AI.info
AI Tools Updates — Page 60
What changed in the AI tools this week — dated, from the makers’ own changelogs.
- Retool Agents — Gemini 4 Argon available in Retool
Gemini 4 Argon is available for Enterprise organizations using their own key through Google and Google Vertex AI.
- Vapi — What's New: Week of October 19, 2026
Claude Haiku 5.5, Gemini 3.8 Flash, and Cartesia Sonic-3.6 are now available.
- Coveo for Workplace — Multi-language support for ServiceNow knowledge articles!
The ServiceNow source provides metadata for language variants, identifies base-language articles and translations, and indexes knowledge base names in the article’s…
- CodeRabbit — GitLab webhook migration with CSV input | Self-managed GitLab
The GitLab webhook migration command now accepts headerless CSV lists of project or group IDs, paths, or URLs.
- Ursa — 2026.10.19 Realtime SaaS
Melia 1 improves transcription accuracy across many languages; output_locale accepts either a single locale or a list, with at most one per language.
- Picsart AI Image Generator — Kling 4.0 is now in Picsart with 30-second and 4K video generation
Kling 4.0 and Kling 4.0 Omni are now available in Picsart for video generation and editing.
- Unsloth — DeepSeek-V4.1 & FLUX-3
Added DeepSeek-V4.1-Flash-1017 and FLUX.3 Image Klein, plus images and videos in chat, FSDP2 training, and MCP setup.
- Novelcrafter — October 19, 2026
Added Gemini 4 Argon and a “Recently released” model picker section, improved model search, and fixed novel import issues.
- Ada AI Agent — Messaging SDKs start a new session only when the end user engages
The Messaging SDKs now wait for engagement before creating a new conversation session.
- Ada AI Agent — MCP server: AI Agent and human agent CSAT in get_ada_metric
get_ada_metric has two new metric types, csat_rate_ai and csat_rate_human.
- Ada AI Agent — Change to the HANDOFF filter
Change to the HANDOFF filter.
- Retool — Gemini 4 Argon available in Retool
Gemini 4 Argon is available in Retool through Google and Google Vertex AI for Enterprise organizations using their own key.
- Retool — Expanded source control managers for multi-instance releases
Enterprise organizations can use supported SCM providers for multi-instance app releases.
- Exa AI — Lite search, recommended search types, and search warnings
Added lite search, recommended lite, auto, and deep search types, and added warnings when /search requests don't do what developers expect.
- Kilo Code — v7.8.15 (release)
Added browser_open permission choices and let primary agents select which agent runs an Agent Manager session.
- Snowflake Cortex Analyst — Tools for Cortex AI Gateway (Preview)
Tools for Cortex AI Gateway.
- OpenHands — v1.28.0
Agents can request secrets through an in-conversation modal.
- Krisp — September Product Updates: Call Center AI
Our latest updates bring clearer translation accuracy reporting, flags for potential voice...
- Krea AI — Krea 3 API is public, and MCP generation tools can quote a cost with preflight
Every workspace can call the Krea 3 API; MCP generation tools can return a cost quote and balance without submitting a job.
- Zilliz Cloud — 🚚 Zero Downtime Migration migrates collections into an existing Dedicated cluster with no downtime, syncing incremental changes for up to 7 days before cutover.
Zero Downtime Migration, Data Backfill, CMEK with Google Cloud clusters, two more BYOC regions, and SSO enforcement exemptions.
- LangChain — 2026-10-12 - LangSmith Cloud update
Added tracing-only API keys and ScaleDown operations and model to the LLM Gateway.
- Lovable — Connect your app to Cube
You can now connect your Lovable app to Cube to read the measures and dimensions your data team defines in its semantic layer.
- Lovable — Faster dictation
Dictation now transcribes while you speak, so your text is ready when you stop recording.
- Hugging Face — TorchNeuron: running HF models on AWS Trainium is easier than ever
TorchNeuron makes running Hugging Face models on AWS Trainium easier.
- Notion AI — Agents, rebuilt to do more for less
Agents were rebuilt to do more for less.
- Lindy — Integrations
Airtable, Intercom, Google Analytics, Slack, ScrapeCreators, ClickUp, Luma, Parallel and Vestaboard; Lindy asks questions and waits for your answer.
- Amazon Bedrock — Structured outputs now available for Anthropic Claude models on both endpoints
Structured outputs is now available for Anthropic Claude models on both the bedrock-runtime and bedrock-mantle endpoints.
- Letta — Release v0.34.17
Improved local compaction, CLI interrupt handling, listener readiness, and message and recovery handling.
- Culture Amp AI Coach — Culture Amp MCP Brings Survey Insights and Performance Data to Your AI Assistant
Culture Amp MCP now helps eligible customers explore engagement insights and prepare performance reviews directly in Claude or ChatGPT.
- Taskade — v6.288.0: Brand Colors for Apps & Fit-to-Screen Mind Maps
Taskade Genesis apps now have brand color and look options, and Mind Map has a Fit to screen button.
- Culture Amp Engage — Culture Amp MCP Brings Survey Insights and Performance Data to Your AI Assistant
Culture Amp MCP now helps eligible customers explore engagement insights and prepare performance reviews directly in Claude or ChatGPT.
- LiteLLM — LiteLLM v1.106.0 adds Claude Haiku 5.5, a Rust Lens service and MCP rate limits
MCP Gateway rate limits now cover every MCP operation, and a new server-level rpm caps all callers of one server.
- Claude Code — 2.1.303
Bug fixes and reliability improvements.
- Lindy — Airtable, Intercom, Google Analytics and more
Read and write records; reply to conversations; run GA4 reports; follow Slack threads and DMs; ask for a decision with a form.
- Fireworks AI — New Serverless Training model: DeepSeek V4.1 Flash 1017
DeepSeek V4.1 Flash 1017 is now available for LoRA workloads on the shared Serverless Training pool, with up to 262K context.
- Fireworks AI — Upcoming Serverless Training deprecation: DeepSeek V4 Flash (0731)
DeepSeek V4 Flash (0731) will be removed from the shared Serverless Training pool on November 2, 2026.
- Fireworks AI — Expanded Small serverless rate-limit tier
The Small tier now includes models with fewer than 800B total parameters, up from 600B; the Medium tier covers models from 800B to less than 1.6T.
- Fireworks AI — New training model: NVIDIA Nemotron 3.5 Super VL
NVIDIA Nemotron 3.5 Super VL 120B A12B is now available for training on all three surfaces, with up to 262K context.
- vLLM — v0.32.0
Added POST /v1/systemone for choice, noul, and score questions over structured state; it is served by default.
- ComfyUI — v0.42.0
Added Krea 3 Image and MOSS-image support; removed retired Arrow 1.x models.