uncategorized
◆  AI OBSERVABILITY

Arize Phoenix

Sources Release notes → arize-phoenix-v17.24.0 NOTES

Phoenix evals move to a pytest plugin, enabling PXI evals to run natively in standard test suites.

└──▷ GET THIS VERSION
$ git clone --branch arize-phoenix-v17.24.0 https://github.com/Arize-ai/phoenix.git
# already have the repo? check out this version:
$ git checkout arize-phoenix-v17.24.0
  • Moves PXI evals to a pytest plugin, allowing Phoenix evals to be executed directly within pytest test suites.
Was this useful?

Langfuse

Sources Release notes → v3.212.0 2 RELEASES · 2026-07-10 NOTES STABLE

Langfuse v3.212.0 adds dashboard widget copy-paste and versioned JSON import/export for dashboard configurations.

└──▷ GET THIS VERSION
$ git clone --branch v3.212.0 https://github.com/langfuse/langfuse.git
# already have the repo? check out this version:
$ git checkout v3.212.0
  • Adds widget menus, copy-paste, and versioned JSON import/export for dashboards, enabling shareable and reproducible dashboard configurations.
1 more release in this issue · 2026-07-10
v3.211.0 NOTES STABLE

Langfuse v3.211.0 adds boolean score filtering, project notification channels, tool-call support in evaluators, and bulk event deletion.

└──▷ GET THIS VERSION
$ git clone --branch v3.211.0 https://github.com/langfuse/langfuse.git
# already have the repo? check out this version:
$ git checkout v3.211.0
  • Adds boolean score filtering to the filters UI, enabling traces and observations to be filtered by boolean-typed scores.
  • Adds project notification channels, allowing teams to configure per-project alerting destinations.
  • Passes tool calls through to both code and LLM-as-a-judge evaluators, enabling evaluations over tool-use traces.
  • Enables select-all batch delete on the v4 events table, supporting bulk cleanup of event records.
  • Automatically blocks evaluator configs when billing is exhausted or non-401 auth LLM errors occur, preventing runaway evaluation spend.
Was this useful?
◆  VECTOR DB RAG

LanceDB

Sources Release notes → python-v0.35.0-beta.0 2 RELEASES · 2026-07-10 NOTES STABLE

LanceDB python-v0.35.0-beta.0 adds an elastic dataloader, OpenTelemetry metrics, Tencent COS/GooseFS support, and expanded expression literal types.

└──▷ GET THIS VERSION
$ git clone --branch python-v0.35.0-beta.0 https://github.com/lancedb/lancedb.git
# already have the repo? check out this version:
$ git checkout python-v0.35.0-beta.0
  • Adds get_lsm_write_spec function to read the installed LSM write spec from a dataset.
  • Aligns Permutation.with_format('torch') behavior with HuggingFace's set_format('torch') convention.
  • Adds an elastic dataloader as an iterable dataset for flexible, streaming data loading.
  • Supports date, datetime, bytes, and Decimal literals in the expression builder.
  • Exposes Lance metrics via OpenTelemetry in both Python and Node.
+1 moreshow less
  • Adds Tencent COS and GooseFS object store support via new feature flags.
└──▷ BREAKING ON UPGRADE
  • !Permutation.with_format('torch') behavior has changed to align with HuggingFace's set_format('torch') — existing code relying on the previous behavior will need to be updated.
1 more release in this issue · 2026-07-10
v0.32.0-beta.0 NOTES STABLE

LanceDB v0.32.0-beta.0 adds an elastic dataloader, OpenTelemetry metrics, Tencent COS/GooseFS support, and richer expression literals.

└──▷ GET THIS VERSION
$ git clone --branch v0.32.0-beta.0 https://github.com/lancedb/lancedb.git
# already have the repo? check out this version:
$ git checkout v0.32.0-beta.0
  • Adds get_lsm_write_spec function to read the currently installed LSM write spec from a table.
  • Exposes Lance metrics via OpenTelemetry in Python and Node, enabling observability integration.
  • Adds Tencent COS and GooseFS object store support via new feature flags.
  • Adds an elastic dataloader as an iterable dataset for flexible data loading pipelines.
  • Supports date, datetime, bytes, and Decimal literals in the expression builder.
+1 moreshow less
  • Aligns Permutation.with_format('torch') behavior with HuggingFace set_format('torch').
└──▷ BREAKING ON UPGRADE
  • !Permutation.with_format('torch') now behaves like HuggingFace's set_format('torch'), which may change output format semantics for existing callers.
Was this useful?

Weaviate

Sources Release notes → v1.38.3 NOTES

Weaviate v1.38.3 adds a /grpc-web endpoint, a runtime GraphQL toggle, hard-link replica movement, and namespace-local RBAC roles.

└──▷ GET THIS VERSION
$ git clone --branch v1.38.3 https://github.com/weaviate/weaviate.git
# already have the repo? check out this version:
$ git checkout v1.38.3
  • Introduces the /grpc-web endpoint, enabling gRPC-Web protocol support for browser and proxy-constrained clients.
  • Adds a runtime toggle for the GraphQL API, allowing operators to enable or disable the GraphQL surface without restarting the node.
  • Adds namespace-local roles, scoping RBAC role definitions to individual namespaces rather than globally.
  • Adds a gate to disallow global non-operator users, restricting operator-only surfaces from namespaced users.
  • Supports automaxprocs for automatic GOMAXPROCS tuning via cgroup v2, improving CPU scheduling in containerized deployments.
+4 moreshow less
  • Propagates raw on-disk object bytes in async replication, reducing serialization overhead during replica sync.
  • Uses batched hashtree-root pre-filtering for many-tenant clusters in async replication, reducing per-hashBeat overhead.
  • Increases the hfresh searchProbe default to 256, improving recall for hybrid-fresh index queries.
  • Optimizes the cycle manager for large multi-tenant collections, reducing overhead when managing many tenants.
Was this useful?
my-toolchain — 0 tools
paste an install list to detect your tools

A brew list, a Brewfile, requirements.txt, a Dockerfile — or just the product names, free-form. Nothing leaves your browser.

    browse all tools →