Doubao Agents

Type: app · Vendor: ByteDance · Language: Web product + API · License: proprietary · Status: active · Status in practice: mature

Links: homepage docs

ByteDance's consumer Doubao chatbot and the Volcano Engine Ark API behind it, offering text/image/video/audio model invocation with documented tool-call building blocks: Function Calling, Web Search, Knowledge Base Search, image processing, and structured output (beta).

Description. Doubao is the consumer ByteDance assistant launched August 2023; by November 2024 it was China's most popular AI chatbot with around 60 million MAU and is powered by ByteDance's Volcano Engine cloud, which the company says processes around 120 trillion tokens per day. Developers reach the same Doubao models through Volcano Engine's Ark platform, whose docs catalogue tool-call building blocks (Function Calling, Web Search a.k.a. 联网内容插件, Knowledge Base Search, image processing) alongside deep thinking, streaming and structured output (beta) modes. The international consumer variant is sold as Cici/Dola and is reportedly powered by Microsoft Azure-hosted OpenAI GPT models rather than Doubao itself.

Agent loop shape. Two surfaces. (1) Consumer Doubao chat app: ByteDance-hosted loop, opaque to users — conversation, search, tool invocations all happen server-side. (2) Volcano Engine Ark API: developer-driven loop similar to other model-vendor APIs — send chat completion request with optional tool definitions and built-in plugins (web search, knowledge base), receive content or tool_calls, execute, return, repeat. Ark also offers deep thinking, streaming, structured-output (beta), context caching and file inputs as request-level toggles.

Primary use cases

consumer chat assistant (Chinese market, mobile + web)
tool-augmented Doubao model invocation via Ark API
web-search-grounded answers via 联网内容插件
private knowledge base search and document Q&A

flowchart TD USER[User] --> SURFACE{Surface} SURFACE -->|consumer| APP[Doubao chat app<br/>web / iOS / Android] SURFACE -->|developer| ARK[Volcano Engine Ark API] ARK --> MODEL[Doubao model] MODEL --> TOOLS{Tool options} TOOLS --> FC[Function Calling] TOOLS --> WS[Web Search 联网内容插件] TOOLS --> KB[Knowledge Base Search] TOOLS --> IMG[Image processing] MODEL --> MODES{Request toggles} MODES --> DT[Deep thinking] MODES --> ST[Streaming] MODES --> SO[Structured output beta] MODES --> CC[Context caching]

Key concepts

Doubao chat app (docs) — ByteDance's consumer AI assistant; web, iOS, Android. ~60M MAU in China as of late 2024.
Volcano Engine Ark (火山方舟) (docs) — Developer platform exposing Doubao models with managed inference, fine-tuning and deployment.
Function Calling (函数调用) → tool-use (docs) — Tool-call API listed under 工具调用 in the Ark docs alongside web search and knowledge base search.
Web Search (联网内容插件) → agentic-rag (docs) — Built-in web-search plugin Ark exposes as a tool option for Doubao models.
Knowledge Base Search (私域知识库搜索) → agentic-rag (docs) — Built-in private knowledge-base retrieval tool the Ark platform offers alongside Doubao chat.
Deep thinking + structured output (beta) → structured-output (docs) — Ark request-level toggles for chain-of-thought style reasoning and JSON-shaped outputs.

Patterns this app implements —

References

Provenance

Last analyzed: 2026-05-24
Last updated: 2026-05-24
Verification status: verified