Agentic Generative BI Agentic Generative BI

GitHub's top-ranked open-source GenBI (Generative BI) platform. It enables users to query databases through natural language, automatically generate SQL, charts, and AI insight reports, and supports multiple data sources and enterprise-grade semantic layer integration. GitHub 排名第一的開源 GenBI(Generative BI)平台。用戶可用自然語言查詢資料庫,自動生成 SQL、圖表與 AI 洞察報告,支援多資料來源與企業級語意層整合。

Natural language to SQL 自然語言轉 SQL

Auto-generated charts and AI insight reports 自動生成圖表與 AI 洞察報告

Enterprise semantic layer integration 企業級語意層整合

From your hardware to business answers. 從您的硬體,到商業解答。

Wren AI's GenBI layer runs directly on the Phison AI Data Platform, so natural-language queries, semantic modeling, and on-prem compute work as one stack — not separate tools you have to wire together.Wren AI 的 GenBI 層直接運行於 Phison AI Data Platform 之上,讓自然語言查詢、語意建模與地端算力整合為單一堆疊,不需自行拼接多套工具。

Demand

People & agents ask 人與 AI 代理提問

Executives, operators, analysts, and AI agents enter through the same governed answer layer. 高階主管、營運人員、分析師與 AI 代理都透過同一個受治理的答案層進行查詢。

Wren role

Business-facing GenBI 面向業務的 GenBI

Wren owns the natural-language experience, semantic model, governance, quality, and auditability. Wren 負責自然語言體驗、語意模型、治理、品質與可稽核性。

Phison role

On-prem AI data platform 地端 AI 資料平台

Phison owns the local runtime: gateways, model services, resource orchestration, storage, and infrastructure. 群聯負責本地運行環境:閘道器、模型服務、資源編排、儲存與基礎設施。

Wren AI with aiDAPTIV Wren AI 搭載 aiDAPTIV

At average 32K input tokens — up to 12× concurrent-user improvement with aiDAPTIV.平均 32K 輸入 Token 下,並行使用者最高 12 倍改善。

Wren AI Appliance (4-GPU) Wren AI 一體機(四卡)
GPU cards GPU 卡數
Model 模型 Qwen3.5-397B-A17B-NVFP4 Qwen3.5-397B-A17B-NVFP4
Input Token 輸入 Token Average 32K 平均 32K
SLA SLA TTFT < 10, TPS > 20 TTFT < 10、TPS > 20
Concurrent Users 並行使用者 12X (3 37)
TTFT TTFT 10X average improvement 平均改善
RTX Pro6000 ×4 Qwen3.5-397B-A17B-NVFP4
Context length 上下文長度
TTFT SLA TTFT SLA
Upper bound 上限 ≤ 10.0 s
TPS SLA TPS SLA
Lower bound 下限 ≥ 20 tok/s
Concurrent users 並行使用者
TTFT

Lower is better 越低越好

TTFT

Lower is better 越低越好

Concurrent users 並行使用者

Token Per Second

Higher is better 越高越好

TPS

Higher is better 越高越好

Concurrent users 並行使用者

Without aiDAPTIV
With aiDAPTIV
SLA threshold
32K context TTFT ≤ 10.0s TP ≥ 20 t/s LLM-only tests; E2E performance is authoritative 僅 LLM 測試;實際效能以端對端為準

Ready to deploy GenBI on Phison? 準備好在群聯平台上部署 GenBI?

Connect your database and start answering business questions in natural language within 10 minutes. 連接您的資料庫,10 分鐘內開始以自然語言回答商業問題。