エージェントとマルチエージェント

OpenViking: Self-evolving Context Database for AI Agents

volcengine/openviking

OpenViking is an open-source context database that unifies agent memory, knowledge RAG, and skills into a single virtual filesystem, enabling AI agents to browse and retrieve context deterministically using a file-system-like protocol.

★ 28Kスター
⑂ 2.2Kフォーク数
459未解決の問題
Python言語
AGPL-3.0ライセンス
Q@project.QualityScore編集部スコア

プロジェクトのスクリーンショット

OpenViking: Self-evolving Context Database for AI Agentsのスクリーンショット OpenViking: Self-evolving Context Database for AI Agentsのスクリーンショット OpenViking: Self-evolving Context Database for AI Agentsのスクリーンショット OpenViking: Self-evolving Context Database for AI Agentsのスクリーンショット OpenViking: Self-evolving Context Database for AI Agentsのスクリーンショット OpenViking: Self-evolving Context Database for AI Agentsのスクリーンショット

概要

OpenViking stores memories, resources, and skills as a unified virtual filesystem under the `viking://` protocol. Instead of querying a black-box vector store, agents use commands like `ls`, `tree`, and `find` to navigate their own context. Content is automatically processed into three tiers—L0 (abstract), L1 (overview), and L2 (details)—and loaded on demand to optimize token usage. Retrieval is performed via directory recursive search, preserving the surrounding context of results and leaving an observable trajectory for debugging. After a session concludes, the system asynchronously extracts user preferences and agent experiences into long-term memory.

主な機能

  • Unified `viking://` filesystem protocol for memories, resources, and skills
  • Three-tier context loading (L0 abstract, L1 overview, L2 details) to reduce token spend
  • Directory recursive retrieval preserving surrounding context
  • Observable retrieval trajectories for debugging
  • Asynchronous session memory extraction
  • OpenViking Studio browser-based playground
  • OpenViking Helper desktop console for local agent setup and session tracing
  • VikingBot AI agent framework

要件、インストール、クイックスタート

Run `pip install openviking --upgrade` to install the package. Then execute `openviking-server init` to run the interactive wizard for provider and model configuration, which writes to `~/.openviking/ov.conf`. Validate the setup with `openviking-server doctor`.

使用方法

Start the server using `openviking-server`. Use the included `ov` CLI client to manage context: `ov status` checks the server, `ov add-resource <url>` ingests data, `ov ls viking://resources/` lists context, `ov tree` visualizes structure, `ov find "query"` performs semantic search, and `ov grep "term"` searches content. To use the VikingBot framework, install with `pip install "openviking[bot]"` and run `openviking-server --with-bot`, then use `ov chat`.

モデルの互換性とユースケース

Supports Volcengine, OpenAI, Codex OAuth, Kimi, GLM, and local Ollama models. Integrations are available for Claude Code, Codex, OpenClaw, Hermes, Cursor, Trae, OpenCode, pi, MCP clients, and LangChain/LangGraph.

ライセンスとリスクに関する注意事項

The main project is licensed under AGPL-3.0. The `crates/ov_cli` component is licensed under Apache 2.0. The `examples` directory is also licensed under Apache 2.0. Third-party components retain their respective original licenses.

Editorial verification 2026-08-09: repository URL, owner, description, license and repository statistics were reviewed. License metadata: AGPL-3.0. README was fetched for the channel draft; re-check repository dependencies, releases and model terms before production use.

リリースとメンテナンス

The repository is actively maintained with 27,964 stars and 459 open issues. The evaluated version mentioned in the README is 0.3.22. The repository was created on 2026-01-05 and last pushed on 2026-08-05.

Firecrawl

firecrawl/firecrawl

★ 161.1KTypeScript

LangChain

langchain-ai/langchain

★ 143.6KPython

RAGFlow

infiniflow/ragflow

★ 86.7KGo