Backfill issue -047 — 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 Agent Frameworks

Agno (formerly Phidata)

v2.6.22 NOTES

Agno v2.6.22 adds TwelveLabsTools, SofyaTools, and SearchApiTools plus a base Toolkit timeout parameter.

└──▷ GET THIS VERSION

$ git clone --branch v2.6.22 https://github.com/agno-agi/agno.git
# already have the repo? check out this version:
$ git checkout v2.6.22

└──▷ USE IT

Integrate TwelveLabs video analysis into an Agno agent to analyze uploaded video content.

python

from agno.tools.twelvelabs import TwelveLabsTools

tools = TwelveLabsTools()
agent = Agent(tools=[tools])

›Adds TwelveLabsTools class for video analysis and multimodal text embedding generation via the TwelveLabs API.
›Adds SofyaTools class exposing search, extract, and research capabilities.
›Adds SearchApiTools class with Google, News, Images, and YouTube search methods.
›Adds timeout parameter to the base Toolkit class, wiring HTTP timeouts across tools and extending timeout support to additional toolkits.

Was this useful?

PydanticAI

v2.4.0 NOTES

PydanticAI v2.4.0 adds GEval, five agentic span evaluators, and splits file-upload security controls into two distinct parameters.

└──▷ GET THIS VERSION

$ git clone --branch v2.4.0 https://github.com/pydantic/pydantic-ai.git
# already have the repo? check out this version:
$ git checkout v2.4.0

›Splits preserve_file_data into allow_uploaded_files (inbound security control) and a separate AG-UI representation opt-in parameter, giving finer-grained control over uploaded file handling.
›Adds GEval evaluator and standard quality metric rubrics for LLMJudge, enabling criteria-driven LLM-as-judge scoring.
›Adds five agentic span-based evaluators — ToolCorrectness, TrajectoryMatch, ArgumentCorrectness, MaxToolCalls, and MaxModelRequests — for evaluating agent execution traces.

└──▷ BREAKING ON UPGRADE

!The preserve_file_data parameter is split into allow_uploaded_files and an AG-UI opt-in; code referencing preserve_file_data will break on upgrade.

Was this useful?

◆ AI Coding Agents

Anthropic Claude Code

v2.1.200 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 v2.1.200 makes 'Manual' the default permission mode and adds opt-in idle timeout for AskUserQuestion dialogs.

└──▷ GET THIS VERSION

$ git clone --branch v2.1.200 https://github.com/anthropics/claude-code.git
# already have the repo? check out this version:
$ git checkout v2.1.200

›Changes --permission-mode manual and "defaultMode": "manual" in config to be the new default permission mode across CLI, --help, VS Code, and JetBrains (previously called 'default').
›Adds opt-in idle timeout for AskUserQuestion dialogs via /config, replacing the previous auto-continue behavior.

Was this useful?

Cline

cli-v3.0.35 NOTES

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

Cline CLI v3.0.35 adds ClinePass for all users and Tencent TokenHub as a new AI provider.

└──▷ GET THIS VERSION

$ git clone --branch cli-v3.0.35 https://github.com/cline/cline.git
# already have the repo? check out this version:
$ git checkout cli-v3.0.35

›Enables ClinePass for all CLI users.
›Adds Tencent TokenHub as a supported AI provider.
›Refreshes the bundled model catalog with the latest provider models.

Was this useful?

Block Goose

v1.41.0 NOTES

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

Goose v1.41.0 adds 60+ new capabilities including new AI providers, CLI commands, hooks, summon subagent controls, ACP improvements, and broad localization.

└──▷ GET THIS VERSION

$ git clone --branch v1.41.0 https://github.com/block/goose.git
# already have the repo? check out this version:
$ git checkout v1.41.0

└──▷ TRY IT

Cap tool response payloads to avoid context bloat when running tools that return large outputs.

$ GOOSE_MAX_TOOL_RESPONSE_SIZE=32768 goose run --recipe my-recipe.yaml

Switch the model mid-session in the CLI without restarting, useful when you want to switch to a cheaper model after planning.

$ /model

›Adds iFlytek Spark and Astron MaaS as new AI providers.
›Adds Fireworks AI as a declarative provider.
›Adds Together AI as a declarative provider.
›Adds OrcaRouter as a declarative OpenAI-compatible provider.
›Adds EmpirioLabs AI as a declarative provider.

