AI.info
AI Tools Updates — Page 17
What changed in the AI tools this week — dated, from the makers’ own changelogs.
- LangGraph — The evaluators list API accepts an agent_id filter to return evaluators attached to the agent’s environments or tagged datasets.
The evaluators list API accepts an agent_id filter.
- LangGraph — Downloading experiment results with “Include annotator name and comment per row” now keeps automated evaluator scores in every feedback column, alongside human annotations.
Downloading experiment results now keeps automated evaluator scores in every feedback column, alongside human annotations.
- LangGraph — Edit and delete in the experiments table now check project permission, to match BE.
Edit and delete in the experiments table now check project permission.
- LangGraph — An evaluator using an organization-scoped model configuration that authenticates with an API key no longer attempts a per-model OAuth token exchange. Amazon Bedrock models configured this way are no l
An evaluator using an API key no longer attempts a per-model OAuth token exchange. Amazon Bedrock models are no longer rejected as unsupported.
- LangGraph — The completed items list in an annotation queue now offers Add to Dataset for selected thread items, importing each thread as a multi-turn example.
The completed items list offers Add to Dataset for selected thread items, importing each thread as a multi-turn example.
- LangGraph — Charts now show shape-aware loading skeletons instead of temporary zero values while data loads. Evaluator spend charts use bar and metric placeholders that preserve the final layout.
Charts show loading skeletons instead of temporary zero values while data loads.
- LangGraph — Categorical evaluator fields resize with the editor pane, keeping descriptions and remove-category buttons inside the feedback configuration card.
Categorical evaluator fields resize with the editor pane.
- AssemblyAI — MiniMax M3 Now Available on the LLM Gateway
MiniMax's minimax-m3 is now available through the LLM Gateway. It supports tool calling, structured outputs, streaming, and the reasoning_effort parameter.
- AssemblyAI — Gemini 3.8 Flash Now Available on the LLM Gateway
Google's gemini-3.8-flash is now available through the LLM Gateway. It has a 1M-token context window and supports tool calling, structured outputs, and streaming.
- AssemblyAI — Kimi K3 Now Available on the LLM Gateway
Moonshot AI's kimi-k3 is now available through the LLM Gateway. It supports tool calling, structured outputs, streaming, and the reasoning_effort parameter.
- Mixedbread Embed — Structured Output for Toast 1
The Responses and Chat Completions APIs now support structured outputs for Toast 1.
- Vapi — What's New: Week of September 14, 2026
Set a same-provider fallback model for Anthropic Bedrock; Composer can create and run simulations.
- Zencoder Coding Agent — August 2026
Added a structured logs panel with task-filtered execution logs and MCP tool-call details; reasoning-effort selection now supports more models.
- CodeRabbit — Requested team overrides | GitHub
Members of a requested GitHub reviewer team can ignore failing Pre-Merge Checks when reviewer-team overrides are enabled.
- Vectara Platform — Confluence Ingests Attachments and Embedded Images
A Confluence pipeline source ingests page attachments as records and inlines images embedded in page bodies.
- Meticulous — ✅ Let Your Agents Review Your Test Run Results
Your agent inspects the diffs in a test run, works out which are regressions, and leaves comments explaining what went wrong.
- Meticulous — 👯 Side-by-Side Replay
Watch the before and after replays side by side.
- Riverside.fm — Replace a published podcast episode's file
Update a live episode by replacing its file in place, without losing your analytics or reshuffling your feed.
- Augment Code — IntelliJ Plugin v0.491.0 Release Notes
The agent now asks for your approval before reading a file that looks sensitive; MCP servers reached over HTTP, SSE, or Streamable HTTP now go through your configured…
- Augment Code — VSCode Extension 0.901.1 Release Notes
Chat submit is now a VS Code keyboard shortcut; workspace-defined Augment hooks now run only in trusted workspaces.
