v0.0.3

What's Changed

• Fix source node ID generation when transforming via bundler by @c9r in #4
• Render stylized markdown output in the REPL by @c9r in #5
• chore: Migrate docs site to separate repository by @c9r in #6
• Implement JSON schema validation, matching, and subtyping by @c9r in #7
• Generate additional schema metadata by @c9r in #8

Full Changelog: v0.0.2...v0.0.3

v0.0.2

Patch Changes

• #3 f57080e Thanks @c9r! - Add current query to AgentContext and use it as the default Index query param. Modify ToolSource, resolveTool, and resolveTools to support functions that return arrays of tools. Taken together, this enables tool indexes to be directly used as tool sources, with the current LLM prompt automatically used as the tool query.

v0.0.1 - Toolcog first release

Hello, World!

This is our first public release of Toolcog, the AI framework for Tool Augmented Generation (TAG). Within it, you'll find everything needed to turn APIs into LLM-enabled tools.

You can read more about TAG and using Toolcog to build AI tools in our documentation.

This Toolcog release includes:

• API: A minimalistic core library of intrinsic LLM functions
• Compiler: A TypeScript compiler plugin that statically analyzes and transforms intrinsic LLM function calls
• Runtime: A lightweight runtime with plugins for popular AI models and integrations for diverse application environments
• Packaging: Compile-time manifest generation of all tools, schemas, prompts, embeddings, and indexes used by a library or application
• Tooling: Compile-time vector embedding prefetching and bundling
• REPL: An interactive read-eval-print loop that combines TypeScript code execution, tool definition, and tool-enabled AI chat

Key Takeaways

Toolcog is the easiest way to build AI tools from existing software apps and APIs. Take advantage of this initial public release to empower your LLMs with access to external software tools and relevant context. Toolcog significantly improves LLM function calling and enables developers to improve tool usage and reliability for AI apps.