+65 moreshow less

›Adds Perplexity as a declarative OpenAI-compatible provider.
›Adds Alibaba (Qwen via DashScope) as a declarative provider.
›Adds Databricks AI Gateway provider.
›Adds NEAR AI Cloud provider.
›Adds Scaleway provider.
›Adds Vercel AI Gateway provider.
›Adds FuturMix provider.
›Adds Routstr as a declarative provider.
›Adds oMLX as a declarative OpenAI-compatible provider.
›Adds Atomic Chat as a declarative OpenAI-compatible provider.
›Adds xAI SuperGrok OAuth subscription provider.
›Adds GLM-5.2 model support.
›Adds MLX model support for local inference.
›Adds gpt-5.5 model support for the ChatGPT Codex provider.
›New --edit session flag to edit a conversation before forking it.
›New /status slash command in the CLI.
›New /model slash command for switching models mid-session.
›New /goal command for agent self-evaluation before finishing.
›New goose review subcommand for local code review.
›New tui command on goose-cli.
›New goose://resume deep link to resume a session.
›New goose://new-session deep link for starting a fresh chat.
›Adds OpenRouter request parameters support.
›Adds OpenAI Responses API store param configurability.
›Adds Azure Entra ID bearer token auth via AZURE_OPENAI_AD_TOKEN.
›Adds Streamable HTTP support in deeplink generator.
›Adds TLS support for ACP serve.
›Exposes available tools in ACP schema.
›Adds ACP thinking effort config option.
›Adds last message snippets for ACP sessions.
›Adds paginated session list in ACP.
›Adds slash commands (built-in, skill, recipe) in ACP server.
›Adds ACP elicitation improvements.
›Adds OTLP logging schema for cross-tool detection.
›Adds hooks feature with PreToolUse denial support.
›Adds summon peek mode for async background tasks.
›Adds summon context parameter for delegate tasks.
›Adds summon working_dir override for delegate tasks.
›Adds structured metadata for summon task load results.
›Adds summon subagent instructions.
›Adds global hints loading from ~/.agents/AGENTS.md.
›Adds session import functionality.
›Adds Harbor eval runner.
›Adds TUI diff viewer.
›Adds cache token tracking for accurate cost reporting.
›Adds configurable tool output size limit via GOOSE_MAX_TOOL_RESPONSE_SIZE.
›Adds GOOSE_FAST_MODEL env var support.
›Adds GOOSE_OAUTH_CALLBACK_PORT for stable OAuth redirect URIs.
›Adds proactive OAuth token refresh to avoid re-auth on every session.
›Adds configurable MAX_CODE_BLOCK_LINES via env vars.
›Adds quarterly option for the scheduler.
›Adds optional --parameters for scheduled recipes.
›Adds worktree-aware directory switcher.
›Adds chat history search feature.
›Adds projects as backend sources with system prompt injection.
›Adds unified thinking effort control across all providers.
›Adds open-plugins generalization and skills.
›Adds structured per-provider config block with non-destructive provider switching.
›Adds GitHub Copilot /responses API support.
›Adds image read tool.
›Adds Nushell terminal and completion support.
›Adds optional api_key configuration for declarative OpenAI-engine providers.
›Adds option to disable automatic update downloads.
›Adds language selection feature in desktop app.
›Adds localization for French, German, Italian, Portuguese, Indonesian, Malay, Vietnamese, Traditional Chinese, Spanish, Korean, Japanese, Hindi, Russian, Turkish, and Simplified Chinese.

Was this useful?

Alibaba Qwen Code

v0.19.6 NOTES

Qwen Code v0.19.6 adds nested sub-agents, session artifact APIs, mobile-mcp, dataviz skill, and configurable cron job expiration.

└──▷ GET THIS VERSION

$ git clone --branch v0.19.6 https://github.com/QwenLM/qwen-code.git
# already have the repo? check out this version:
$ git checkout v0.19.6

›Adds visionModelId field to workspace provider status; exposed in the daemon API and web-shell model dialog.
›Introduces session artifact APIs in the daemon (feat(daemon): add session artifact APIs).
›Vendors mobile-mcp with opt-in 0–1000 relative coordinate mode for mobile automation.
›Makes recurring cron/loop job expiration configurable in the scheduler.
›Allows sub-agents to spawn nested sub-agents up to a configurable depth.

