AI.info
AI Tools Updates — Page 22
What changed in the AI tools this week — dated, from the makers’ own changelogs.
In inglese
- Appsmith AI — Release v2.4🌈
Appsmith AI reaches end of life on September 30, 2026; its data sources will stop working after that date.
- Windsurf — September 9, 2026
Archiving a session can close child sessions' PRs; sidebar lists multiple PRs; file paths open diffs; Desktop works on touch devices.
- Fireworks AI — Training skill for coding agents
A Fireworks training skill is available for coding agents to plan a run, estimate cost, and wait for approval before spend.
- Fireworks AI — Training cost estimator
The training cost estimator helps estimate what a training job will cost before you run it.
- AssemblyAI — Public Models Endpoint With Pricing for LLM Gateway
The LLM Gateway now exposes a public list-models endpoint that returns every available model with its supported parameters, context length, pricing, provider, and…
- Browserbase — Webhooks in Functions
Webhooks are now available for Browserbase Functions. Subscribe to events like Function pending, running, completed, and failed.
- Renderforest AI Logo Maker — GPT Image 2.5 on Renderforest
GPT Image 2.5 is now available in the AI Image Generator with two new models: Sunburst and Flare.
- Retool Agents — Self-hosted Retool 4.34, 4.0 stable updates
Retool 4.34.2 and 4.0.15 patches are available on the Stable channel with bug fixes and security patches.
- Brand24 — Discover more TikTok mentions with video transcriptions
Brand24 can detect keywords in TikTok video transcriptions and show spoken fragments with timestamps.
- scite — Citation Surfing with Scite MCP
Scite MCP includes a citation graph tool that lets AI agents like ChatGPT and Claude explore how papers are connected through citations.
- SaneBox — SaneBox is now on Android
SaneBox launched an Android app on Google Play.
- CrewAI — 1.15.21
Added telemetry for checkpoint runtime and CLI usage.
- Appsmith — Release v2.4🌈
Appsmith AI datasources will stop working September 30, 2026; model lists were refreshed for OpenAI, Anthropic, and Google AI.
- Vectara Platform — Document Deletes Report a Precise Outcome
DELETE /v2/corpora/{corpus_key}/documents/{document_id} now responds once the document has actually been removed, and reports five outcomes it previously did not…
- Unsloth — Large Performance Gains + Fixes
Adds faster diffusion, AMD Vulkan performance improvements, PyTorch 2.11, MLX updates, and fixes for RAG uploads and AppImage lag.
- ComfyUI — v0.35.0
Added Recraft V4 Styles, WAN3-Prime support and AVIF saving; dropped Veo 2 and Veo 3.0 models and updated workflow templates.
- Retool — Self-hosted Retool 4.34, 4.0 stable updates
Stable patches Retool 4.34.2 and 4.0.15 are available with bug fixes and security patches.
- Retool — Retool 4.54 Edge for self-hosted instances
Retool 4.54 is available on the Edge release channel for self-hosted organizations.
- OpenHands — v1.17.0
Added provider connections and local Planner support, conversation tags and filtering, automation run outcomes and permissions, and unified cloud settings.
- Krea AI — Elements in Seedance Studio, GPT Image 2.5 models, MiniMax H3 Max Turbo with native 1080p, postpaid invoicing, and Kling retirements
Elements save characters, locations, and objects; GPT Image 2.5 models are available through the API; MiniMax H3 Max Turbo adds native 1080p.
- Notion AI — Control which AI models your agents can use
You can control which AI models your agents can use.
- ON1 Photo RAW — Introducing ON1 Photo RAW 2027: A New RAW Processing Engine, Improved NoNoise AI, an AI Culling Assistant and More
Photo RAW 2027 adds a new RAW processing engine, improved NoNoise AI, and an AI culling assistant.
- Raycast — 🦙 Bring Your Own Model
Added custom OpenAI-compatible providers, local Ollama models, and OpenRouter models to Raycast AI.
- Napkin AI — v1.1.24 (2026-09-08)
Credits are charged once per visual request.
- Swapcard — Anonymous Group Registration: Buy the Passes Now, Assign Them Later 🎟️
Buyers can purchase a block of registrations and assign the passes later.
- Kittl — Your brand, in every asset you make
Brand Studio lets users create a brand kit from a website URL or from scratch, and applies it to generations.
- Murf AI — Instant Voice Cloning is Now Live
Enterprise customers can create AI voice clones through the Murf API using a short recording; clones can speak across all locales supported by Falcon 2.
- Culture Amp AI Coach — Uncover the "Why" Behind Your Data with AI Follow-Ups
AI automatically asks relevant follow-up questions based on a participant's rating or comment; view conversations in the Comments report.
- Arize Phoenix — Select OpenAI GPT-6 in the prompt playground
Select OpenAI’s gpt-6-astra in the prompt playground model picker and run prompts against it.
- Arize Phoenix — Inspect remote evaluator runs as traces
Remote evaluator runs emit a trace for each evaluated row into your evaluations project, with spans capturing request, response, status code, and latency.
- Arize Phoenix — Add and annotate entire sessions and traces in annotation queues
Send sessions and traces to an annotation queue and review them at session or trace granularity.
- Jellypod — Videos, Shorts, and Voiceovers in the API and MCP
Create narrated videos through the API or MCP, using the same product choices as Studio.
- Jellypod — Chalkboard, Blueprint, and a Better Whiteboard
Add Chalkboard and Blueprint board styles and use an improved Whiteboard Explainer.
- PromptLayer — Deployment 1
Added Gemini 3.8 Flash support across Google AI Studio, Vertex AI, and OpenRouter; updated GPT-5.6 Sol and Sol Pro pricing.
- Arize AX — Inspect remote evaluator runs as traces
Remote evaluator runs now emit a trace for each evaluated row into your space’s evaluations project.
- Arize AX — Add and annotate entire sessions and traces in annotation queues
You can send full multi-turn sessions and complete traces to an annotation queue and review them at session or trace granularity.
- Arize AX — Select OpenAI GPT-6 in the prompt playground
You can now select OpenAI’s gpt-6-astra in the prompt playground model picker and run prompts against it.
- Deepgram — Hold a function call until the user's turn is confirmed
Hold function calls until speech-to-text confirms the user has finished speaking.
- Wordly — Portal 4.11.0
We've added a dedicated page for custom fields, and Org Admins in an SSO realm can create, edit, and delete custom fields.
- Wordly — Android 4.7.0 and iOS 4.7.0
We've added Silent Theater integration for presenters and attendees.