Reading
Keelson is a locally-run AI agent framework that supports multiple providers (Copilot and Claude), a pluggable Rib extension architecture, and a YAML-based deterministic workflow engine, allowing users to fully control their AI toolchain.
Section 01
Keelson is a locally-run AI agent framework that supports multiple providers (GitHub Copilot, Claude), a pluggable Rib extension architecture, and a YAML-based deterministic workflow engine, allowing users to fully control their AI toolchain. The project is open-source (Apache 2.0 license), currently in the Alpha stage, maintained by Daniel Scholl, and its source code is hosted on GitHub.
Section 02
With the popularity of AI coding assistants like GitHub Copilot and Claude, existing solutions have issues such as being tied to specific cloud services or lacking scalability. Keelson emerged as a response, with its core philosophy being 'Your machine, your rules'—it runs on local laptops, directly calls local CLI without going through hosted services, and supports on-demand expansion via a modular architecture.
Section 03
The name Keelson comes from a nautical term (a longitudinal beam above the keel), metaphorically representing its role as a foundational framework: the Keelson core provides multi-provider support, persistent state, workflow engine, and browser interface; Ribs (as in ship's ribs) are attached functional modules that communicate with the core via standardized interfaces to integrate specific functions.
Section 04
Keelson supports the GitHub Copilot SDK and Claude Agent SDK, and is designed to easily add more providers. It abstracts differences between various SDKs through a unified IAgentProvider interface, allowing upper-layer applications to seamlessly switch underlying providers to adapt to changes in the AI service market (new models, price adjustments, etc.), enhancing flexibility and risk resistance.
Section 05
A Rib is an independent npm package (following the naming convention @keelson/rib-<name>) that implements the Rib interface to integrate with the core. During startup, it automatically scans the node_modules/@keelson/ directory to discover Ribs, which can be filtered and activated via environment variables. Advantages: separation of concerns (lean core), composability (install on demand), maintainability (independent development and version control).
Section 06
Keelson uses YAML to define deterministic workflows, which are human-readable, version-controllable, and auditable. Supported node types include prompt (AI prompt), bash (execute command), loop (loop), approval (manual approval), etc. Dependencies between nodes are defined via depends_on, and data is passed using $nodeId.output. Drawing on the Archon project, it is TypeScript-typed and adapted to the Bun+SQLite stack, suitable for repeatable and compliant scenarios.
Section 07
Keelson follows the local-first principle: state is stored in a local SQLite database, credentials are saved in the system keychain via keytar, and workflows are executed locally. The browser interface only accepts local connections (http://127.0.0.1:5173), and sensitive data never leaves the machine, making it suitable for scenarios involving confidential code, proprietary algorithms, or regulated data.
Section 08
Applicable Scenarios: Multi-provider strategy (flexible switching), sensitive data processing (local execution), auditable workflows (deterministic), custom tool integration (Rib extension), technical exploration (hackable codebase).
Summary: Keelson returns to the essence of tools, with simplicity, locality, and controllability at its core. Through modular and deterministic design, it provides developers with the option to control their AI toolchain, making it attractive in the era of data sovereignty. The project's tech stack includes Bun, React19, TypeScript, and SQLite, and it is currently in the Alpha stage.
Keep going with more reads from the same topic.
Nornir MCP Server is an enterprise-level server based on the Model Context Protocol (MCP). It seamlessly integrates large language models (such as Claude) with the Nornir network automation framework, supporting natural language orchestration for multi-vendor network devices (Cisco, Arista, Juniper, etc.), and providing production-grade features like a dual-engine architecture (NAPALM + Netmiko), intelligent filtering, and a secure sandbox.
Bibliothèque Française LLM is a structured indexing and annotation project for French public domain literature designed specifically for large language models (LLMs). It integrates multiple authoritative sources such as DraCor, Common Corpus, and Wikisource, providing metadata indexing categorized by genre, author, and era, as well as in-depth annotations for dramatic texts (including characters, lines, stage directions, etc.). Its aim is to enable LLMs to efficiently read and understand classic French literary works.
Splinter is a minimalist, high-performance key-value (KV) and vector storage system enabling zero-latency inter-process communication via shared memory and atomic operations. With only 766 lines of core code, it supports millions of operations per second and 768-dimensional vector storage, offering a new architectural approach for local LLM inference and data-intensive applications.
A complete cloud-native AI application development guide for beginners, building a simple generative AI chatbot using Amazon EC2, Apache, Python CGI, and Amazon Bedrock, covering architecture design, IAM permission configuration, security best practices, and cost optimization suggestions.