feat(dotAI): Dot AI LangChain4J - Google AI Gemini

[ihoffmann-dot]

Summary

Adds Google AI (Gemini API / AI Studio) as a supported provider under the identifier
google_ai — distinct from the existing vertex_ai, which reaches the same model
family through Google Cloud. This is the consumer-facing Gemini API: a single
AI Studio API key, no GCP project/location/service-account setup.

Full model-type coverage — the only provider besides OpenAI with all four:

• Add langchain4j-google-ai-gemini dependency (version via langchain4j BOM)
• Add GoogleAiGeminiModelProviderStrategy registered in LangChain4jModelFactory.STRATEGIES
• Chat + streaming chat via GoogleAiGeminiChatModel / GoogleAiGeminiStreamingChatModel
• Embeddings via GoogleAiEmbeddingModel — dimensions maps to outputDimensionality
• Image generation via GoogleAiGeminiImageModel — size maps to imageSize (1K/2K)
• Add 11 unit tests in LangChain4jModelFactoryTest

Configuration

{
  "chat": {
    "provider": "google_ai",
    "model": "gemini-2.0-flash",
    "apiKey": "AIza...",
    "maxTokens": 4096,
    "temperature": 0.7
  },
  "embeddings": {
    "provider": "google_ai",
    "model": "gemini-embedding-001",
    "apiKey": "AIza...",
    "dimensions": 768
  },
  "image": {
    "provider": "google_ai",
    "model": "gemini-2.5-flash-image",
    "apiKey": "AIza...",
    "size": "1K"
  }
}

Notes

• API key comes from Google AI Studio — no GCP project required.
• size for image uses Gemini's format (1K, 2K), not OpenAI's 1024x1024.
• maxRetries is not supported by the streaming chat model and is ignored with a warning (same behavior as the OpenAI provider).
• endpoint optionally overrides the default base URL on all three model types.

Related Issue

This PR fixes #36134
EPIC: dotAI Multi-Provider Support #33970

[claude]

Claude finished @ihoffmann-dot's task in 1m 15s —— View job

---

Rollback Safety Analysis

• Read rollback-unsafe categories reference
• Get full PR diff
• Analyze diff against all categories
• Apply label

Verdict: ✅ Safe to Roll Back

All changes in this PR are purely additive — no database schema, Elasticsearch mapping, content model, REST API contract, or VTL viewtool changes are involved.

Category	Result
C-1 Structural Data Model Change	✅ No DB schema changes
C-2 Elasticsearch Mapping Change	✅ No ES mapping changes
C-3 Content JSON Model Version Bump	✅ No contentlet model changes
C-4 DROP TABLE / DROP COLUMN	✅ No destructive DDL
H-1 One-Way Data Migration	✅ No data migrations
H-2 RENAME TABLE / RENAME COLUMN	✅ No renames
H-3 PK Restructuring	✅ No PK changes
H-4 New ContentType Field Type	✅ No new field types
H-5 Storage Provider Change	✅ No storage provider changes
H-6 DROP PROCEDURE / FUNCTION	✅ No procedure drops
H-7 NOT NULL Column Without Default	✅ No column additions
H-8 VTL Viewtool Contract Change	✅ No viewtool changes
M-1 Non-Broadening Column Type Change	✅ No type changes
M-2 Push Publishing Bundle Format	✅ No bundle format changes
M-3 REST / GraphQL API Contract Change	✅ No API contract changes
M-4 OSGi Plugin API Breakage	✅ No OSGi interface changes

What changed:

• New Maven dependency langchain4j-google-ai-gemini (version from BOM)
• New GoogleAiGeminiModelProviderStrategy class (package-private, internal)
• STRATEGIES list in LangChain4jModelFactory extended with the new strategy
• Javadoc updates in ProviderConfig.java and LangChain4jModelFactory.java
• 11 new unit tests

Rolling back to N-1 simply means the google_ai provider identifier is unrecognized — existing configurations using other providers are unaffected, and no stored data or index state is changed.

[github-actions]

🤖 Codex Review — openai.gpt-5.5

**[> [🟡 Medium] dotCMS/pom.xml:517 — New dependency is added directly to dotCMS/pom.xml without showing dependency management in bom/application/pom.xml. dotCMS convention requires dependency versions to be managed in the BOM; if this artifact is not already managed there, Maven builds will fail with a missing version.

---

_{Run: #27393353859 · tokens: in: 4253 · out: 597 (reasoning: 516) · total: 4850}