Skip to content

CLAUDE.md

Latest commit

 

History

History
75 lines (55 loc) · 3.5 KB

CLAUDE.md

File metadata and controls

75 lines (55 loc) · 3.5 KB

CLAUDE.md — LabelStudio SDK

Overview

Auto-generated C# SDK for Label Studio — open-source data labeling and annotation platform for ML with 30k+ GitHub stars. Supports projects, tasks, annotations, ML backends, webhooks, and storage integrations (S3, GCS, Azure, Redis). OpenAPI spec from the official HumanSignal/label-studio-client-generator Fern definition.

Build & Test

dotnet build LabelStudio.slnx
dotnet test src/tests/IntegrationTests/

Test Modes (Testcontainers)

Tests use Testcontainers with automatic environment detection:

  • Debug builds (dotnet test): Connect to local instance at localhost:8080 (requires LABEL_STUDIO_API_KEY)
  • Release/CI builds (dotnet test -c Release): Automatically spin up heartexlabs/label-studio Docker container with preset API token (3-minute startup timeout)

Auth

Token-based auth (Authorization: Token <api_key>):

var client = new LabelStudioClient(apiKey); // LABEL_STUDIO_API_KEY env var

The SDK converts Bearer auth to Label Studio's Token scheme via a PrepareRequest hook.

Key Files

  • src/libs/LabelStudio/openapi.yaml — OpenAPI spec (downloaded from HumanSignal/label-studio-client-generator)
  • src/libs/LabelStudio/generate.sh — Downloads spec, converts apiKey auth to http/bearer, runs autosdk
  • src/libs/LabelStudio/Generated/ — Never edit — auto-generated code (~2695 files)
  • src/libs/LabelStudio/Extensions/LabelStudioClient.Auth.cs — PrepareRequest hook: Bearer → Token
  • src/libs/LabelStudio/Extensions/LabelStudioClient.Tools.cs — MEAI AIFunction tools
  • src/tests/IntegrationTests/Environment.cs — Testcontainers setup (Docker in CI, localhost in debug)
  • src/tests/IntegrationTests/Tests.cs — Assembly-level init/cleanup
  • src/tests/IntegrationTests/Examples/ — Example tests (also generate docs)

Spec Notes

Label Studio uses apiKey auth in the spec with Token prefix. The generate.sh converts it to http/bearer for AutoSDK, and PrepareRequest rewrites Bearer → Token at runtime.

Sub-client Pattern

The API is organized into sub-clients:

  • client.Projects2.ListAsync(...) — List labeling projects
  • client.Projects2.GetAsync(...) — Get project details
  • client.Projects2.CreateAsync(...) — Create a project
  • client.Tasks2.List2Async(...) — List tasks in a project
  • client.Tasks2.GetAsync(...) — Get task details
  • client.Annotations.CreateAsync(...) — Create an annotation
  • client.Annotations.ListAsync(...) — List annotations for a task
  • client.Ml.ListAsync(...) — List ML backends
  • client.Webhooks.ListAsync(...) — List webhooks
  • client.StorageS3.* / client.StorageGcs.* / client.StorageAzure.* — Cloud storage
  • client.Organizations.* — Organization management
  • client.Users.* — User management
  • client.Export.* — Data export
  • client.Import.* — Data import
  • client.Views.* — Data manager views

MEAI

AIFunction tools for use with any IChatClient:

  • AsListProjectsTool() — List labeling projects with search/pagination
  • AsGetProjectTool() — Get project details by ID
  • AsListTasksTool() — List tasks in a project
  • AsCreateAnnotationTool() — Create an annotation on a task

Base URL

Default base URL is http://localhost:8000 (Label Studio is typically self-hosted). Set LABEL_STUDIO_BASE_URL env var or pass a custom baseUri to the constructor for different deployments.