---

title: Dune (by polygala-ai) IC Memo company_id: polygala_ai_dune company_name: Dune (by polygala-ai) dd_case_id: korok-orchestrator-workflow:1774322535.242299 canonical_document: ic_memo created_at: '2026-03-24T03:29:03+00:00' last_updated_at: '2026-03-24T03:29:03+00:00' update_history:

• '2026-03-24T03:29:03+00:00 :: Initial creation based on preliminary due diligence analysis.'

---

Canonical DD IC Memo Case ID: korok-orchestrator-workflow:1774322535.242299 Last updated: 2026-03-24T03:29:03+00:00 Update note: Initial creation based on preliminary due diligence analysis.

投资委员会备忘录 (IC Memo)：Dune (by polygala-ai)

⚠️ 本尽职调查报告基于公开信息和 AI 分析生成。它应作为辅助手段，而非替代人类判断、背景调查及与创始人的直接交流。

I. 执行摘要 (Executive Summary)

Dune 是一个由资深数据库专家 Dorian Zheng (前 StarRocks 负责人、腾讯微信数据库专家) 发起的极早期开源 TypeScript 项目。该项目旨在通过底层的沙盒隔离技术，构建更安全、高性能的 AI Agent 和 RAG 工作流框架。虽然团队技术背景极其出色（9/10），但由于产品处于 MVP 前期且市场竞争异常激烈，目前的结论是：保持观望 (Watch)，重点跟踪其能否将底层技术优势转化为开发者生态。

II. 公司概况 (Company Overview)

• 项目名称：Dune (由 Polygala-ai 开发)
• 定位：用于构建 AI Agent 和复杂 RAG 工作流的 TypeScript 框架。
• 差异化卖点：本地优先 (Local-first)、硬件级沙盒隔离、强类型任务编排。

III. 团队 (Team)

• 核心创始人：Dorian Zheng。
• 技术履历：国内顶尖的基础架构专家。在 StarRocks 任职 Team Lead 期间积累了深厚的分布式系统和高性能数据库经验。在微信期间担任数据库专家工程师。
• 综合评估：团队在“底层基础架构”层面的技术实力处于国内顶尖水平。目前的挑战在于如何跨越从“底层工具”到“开发者应用层框架”的商业化鸿沟。

IV. 市场机会 (Market Opportunity)

• 赛道背景：随着 LLM 向 Agent 演进，开发者急需更可控、更安全的工作流编排工具。
• 市场规模：Agent 编排层正处于爆发期，虽然目前 LangChain 占据主导，但在生产级稳定性、本地安全性和 TS 原生支持上仍有巨大空缺。

V. 产品与技术 (Product & Technology)

• 技术底座：极大可能集成了创始人的另一开源项目 BoxLite。BoxLite 是一个高性能的轻量级沙盒隔离环境（GitHub 1.5k+ Stars）。
• 产品现状：目前是一个极简的 TS 库，允许开发者定义 Agent 状态机和任务流。
• 核心优势：相比纯应用层框架，Dune 具备更底层的硬件/沙盒控制能力，适合处理敏感数据或需要严格执行环境隔离的任务。

VI. 竞争格局 (Competitive Landscape)

• 头部玩家：LangChain.js, Vercel AI SDK, Mastra, CrewAI。
• 竞争挑战：Dune 目前缺乏文档生态、预置组件库以及广泛的社区认可。其“沙盒隔离”能否成为刚需，尚需市场验证。

VII. 牵引力与财务概况 (Traction & Financial Overview)

• GitHub 表现：22 Stars, 4 Forks (截至 2026 年 3 月)。
• 活跃度：极其活跃。单人 30 天内完成 105 次 Commit。
• 财务状况：极早期，目前以开源社区探索为主，尚未有公开的融资记录。

VIII. 投资逻辑 (Investment Merits)

• 一流人才：创始人具备打造顶级开源项目（StarRocks）的成功经验。
• 技术护城河：BoxLite 带来的底层隔离能力是其差异化竞争的核心，难以被应用层框架轻易复制。

IX. 风险因素与对冲 (Risk Factors & Mitigants)

• 关键人物风险：目前处于“超级单兵”状态。对冲建议：观察其能否在 3 个月内招募到首批核心代码贡献者。
• 赛道过热风险：Agent 框架层可能面临洗牌。对冲建议：关注其在特定垂直行业（如金融、网络安全）的切入效果。

