QueryPie AI

Partner Release Dashboard

Release schedules, included improvements, and planned roadmaps for Regular Release, Lingo, and NotePie.

Regular Release(Sprint 14)

AIP features and improvements released regularly by Sprint.

※ Feature scope may change depending on development status until QA begins. Please keep this in mind.

QA
June 9 (Tue) - June 15 (Mon)
Release and manual publication
June 16 (Tue)

MCP Improvements

  • Manually refresh the Custom Remote MCP tool list
  • Reuse Client ID and Secret values for configurable integrations
  • Add MCP tools in chat without Preset configuration during conversations
  • Prepare Kintone MCP integration

Admin Enhancements

  • View the current accumulated remaining credits on the settlement screen
  • Manage external app access settings by partner and organization
  • Manage default and maximum chat recursion limits
  • Manage automatic credit renewal settings
  • Allow organization admins to review usage by agent
  • Bulk activate or deactivate users in Admin > Users

Agent Enhancements

  • Use preconfigured MCP integrations in organization agents
  • Change the agent option maximum output tokens from 100 to 500
  • Support standard chat hints and basic tool specifications in Agent
  • Provide chat-only links dedicated to specific agents
  • Load required tools on demand during conversations, like Claude's on-demand tool loading
  • Set agent sharing scope with tags

My Drive Improvements

  • Safely download Sandbox-generated files from My Drive with user-authenticated links
  • Delete files from the My Drive context menu

UX Improvements

  • Distinguish personal and organization agents in the agent list
  • Simplify low-usage widget save menus
  • Preview My Drive, Skill, and Knowledge content in the viewer
  • Change the file upload limit to 200MB per turn

Security and Administration

  • Securely manage and reference API keys and secrets when running external APIs in Skills

Web Builder Improvements

  • Set customer domains for Web Builder apps
  • Manage and delete published apps in Web Builder
  • Configure Web Builder publishing scope more flexibly

Sandbox Improvements

  • Access local networks from Sandbox through EdgeTunnel

Skill Improvements

  • Call required Skills directly during conversations

Knowledge Improvements

  • Provide embedding param options for Knowledge Bundles

Long-Term Roadmap

Product direction for partner sharing.

QLM (QueryPie Language Model)-centered Auto Router

QA
QA in progress
Release target
Targeting June release

This release introduces Auto Model Router, which automatically selects the most appropriate model based on the user's query. Auto selects the optimal model according to the complexity and characteristics of the request, providing both response quality and cost efficiency without requiring users to choose a model manually.

The default model for Auto Router is QLM (QueryPie Language Model), developed and operated by QueryPie. QLM is an enterprise-grade LLM optimized for the QueryPie environment based on open-source models, and is designed to provide high-quality responses across most general business scenarios.

QLM runs on H200 GPU infrastructure directly operated by QueryPie, so customer prompts, internal company information, and personal data are not used for model training or sent outside. Because QueryPie develops and operates the model directly, QueryPie can take responsibility for and assure the data processing and operating environment.

QLM targets Claude Sonnet-level work performance while providing significantly higher cost efficiency. Auto Router therefore sends most requests to QLM first for fast and economical processing, and uses other models only when needed.

This allows customers to experience optimal AI performance without worrying about model selection, and to operate AI services at more reasonable and predictable cost even as organization-wide AI usage grows.