AI.info
AI Tools Updates — Page 51
What changed in the AI tools this week — dated, from the makers’ own changelogs.
En anglais
- Ada AI Agent — Show queue position and wait time during Salesforce Messaging handoffs
End users waiting for a live agent after a Salesforce Messaging handoff now see a queue banner in the chat widget.
- Ada AI Agent — API tool response time in the API usage report
The API usage report now shows how long your API tools take to respond, across all of your conversations.
- Hex — Pull agent usage data and import Ossie models
Workspace Admins on Enterprise plans can access agent usage and credit data via API, MCP server, or CLI, and import Apache Ossie models.
- Kilo Code — JetBrains 7.1.12
Added project-scoped agent usage aggregation, custom auto routers, a session host registry, and a Kilo Cloud session adapter.
- Keploy — v3.6.159
Dropped distrusted eMudhra roots, took NSS roots from x509roots, and built with Go 1.27.3.
- OpenHands — v1.31.0
Self-hosted instances hide the GitHub token secret field when GitHub is connected.
- OpenHands — v1.30.0
The setup choice modal reflects the built-in integration's connection state.
- llama.cpp — b11821
Refactored SYCL Intel XMX multi-column mul_mat_vec_q support for reordered K-quant and Q8_0 weights.
- llama.cpp — b11813
Pixtral now uses the Tekken pre-tokenizer instead of the Llama 3 pre-tokenizer.
- llama.cpp — b11819
Added support for Nemotron non-standard GQA, BPE tokenization, and domyn-small pre-type.
- llama.cpp — b11814
Added the clean_spaces GGUF key.
- llama.cpp — b11818
Added typed content input support for the RERANK endpoint.
- JetBrains AI Assistant — Air Governance: Manage AI Agents, Access, and Spend Throughout Your Organization
JetBrains Air Governance is generally available to organizational customers for managing AI agents, models, providers, spending limits, and usage.
- Everlaw Ediscovery Platform — Everlaw for Word Brings Your Evidence Directly to Where You Draft
Everlaw for Word brings evidence directly to where users draft.
- Everlaw Ediscovery Platform — Everlaw Brings Frontier AI Closer to the Evidence with ChatGPT Plugin
A new plugin connects Everlaw’s governed evidence layer with ChatGPT Enterprise.
- n8n — n8n@1.123.86
- n8n — n8n@2.45.2
Added CSV export for migration reports; updated type-policy enforcement, Postgres handling, and OAuth confirmation and email logos.
- n8n — n8n@2.44.6
Fixed the Postgres node overwriting global pgtypes.
- n8n — stable
Fixed the Postgres node overwriting global pgtypes.
- Ollama — v0.40.9
- LM Studio — Bionic 1.1.8
Support for previewing Word documents, spreadsheets, presentations, and HTML files; decision model inference and management through the REST API.
- Zed — Zed v1.25.2
Added inlay hint text edits by double-clicking, external agent context compaction and session notices, and three models to Amazon Bedrock.
- Mistral AI — Mistral Advances AI for Engineering with NVIDIA
Mistral and NVIDIA are developing AI-native engineering workflows powered by NVIDIA accelerated computing.
- Rox — Find your workflows in Automations and prospecting in Outbound
Automations provides the workflows you build to automate repetitive tasks. Outbound provides automated prospecting.
- Resemble AI — Live Signal, Identity and Intelligence on the meeting bot
The meeting bot now runs Signal, Identity and Intelligence live, with in-call chat alerts and a post-call report email.
- Ask Sage — Ask Sage v3.10.3
Makes automatic token increases apply in real time and restores streaming and prompt caching for Claude models on AWS Bedrock GovCloud.
- Ask Sage — Ask Sage v3.10.4
Fixes password resets so users can complete them from the emailed reset link.
- Truss — v0.18.36rc2
Added batch System One API, train examples lookup, CLI training updates, and pending job messages on push.
- Truss — v0.18.36
Added batch System One API, train examples lookup, CLI training updates, and pending job messages on push.
- Consensus — Threads now available in the Consensus API and MCP
The Consensus API and MCP now support Threads, allowing users to create and manage research Threads, ask follow-up questions, and retrieve previous Threads.
- CoreWeave Cloud — CoreWeave Launches the CoreWeave Zero Trust AI Framework, Keeping Customers in Control of Access to Their Models, Data and Runtime Environments
Six layers of customer-controlled security, from encryption keys to runtime, designed to keep CoreWeave out of the customer's trust chain
- CoreWeave Cloud — CoreWeave Unveils CoreWeave Physical AI Launchables, Taking the Setup Work Out of Robotics and AV Development
Engineers get tested configurations for the software they already use, so time goes into the work instead of the plumbing
- Grammarly — Bug fixes and performance optimization
- Higgsfield AI — Kling 4.0
Kling 4.0 is now on Higgsfield for generating videos from text, images, or references and editing finished clips.
- Meshy — Added the X-Meshy-Metadata header for Studio and Enterprise teams to every endpoint that creates a task, except Creative Lab endpoints.
Added the X-Meshy-Metadata header to task-creating endpoints; read key-value pairs in the metadata field of Usage records.
- LiteLLM — LiteLLM v1.104.5 counts Claude tokens on Bedrock Mantle when Bedrock Runtime cannot
Token counts for newer Claude models on Bedrock come from the Bedrock Mantle endpoint instead of a local estimate that undercounts.
- LiteLLM — LiteLLM v1.106.2, v1.105.3 and v1.103.7 fix spend on failed requests and batch cost
Completed batches keep their spend through malformed results and temporary accounting failures, and Bedrock inference profiles are priced.
- Tonic Textual — v499
Bug fixes.
- Tonic Textual — v500
Fixed incorrect text sizes and source styles in V5 PDF synthesis for unembedded fonts, numeric text, and overlapping text layers.
- Harness AI (AIDA) — Bring Your Own IdP to Harness MCP | OAuth Proxy Mode
Harness MCP OAuth proxy mode lets enterprise developers use their own identity provider without static tokens or shared credentials.