- RunPod Serverless — Nano Banana Edit
The Nano Banana Edit public endpoint will be retired on September 28, 2026. Migrate to Nano Banana 2 Edit.
- Rossum — Important Announcement: Deprecation of Node.js 22
New extensions must be created using the Node.js 24 runtime; existing extensions should be updated and switched to the Node.js 24 runtime.
- LangChain — Evaluator side panels disable Save while the configuration is invalid and display the reasons. Correcting the configuration enables saving again.
Evaluator side panels disable Save while the configuration is invalid and display the reasons.
- LangChain — Edit and delete in the experiments table now check project permission, to match BE.
Edit and delete in the experiments table now check project permission.
- LangChain — The completed items list in an annotation queue now offers Add to Dataset for selected thread items, importing each thread as a multi-turn example.
The completed items list in an annotation queue now offers Add to Dataset for selected thread items.
- LangChain — Downloading experiment results with “Include annotator name and comment per row” now keeps automated evaluator scores in every feedback column, alongside human annotations.
Downloading experiment results now keeps automated evaluator scores in every feedback column, alongside human annotations.
- LangChain — Charts now show shape-aware loading skeletons instead of temporary zero values while data loads. Evaluator spend charts use bar and metric placeholders that preserve the final layout.
Charts now show loading skeletons instead of temporary zero values while data loads.
- LangChain — An evaluator using an organization-scoped model configuration that authenticates with an API key no longer attempts a per-model OAuth token exchange. Amazon Bedrock models configured this way are no l
An evaluator using an organization-scoped model configuration that authenticates with an API key no longer attempts a per-model OAuth token exchange.
- LangChain — Categorical evaluator fields resize with the editor pane, keeping descriptions and remove-category buttons inside the feedback configuration card.
Categorical evaluator fields resize with the editor pane.
- LangChain — The evaluators list API accepts an agent_id filter to return evaluators attached to the agent’s environments or tagged datasets.
The evaluators list API accepts an agent_id filter.
- ElevenLabs — ElevenAgents
Added call queueing and hold audio, music_v2_5 support, and test version metadata.
- Jellypod — Add Visual Transitions Between Shots
Choose from seven shot transition styles and adjust their duration.
- Devin — September 11, 2026
Machine startup status is clearer, sidebar keyboard navigation is expanded, nested groups stay open, and primary buttons use a neutral style.
- Truss — Model API Deprecation (GLM 4.7, Kimi K2.7, Kimi K2.6, Inkling, Inkling Small, DeepSeek v4 Pro)
GLM 4.7, Kimi K2.7, Kimi K2.6, Inkling, Inkling Small, and DeepSeek v4 Pro will be deprecated at 5pm PT September 25th.
- Windsurf — September 11, 2026
Computer tab shows machine provisioning; sidebar supports arrow-key navigation; nested groups stay expanded; primary buttons use neutral styling.
- Pinecone — Higher namespace limit on the Enterprise plan
The Enterprise limit for namespaces per serverless index is now 1,000,000.
- Baseten — Model API Deprecation (GLM 4.7, Kimi K2.7, Kimi K2.6, Inkling, Inkling Small, DeepSeek v4 Pro)
GLM 4.7, Kimi K2.7, Kimi K2.6, Inkling, Inkling Small, and DeepSeek v4 Pro will be deprecated at 5pm PT September 25th.
- CivitAI — Weekly Update: September 5 to September 12, 2026
Ideogram 4.0 is in the generator; MiniMax H3 is the default video engine; ControlNet for MiniMax H3.
- Recall — Recall release notes, September 12, 2026 : Save your books with EPUB support, capture text on social posts, and add up to 10 images at a time.
EPUB books can be uploaded, read, searched, and chatted with; users can add up to 10 images at a time, and Recall reads text in social-post images.
- Kimi — 3.2.8 (2026-09-12)
Anti-sleep setting now offers three options: keep running when the screen is off, keep the screen on, and off.