ArkStack builds upon Chidori, a reactive runtime for building durable AI agents. It extends Chidori’s capabilities by integrating smolagents, a lightweight library for code-driven AI agents, and adding features like vector store integration, smolagents remote tools support, and enhanced debugging. ArkStack provides a feature rich frameowrk for developers to build, and debug AI agents with ease.
For more details read the docs
ArkStack leverages Chidori’s reactive runtime to orchestrate interactions between agents and their components. It supports Python and JavaScript code execution, enabling developers to build dynamic workflows with pause/resume and time-travel debugging.
With smolagents, developers can create code-driven agents that generate and execute Python code, reducing the number of steps required to complete tasks and improving efficiency.
ArkStack supports vector stores for tasks like retrieval-augmented generation (RAG), semantic search, and memory management. It includes a local vector store for development and integrates with external databases like Pinecone and Weaviate.
ArkStack provides a marketplace and hub for smolagents remote tools, enabling developers to dynamically discover, share, and integrate tools into agent workflows.
ArkStack offers advanced debugging features, including time-travel debugging, execution comparison, and interactive debugging. Comprehensive metrics dashboards provide real-time insights into agent performance.
ArkStack supports hierarchical multi-agent systems, enabling agents to collaborate on complex tasks through managed agents and communication protocols.
ArkStack ensures safe execution of generated code with secure environments, including the local Python interpreter and E2B sandbox (inherited from smolagents :D).
- Reactive subscriptions between nodes
- Branching and time travel debugging, reverting execution of a graph
- Node.js, Python, and Rust support for building and executing graphs
- Simple local vector db for development
- [~] Integrate smolagents library for code generation execution (in-progress)
- Combine Chidori’s time-travel debugging with smolagents’ execution logs for a seamless debugging experience.
- Expand the tool registry to support dynamic tool discovery and registration.
- Analysis tools for comparing executions
- Adding support for more vector databases
- Adding support for other LLM sources
- Adding support for more code interpreter environments
- Port smolagents to TypeScript
- Launch tool marketplace
- Agent re-evaluation with feedback
- Add metrics, dashboards, and alerts for monitoring agent performance and behavior.
- Enable agents to dynamically adapt to changing contexts or environments during execution.
ArkStack is under the MIT license.
Star the GitHub repo to stay updated and contribute!
Our framework is inspired by the work of many others, including:
- chidori - providing the reactive runtime
- smolagents - providing First-class support for Code Agents, i.e. agents that write their actions in code (as opposed to "agents being used to write code")
- Temporal.io - providing reliability and durability to workflows
- Eve - developing patterns for building reactive systems and reducing accidental complexity
- Timely Dataflow - efficiently streaming changes
- Langchain - developing tools and patterns for building with LLMs