docs: Add SDK CHANGELOG.md for v0.2.0 release

Author: Deepthi-Chand

docs/sdk/CHANGELOG.md

@@ -0,0 +1,93 @@
+# 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).
+
+## [0.2.0] - 2025-11-12
+
+### Added
+
+- **AI Model Execution Support**
+  - `call_model(model_id, input_text, parameters)` - Synchronous model execution via SDK
+  - `call_model_async(model_id, input_text, parameters)` - Asynchronous execution (placeholder)
+  - Support for multiple providers: OpenAI, Llama (Ollama, Together AI, Replicate), HuggingFace, Custom APIs
+  - New API endpoints: `POST /api/aimodels/{id}/call/` and `POST /api/aimodels/{id}/call-async/`
+
+- **HuggingFace Integration**
+  - `hf_use_pipeline` field - Use Pipeline inference API
+  - `hf_auth_token` field - Authentication token for gated models
+  - `hf_model_class` field - Model head specification (8 variants)
+  - `hf_attn_implementation` field - Attention function configuration
+  - `framework` field - PyTorch or TensorFlow selection
+  - Support for 8 model classes: CausalLM, Seq2SeqLM, SequenceClassification, etc.
+
+- **CRUD Operations for AIModels**
+  - `create(data)` - Create new AI models
+  - `update(model_id, data)` - Update existing models
+  - `delete_model(model_id)` - Delete models
+  - PATCH HTTP method support in base client
+
+- **Backend Services**
+  - `ModelAPIClient` - Multi-provider API client with retry logic and statistics tracking
+  - `ModelHFClient` - HuggingFace inference client with pipeline and manual loading support
+  - Automatic GPU/CPU device selection
+  - Request template system with placeholder replacement
+
+- **New Enums**
+  - `AIModelProvider.HUGGINGFACE` - HuggingFace provider
+  - `AIModelFramework` - PyTorch/TensorFlow framework selection
+  - `HFModelClass` - 8 model head types
+
+- **Security & Access Control**
+  - Organization-based model access control
+  - Enhanced authentication checks for private models
+  - Secure API key storage with encryption
+
+### Changed
+
+- Enhanced GraphQL queries to include HuggingFace-specific fields
+- Improved error handling for missing endpoints
+- Updated response format with standardized structure
+
+### Fixed
+
+- Type annotations for full mypy compatibility
+- Async/sync execution context handling
+- Response extraction for various providers
+- Union type issues with AnonymousUser
+
+### Technical
+
+- Full mypy type checking compatibility
+- Comprehensive type annotations throughout codebase
+- Connection pooling with httpx
+- Retry logic with exponential backoff (3 attempts)
+- Endpoint statistics tracking
+- Proper handling of Union types
+
+### Dependencies
+
+- Added `httpx>=0.28.1` - Async HTTP client
+- Added `tenacity>=9.1.2` - Retry logic
+- Added `torch>=2.9.0` - PyTorch support (optional)
+- Added `transformers>=4.57.1` - HuggingFace models (optional)
+- Added `nest-asyncio>=1.6.0` - Nested event loop support
+
+## [0.1.0] - 2025-XX-XX
+
+### Added
+
+- Initial release of DataSpace SDK
+- Basic AIModel CRUD operations (read-only)
+- Dataset management
+- UseCase management
+- Search functionality
+- GraphQL support
+- Keycloak authentication
+- Organization and user management
+- Base API client with authentication
+
+[0.2.0]: https://github.com/CivicDataLab/DataSpaceBackend/compare/v0.1.0...v0.2.0
+[0.1.0]: https://github.com/CivicDataLab/DataSpaceBackend/releases/tag/v0.1.0