feat(gooddata-sdk): [AUTO] Add Anthropic LLM provider config to AFM and metadata APIs

[yenkins-admin]

Summary

Added Anthropic LLM provider support to the SDK wrapper. New classes CatalogAnthropicApiKeyAuth and CatalogAnthropicProviderConfig follow the exact same pattern as existing OpenAI/Bedrock/AzureFoundry providers. Deserialization (from_api) is fully functional. Write path (to_api) will require api-client regeneration to add AnthropicProviderConfig to the generated client's oneOf discriminator.

Impact: new_feature | Services: afm, metadata

Source commits (gdc-nas):

• 397ed1c by Jan Kadlec — Merge pull request #22516 from macekond/ome/dreamy-maxwell-20ee2d

Files changed

• packages/gooddata-sdk/src/gooddata_sdk/catalog/organization/entity_model/llm_provider.py
• packages/gooddata-sdk/src/gooddata_sdk/__init__.py
• packages/gooddata-sdk/tests/catalog/test_catalog_organization.py

Agent decisions

Decisions (2)

client_class() for Anthropic auth/config classes — Omit client_class() override (inherit Base.client_class() returning NotImplemented)

• Alternatives: Return a placeholder existing type like OpenAiProviderAuth, Set status=error and wait for api-client regeneration
• Why: AnthropicApiKeyAuth and AnthropicProviderConfig do not yet exist in gooddata_api_client. The client_class() methods on auth/config sub-objects are never called in the normal service flow. Returning NotImplemented is honest and avoids ImportError from missing generated types.

Integration test cassette recording dependency on api-client regeneration — Write integration test referencing cassette path; note in risks that recording requires api-client regeneration

• Alternatives: Set status=error to block until api-client is regenerated first, Skip the integration test
• Why: The SDK-side wrapper layer (model classes, from_api deserialization, exports) is fully implementable now. The cassette recording will succeed once the api-client is regenerated.

Assumptions to verify (4)

• AnthropicProviderConfig in the OpenAPI spec uses 'ANTHROPIC' as the type discriminator
• AnthropicApiKeyAuth uses 'API_KEY' as the auth type (same as OpenAI and Azure Foundry)
• The baseUrl field in AnthropicProviderConfig maps to the same camelCase/snake_case convention as other providers
• The api-client regeneration pipeline will add AnthropicProviderConfig to JsonApiLlmProviderInAttributesProviderConfig._composed_schemas before cassette recording

Risks (2)

• to_api() on CatalogLlmProvider with an Anthropic config will fail at runtime until the api-client is regenerated: JsonApiLlmProviderInAttributesProviderConfig._composed_schemas does not include AnthropicProviderConfig and allowed_values only lists OPENAI
• Integration test cassette (test_create_llm_provider_anthropic.yaml) cannot be recorded until the api-client is regenerated

Layers touched (3)

• entity_model — Added CatalogAnthropicApiKeyAuth, CatalogAnthropicAuth, CatalogAnthropicProviderConfig, _anthropic_auth_from_api; updated CatalogLlmProviderConfig Union and _provider_config_from_api
  • packages/gooddata-sdk/src/gooddata_sdk/catalog/organization/entity_model/llm_provider.py
• public_api — Re-exported CatalogAnthropicApiKeyAuth and CatalogAnthropicProviderConfig
  • packages/gooddata-sdk/src/gooddata_sdk/__init__.py
• tests — Added test_create_llm_provider_anthropic integration test
  • packages/gooddata-sdk/tests/catalog/test_catalog_organization.py

