# Changelog

## v0.1.0 — Complete Implementation

### Breaking: new module hierarchy
All modules reorganized to match HuggingFace's service groupings:

- `HuggingfaceClient.Inference.*` — all inference (was `HuggingfaceInference.*`)
- `HuggingfaceClient.Hub.*` — all hub APIs (was `HuggingfaceHub.*`)
- `HuggingfaceClient.Jinja` — templates (was `HuggingfaceJinja`)
- `HuggingfaceClient.Agent` — agents (was `HuggingfaceAgent`)
- `HuggingfaceClient.Error.*` — errors (top-level, shared)
- `HuggingfaceClient.Config` — env vars & config

### New service folders
```
hub/repositories/   repos, commits, files, snapshots, tags
hub/models/         Models Hub
hub/datasets/       Datasets Hub + Viewer
hub/spaces/         Spaces management
hub/discovery/      search, collections, papers
hub/collaboration/  discussions, PRs, webhooks, webhooks_server
hub/storage/        buckets, cache, filesystem (hf://)
hub/compute/        jobs, autotrain
hub/deployment/     inference endpoints
hub/enterprise/     organizations, gated_repos, oauth
hub/community/      users
hub/metadata/       model cards
hub/evaluation/     evaluate, leaderboards
```

### New modules
- `HuggingfaceClient.Inference.TGI` — Text Generation Inference server client
- `HuggingfaceClient.Hub.AutoTrain` — AutoTrain training API
- `HuggingfaceClient.Hub.Leaderboards` — Hub leaderboards
- `HuggingfaceClient.Hub.Evaluate` — metrics computation
- `HuggingfaceClient.Hub.Search` — full-filter model/dataset/space search
- `HuggingfaceClient.Hub.Organizations` — org member management
- `HuggingfaceClient.Hub.FileSystem` — `hf://` URI filesystem interface
- `HuggingfaceClient.Hub.WebhooksServer` — webhook payload verification
- `HuggingfaceClient.Config` — environment variable management

## v0.2.0 — Production-grade release
- Renamed from `huggingface_inference` to `huggingface_client`
- Added: Jobs, Buckets, Discussions, Webhooks, Gated repos, Inference Endpoints
- Added: Papers, Spaces, Users, Tags, Dataset Viewer, Cache, Metadata, Collections
- Added: HfFileSystem, AutoTrain, Leaderboards, Evaluate, Organizations
- Jinja2: 78/78 tests passing

## v0.1.0 — Initial release
- 34 inference tasks, 25+ providers
- Basic Hub API (repos, files, models, datasets)
- Jinja2 engine, MCP Agent