+7 moreshow less

›Adds a bundled dataviz skill to core.
›Adds daemon UI support for vision model selection in the web-shell.
›Displays nested sub-agents as a tree in the web-shell tasks panel.
›Adds sessionless workspace memory forget and dream operations.
›Improves slash command discovery with a taller menu, group counts, and fuzzy search.
›Syncs vendored cua-driver from 0.6.8 to 0.7.0.
›Shows the qwen-code version in the sidebar footer.

Was this useful?

◆ Local LLM Runtimes

llama.cpp

b9867 2 RELEASES · 2026-07-03 NOTES STABLE

llama.cpp b9867 adds spec-draft-p-min support to DFlash speculative decoding.

└──▷ GET THIS VERSION

$ git clone --branch b9867 https://github.com/ggml-org/llama.cpp.git
# already have the repo? check out this version:
$ git checkout b9867

›Adds spec-draft-p-min parameter support to DFlash speculative decoding, with n_min and n_max guards to bound draft token counts.

1 more release in this issue · 2026-07-03

b9864 NOTES STABLE

Adds sse_ping_interval as a per-request body field so slow prefill never drops healthy SSE connections

└──▷ GET THIS VERSION

$ git clone --branch b9864 https://github.com/ggml-org/llama.cpp.git
# already have the repo? check out this version:
$ git checkout b9864

›Adds sse_ping_interval as a typed per-request body field in the server's request schema, accepting positive integers or -1 (disabled), with hard limits of -1 to INT32_MAX; the global CLI default remains 30 so existing API clients see no behavior change.
›Server now pings silent SSE streams every 1 second and kicks connections only after 3 seconds, preventing slow prefill from dropping healthy long-running connections.
›WebUI automatically sends sse_ping_interval: 1 in each request body to enforce the 3-second visibility-kick contract without requiring server-side reconfiguration.

Was this useful?

Other / Uncategorized

◆ AI OBSERVABILITY

Arize Phoenix

arize-phoenix-client-v2.12.0 NOTES

Phoenix client v2.12.0 adds REST endpoints for assigning annotation configs to projects and managing dataset labels.

└──▷ GET THIS VERSION

$ git clone --branch arize-phoenix-client-v2.12.0 https://github.com/Arize-ai/phoenix.git
# already have the repo? check out this version:
$ git checkout arize-phoenix-client-v2.12.0

›Adds REST API support for assigning annotation configs to projects.
›Adds REST endpoints for dataset label management.

Was this useful?

Langfuse

v3.205.0 2 RELEASES · 2026-07-03 NOTES STABLE

Langfuse v3.205.0 adds Parquet export, widget creation via MCP/API, and richer trace graph and session views.

└──▷ GET THIS VERSION

$ git clone --branch v3.205.0 https://github.com/langfuse/langfuse.git
# already have the repo? check out this version:
$ git checkout v3.205.0

›Promotes Parquet to a first-class, whitelisted export fileType for blob exports, alongside existing formats.
›Enables dashboard widget creation via the API and MCP (Model Context Protocol) integration.
›Adds real v4 session support in the Sessions view with a new View model and visible filters.
›Improves the trace graph view with live framing, a collapsible per-project panel, and broader availability across traces.
›Suggests observed metadata keys with their types in the search bar for faster filter building.

+3 moreshow less

›Collapses the filter sidebar to a rail and moves the date picker into the page header across table views.
›Improves the code evaluator editing experience in the Evals section.
›Shows the playback transport control only on traces that have playback content.

1 more release in this issue · 2026-07-03

v3.204.0 NOTES STABLE

Langfuse v3.204.0 adds SDK attribution persistence, auto-collapsing tool calls, and a custom graph renderer with trace playback.

└──▷ GET THIS VERSION

$ git clone --branch v3.204.0 https://github.com/langfuse/langfuse.git
# already have the repo? check out this version:
$ git checkout v3.204.0

›Adds trace playback and a custom graph renderer for visualizing trace execution in the trace UI.
›Persists SDK attribution data during ingestion, enabling source tracking across SDKs.
›Auto-collapses tool calls in the trace UI when a trace contains many tool calls, reducing visual noise.

Was this useful?

// END OF TRANSMISSION

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

The AI Toolchain — issue -047, July 3, 2026