X. 估值与收益分析 (Valuation & Returns Analysis)

• 当前估值：种子轮前阶段，估值溢价将主要来自创始人的过往成功履历。
• 潜在回报：若能成为 TS 生态下的主流 Agent 基础设施，具有极高的战略价值。

XI. 建议与下一步行动 (Recommendation & Next Steps)

• 最终建议：观望 (Watch)。
• 下一步行动：
  1. 跟踪 BoxLite 在 Dune 中的实际落地进度。
  2. 观察其 GitHub 增长曲线，看是否能在 3-6 个月内突破 500 Stars 门槛。
  3. 与创始人建立初步联系，了解其商业化路线图。

Investment Committee Memo (IC Memo): Dune (by polygala-ai)

⚠️ This due diligence report is generated from publicly available information and AI analysis. It should supplement, not replace, human judgment, reference calls, and direct founder interactions.

I. Executive Summary

Dune is an extremely early-stage open-source TypeScript project initiated by veteran database expert Dorian Zheng (former StarRocks Team Lead, Tencent WeChat Database Expert). The project aims to build a more secure and high-performance framework for AI Agents and RAG workflows using underlying sandbox isolation technology. While the team’s technical pedigree is exceptional (9/10), given the early MVP stage and intense market competition, the current recommendation is to Watch. We will track whether their technical moat can translate into developer adoption.

II. Company Overview

• Project Name: Dune (developed by Polygala-ai)
• Positioning: A TypeScript framework for orchestrating AI Agents and complex RAG workflows.
• Differentiators: Local-first architecture, hardware-level sandbox isolation, and strongly typed task orchestration.

III. Team

• Key Founder: Dorian Zheng.
• Technical Background: A premier infrastructure expert. He accumulated deep experience in distributed systems and high-performance databases as a Team Lead at StarRocks and as a Database Expert at Tencent WeChat.
• Assessment: The team possesses top-tier technical strength in "low-level infrastructure." The current challenge is crossing the commercial chasm from building "infra tools" to building "application-layer developer frameworks."

IV. Market Opportunity

• Sector Context: As LLMs evolve toward Agents, developers urgently need more controllable and secure workflow orchestration tools.
• Market Size: The Agent orchestration layer is exploding. While LangChain is dominant, there remain significant gaps in production stability, local security, and native TS support.

V. Product & Technology

• Tech Stack: Highly likely to integrate BoxLite, the founder's other open-source project (a high-performance, lightweight sandbox environment with 1.5k+ GitHub Stars).
• Current State: A minimal TS library allowing developers to define Agent state machines and task flows.
• Core Edge: Compared to pure application-layer frameworks, Dune offers deeper hardware/sandbox control, making it suitable for sensitive data processing or tasks requiring strict environment isolation.

VI. Competitive Landscape

• Main Competitors: LangChain.js, Vercel AI SDK, Mastra, CrewAI.
• Challenges: Dune currently lacks a documentation ecosystem, pre-built component libraries, and broad community recognition. Whether "sandbox isolation" becomes a must-have remains to be proven.

VII. Traction & Financial Overview

• GitHub Metrics: 22 Stars, 4 Forks (as of March 2026).
• Activity level: Extremely high. Solo execution of 105 commits in the last 30 days.
• Financials: Extremely early stage; currently focused on open-source exploration with no public funding records.

VIII. Investment Merits

• Tier-1 Talent: The founder has a proven track record of building top-tier open-source projects (StarRocks).
• Technical Moat: The low-level isolation capabilities provided by BoxLite are a core differentiator that is difficult for application-layer frameworks to replicate.

IX. Risk Factors & Mitigants

• Key-Person Risk: Currently a "solo player" operation. Mitigant: Monitor the ability to recruit the first cohort of core contributors within 3 months.
• Market Saturation: The Agent framework layer faces potential consolidation. Mitigant: Focus on traction in specific verticals like finance or cybersecurity where isolation is critical.

X. Valuation & Returns Analysis

• Current Valuation: Pre-seed stage. Valuation premium will primarily drive from the founder's previous successes.
• Potential Upside: High strategic value if it becomes the dominant Agent infrastructure in the TS ecosystem.

XI. Recommendation & Next Steps

• Final Verdict: Watch.
• Next Steps:
  1. Track the actual integration progress of BoxLite into the Dune runtime.
  2. Monitor GitHub growth curves, looking for a 500+ Stars milestone in the next 3-6 months.
  3. Establish initial contact with the founder to understand the commercialization roadmap.