Data

Tools

Indexes

Applications

Experiments

Open Source Science

Recent Releases of https://github.com/kruskal-labs/toolfront

https://github.com/kruskal-labs/toolfront - v0.2.10

Launch a dedicated docs site at docs.toolfront.ai

- Python
Published by antidmg 6 months ago

https://github.com/kruskal-labs/toolfront - v0.2.9: better type hinting with `Table`

Users can now specify

python answer: db.Table = db.ask(...)

or

```python class Row(BaseModel): name: str email: str phone_number: int

answer: db.Table[Row] = db.ask(...) ```

- Python
Published by antidmg 7 months ago

https://github.com/kruskal-labs/toolfront - v0.2.8

What's Changed

  • Improved error handling for optional dependencies
  • General code cleanup and improvements

Full Changelog: https://github.com/kruskal-labs/toolfront/compare/v0.2.4...v0.2.8

- Python
Published by antidmg 7 months ago

https://github.com/kruskal-labs/toolfront - v0.2.1

What's Changed

  • fix: DataFrame type hint execution (#31)

Bug Fixes

  • Fixed DataFrame type hint execution issues

Internal Changes

  • Refactored Docker build/push to separate script for better release management

Full Changelog: https://github.com/kruskal-labs/toolfront/compare/v0.2.0...v0.2.1

- Python
Published by antidmg 7 months ago

https://github.com/kruskal-labs/toolfront - ToolFront SDK

Changelog

v0.2.0 - ToolFront SDK

Summary

This is a major refactoring of the architecture that exposes ToolFront as an SDK instead of a standalone MCP server. This makes ToolFront's functionality much more accessible, composable, and powerful.

Key Features

  • Built-in ask() method with streaming support across all data sources

  • Structured Outputs: Type-annotated responses automatically structure data (primitives, Pydantic models, DataFrames)

  • Backward Compatibility: Existing MCP server functionality preserved

  • Enhanced Database Support: Added 15+ new database connectors (Athena, ClickHouse, Trino, etc.) powered by Ibis, which is a more battle-tested library for data connectivity and gives us support for more diverse datasources out-of-the-box

Architecture Changes

  • Extracted core MCP tooling logic into reusable SDK components

  • Added direct LLM integration via Pydantic-AI

  • Maintained MCP server as thin wrapper around SDK

Migration Path

  • MCP users: No breaking changes, existing configurations work as-is

  • New users: Can use SDK directly without MCP knowledge

  • Developers: Can compose ToolFront into existing applications

- Python
Published by antidmg 7 months ago

https://github.com/kruskal-labs/toolfront - Release v0.1.11

Fixes a minor bug with API connection caching introduced in v0.1.10.

(notes from v0.1.10):

Document Library Support

Added document library functionality to ToolFront, enabling AI agents to search and read documents alongside database and API operations.

Supported Document Types: - PDF (.pdf) - with smart pagination - Microsoft Office - Word (.docx), PowerPoint (.pptx), Excel (.xlsx, .xls) - Text formats - JSON, Markdown, TXT, XML, YAML/YML, RTF

Key Features: - Intelligent Pagination: Percentile-based navigation for large documents - Performance Optimized: Caching layer for improved response times - No External Dependencies: BM25 implementation without embeddings

🔧 Technical Improvements

  • Modular Architecture: Separated tools into API, database, and library modules
  • Improved Caching: Improved performance for schema discovery and document operations

📦 Installation

Document support is available as an optional dependency:

```bash

Install with document support

uv pip install -e ".[document]" ```

🔧 Usage

Add a document library to your ToolFront configuration:

```bash

Point to any directory containing documents

file:///path/to/your/documents

Example: Documents in your Downloads folder

file:///Users/username/Downloads ```

🐛 Bug Fixes

  • Fixed various edge cases in search algorithms
  • Improved error handling for malformed documents
  • Better connection management for large file operations

This release expands ToolFront's capabilities, allowing AI agents to work with both structured data (databases) and unstructured data (documents) in a unified interface.

Full Changelog: https://github.com/kruskal-labs/toolfront/compare/v0.1.9...v0.1.10

- Python
Published by antidmg 7 months ago

https://github.com/kruskal-labs/toolfront - 🚀 Release v0.1.10

Document Library Support

Added document library functionality to ToolFront, enabling AI agents to search and read documents alongside database and API operations.

Supported Document Types: - PDF (.pdf) - with smart pagination - Microsoft Office - Word (.docx), PowerPoint (.pptx), Excel (.xlsx, .xls) - Text formats - JSON, Markdown, TXT, XML, YAML/YML, RTF

Key Features: - Intelligent Pagination: Percentile-based navigation for large documents - Performance Optimized: Caching layer for improved response times - No External Dependencies: BM25 implementation without embeddings

🔧 Technical Improvements

  • Modular Architecture: Separated tools into API, database, and library modules
  • Improved Caching: Improved performance for schema discovery and document operations

📦 Installation

Document support is available as an optional dependency:

```bash

Install with document support

uv pip install -e ".[document]" ```

🔧 Usage

Add a document library to your ToolFront configuration:

```bash

Point to any directory containing documents

file:///path/to/your/documents

Example: Documents in your Downloads folder

file:///Users/username/Downloads ```

🐛 Bug Fixes

  • Fixed various edge cases in search algorithms
  • Improved error handling for malformed documents
  • Better connection management for large file operations

This release expands ToolFront's capabilities, allowing AI agents to work with both structured data (databases) and unstructured data (documents) in a unified interface.

Full Changelog: https://github.com/kruskal-labs/toolfront/compare/v0.1.9...v0.1.10

- Python
Published by antidmg 7 months ago

https://github.com/kruskal-labs/toolfront - v0.1.9: Standardize cache directory

This release standardizes the cache directory path for improved consistency.

- Python
Published by antidmg 8 months ago

https://github.com/kruskal-labs/toolfront - v0.1.8

Added OpenAPI specification support, authentication routing, secret obfuscation, and performance optimizations

- Python
Published by antidmg 8 months ago

https://github.com/kruskal-labs/toolfront - v0.1.7

Changelog

Added

  • API endpoint support for database access via REST APIs given any OpenAPI or Swagger specification.

Fixed

  • Secret obfuscation in database connection URLs
  • OpenAPI endpoint discovery functionality

- Python
Published by antidmg 8 months ago

https://github.com/kruskal-labs/toolfront - Release v0.1.6

Optional Database Dependencies

  • Database drivers are now optional extras, keeping the core package lightweight
  • Install only what you need: uvx "toolfront[postgresql,mysql]"
  • Use [all] to install all supported database drivers
  • Added PostgreSQL aliases: postgres, psql

This architectural change reduces installation size and complexity for users who only need specific databases.

- Python
Published by antidmg 8 months ago

https://github.com/kruskal-labs/toolfront - v0.1.2

Fixed bug where MCP server database connections with passwords were failing due to URL parsing issue.

- Python
Published by antidmg 8 months ago