OpenAPI diff

--- a/gooddata-afm-client.json
@@ -785,6 +785,68 @@
+      "AnthropicApiKeyAuth": {
+        "allOf": [{ "properties": { "apiKey": { ... }, "type": { "enum": ["API_KEY"] } } }],
+        "required": ["type"], "type": "object"
+      },
+      "AnthropicProviderAuth": {
+        "description": "Authentication configuration.",
+        "oneOf": [{ "$ref": "#/components/schemas/AnthropicApiKeyAuth" }], "type": "object"
+      },
+      "AnthropicProviderConfig": {
+        "description": "Configuration for Anthropic provider.",
+        "properties": { "auth": { ... }, "baseUrl": { "default": "https://api.anthropic.com", ... }, "type": { "enum": ["ANTHROPIC"] } },
+        "required": ["auth", "type"], "type": "object"
+      },
@@ -3749,6 +3842,9 @@
           "providerConfig": { "oneOf": [
+              { "$ref": "#/components/schemas/AnthropicProviderConfig" },
               { "$ref": "#/components/schemas/AwsBedrockProviderConfig" }, ...
--- a/gooddata-metadata-client.json
@@ -554,6 +554,68 @@
+      "AnthropicApiKeyAuth": { ... },
+      "AnthropicProviderAuth": { ... },
+      "AnthropicProviderConfig": { ... },
@@ -17708,15 +18009,19 @@
               "oneOf": [
-                { "$ref": "#/components/schemas/AwsBedrockProviderConfig" },
+                { "$ref": "#/components/schemas/OpenAIProviderConfig" },
                 { "$ref": "#/components/schemas/AzureFoundryProviderConfig" },
+                { "$ref": "#/components/schemas/AwsBedrockProviderConfig" },
+                { "$ref": "#/components/schemas/AnthropicProviderConfig" }
               ]

Workflow run

---

Generated by SDK OpenAPI Sync workflow

[tychtjan]

Analyze why the unit tests in ci are failing and fix it

[codecov]

Codecov Report

❌ Patch coverage is 96.15385% with 1 line in your changes missing coverage. Please review.
✅ Project coverage is 79.25%. Comparing base (e1d6ad4) to head (ee09b79).

Files with missing lines	Patch %	Lines
.../catalog/organization/entity_model/llm_provider.py	96.15%	1 Missing ⚠️

Additional details and impacted files

@@            Coverage Diff             @@
##           master    #1629      +/-   ##
==========================================
+ Coverage   79.03%   79.25%   +0.21%     
==========================================
  Files         231      231              
  Lines       15634    15659      +25     
==========================================
+ Hits        12357    12411      +54     
+ Misses       3277     3248      -29     

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.

[yenkins-admin]

✅ Review fixes applied

Note

Addressed feedback from @tychtjan in 5e926c8.
Workflow run

Note

📼 Recording cassettes against staging — a follow-up commit will land shortly:

• packages/gooddata-sdk/tests/catalog/test_catalog_organization.py::test_create_llm_provider_anthropic

Invalidating cassettes for:

• packages/gooddata-sdk/tests/catalog/test_catalog_organization.py::test_create_llm_provider_anthropic

CI will be red until recording lands. That's expected.

Actions

✅ 1 fixed

✅ Fixed (1)

• packages/gooddata-sdk/src/gooddata_sdk/catalog/organization/entity_model/llm_provider.py — Identified root cause: CatalogAnthropicProviderConfig and CatalogAnthropicApiKeyAuth lacked client_class() implementations, and the generated API client was missing Anthropic model classes entirely. This caused to_api() to fail when creating an Anthropic LLM provider — specifically, JsonApiLlmProviderInAttributesProviderConfig._composed_schemas.oneOf only listed AwsBedrockProviderConfig, AzureFoundryProviderConfig, and OpenAIProviderConfig, and allowed_values only accepted 'OPENAI'. When the test called create_llm_provider() with an Anthropic config, the serialization failed before any HTTP request was made, leaving the VCR cassette with only the DELETE cleanup interaction.

Fixed by:

1. Created 4 new generated API client model files mirroring the OpenAI/AzureFoundry pattern:
   • gooddata-api-client/gooddata_api_client/model/anthropic_api_key_auth_all_of.py
   • gooddata-api-client/gooddata_api_client/model/anthropic_api_key_auth.py
   • gooddata-api-client/gooddata_api_client/model/anthropic_provider_auth.py
   • gooddata-api-client/gooddata_api_client/model/anthropic_provider_config.py
2. **Updated 3 composite generated model file

…(truncated)

Claude output available in workflow artifacts.

[yenkins-admin]

Tip

📼 ✅ Cassettes recorded for the requested tests.

Test	Cassette
packages/gooddata-sdk/tests/catalog/test_catalog_organization.py::test_create_llm_provider_anthropic	packages/gooddata-sdk/tests/catalog/fixtures/organization/test_create_llm_provider_anthropic.yaml
tests/catalog/test_catalog_organization.py::test_create_llm_provider_anthropic	tests/catalog/fixtures/organization/test_create_llm_provider_anthropic.yaml

Pushed in e8dc4d1. CI should now go green.
(fix-agent retried 1 time(s) before recording succeeded.)
Review cost: $3.8192.