# Changelog

All notable changes to this project will be documented in this file.

The format is based on [Keep a Changelog](https://keepachangelog.com/en/1.0.0/),
and this project adheres to [Semantic Versioning](https://semver.org/spec/v2.0.0.html).

## [1.0.6] - 2025-10-23

### Fixed
- Explicitly exclude `body` property from GET and HEAD requests to prevent HTTP 422 errors
- Build fetch options more carefully to avoid accidentally including request body in GET requests

## [1.0.5] - 2025-10-23

### Fixed
- Fixed HTTP 422 errors on result endpoint by not setting Content-Type header for GET requests
- Only set Content-Type: application/json for POST, PUT, and PATCH methods

## [1.0.4] - 2025-10-23

### Fixed
- Fixed HTTP 405 errors on result/status endpoints by using URL-based API pattern
- Updated tool signatures to match Python implementation and fal.ai queue API design

### Changed
- **BREAKING**: `result`, `status`, and `cancel` tools now accept a `url` parameter instead of `app_id` + `request_id`
- `generate` tool now returns complete queue response including `response_url`, `status_url`, and `cancel_url`
- Response status values now use fal.ai standard: "IN_QUEUE", "IN_PROGRESS", "COMPLETED"
- Cancel endpoint now uses PUT method instead of POST (matches API requirements)

### Migration
If you were using:
```javascript
const gen = await generate({app_id: "fal-ai/flux/dev", input_data: {...}});
const result = await result({app_id: "fal-ai/flux/dev", request_id: gen.request_id});
```

Update to:
```javascript
const gen = await generate({app_id: "fal-ai/flux/dev", input_data: {...}});
const result = await result({url: gen.response_url});
const status = await status({url: gen.status_url});
```

## [1.0.3] - 2025-10-23

### Fixed
- Fixed search functionality to use the correct API endpoint with `keywords` parameter
- Fixed response structure to handle `items` field from fal.ai API (was incorrectly expecting `models`)
- Updated schema endpoint to use the correct OpenAPI endpoint that matches Python implementation
- Fixed TypeScript build errors related to duplicate property declarations

### Changed
- Search now uses server-side filtering via `keywords` parameter instead of client-side filtering
- All model discovery tools now properly transform API response structure

## [1.0.2] - 2025-10-23

### Changed
- Implemented client-side filtering for search (temporary solution, now fixed in 1.0.3)

## [1.0.1] - 2025-10-23

### Fixed
- Added explicit GET method to result and status endpoints
- Fixed schema endpoint to use full app_id directly
- Added default GET method to publicRequest
- Fixed undefined args parameter in tool handlers

## [1.0.0] - 2025-10-23

### Added
- Initial release of TypeScript MCP server for fal.ai
- Model discovery tools: `models`, `search`, `schema`
- Generation tools: `generate`, `result`, `status`, `cancel`
- Storage tool: `upload`
- Full TypeScript support with type definitions
- Comprehensive documentation and examples
- Support for Claude Desktop, Cursor, and other MCP clients
- Environment-based API key configuration

### Features
- List and search fal.ai model gallery
- Get detailed model schemas
- Generate content using AI models (images, videos, etc.)
- Queue management for long-running generations
- File upload to fal.ai CDN
- Proper error handling and validation

### Technical
- Built with @modelcontextprotocol/sdk
- Uses @fal-ai/client for API interactions
- ES2022 target with Node 18+ support
- TypeScript 5.7+ with strict mode
- Source maps and type declarations included