Gemini API Enterprise Development Guide
Helps developers build enterprise Gemini applications on Agent Platform with Google's unified Gen AI SDK.
The skill recommends environment variables or ADC, discourages hard-coded credentials, and identifies enterprise endpoints and some data sources. It does not specify least-privilege IAM roles, sensitive-data handling, user confirmation, isolation of external effects, or rollback, so points are deducted. No red-line risk is evident.
It covers multiple SDKs, models, and workflows and includes a blocked-response handling example. However, the supplied files contain no tests, CI, pinned versions, or reproducibility evidence; some examples have unstated dependencies or uncertain parameter compatibility. Under static calibration it cannot exceed 10, so points are deducted.
The trigger conditions, supported SDKs, and major capabilities are reasonably clear, and the required Google Cloud credentials and API enablement are stated. Non-fit cases, input/output boundaries, regional availability, mainland-China reachability, and cost constraints are insufficiently documented, so points are deducted.
The material is organized by topic and includes installation, migration, reference files, licensing, and a GitHub issue path. It lacks skill-specific versioning, changelog, named maintenance responsibility, stability labels, FAQ, and systematic troubleshooting; dependencies and model versions are partly dynamic or preview-based, so points are deducted.
The quick starts and references cover text, multimodal input, tools, caching, batch processing, Live API, media generation, and tuning, supporting the core integration task. There is no repository-contained output verification, examples require environment setup and possibly light correction, and alternatives or cost-benefit tradeoffs are not sufficiently compared, so the static ceiling of 7 applies.
Official Google Cloud documentation, a migration guide, and a GoogleCloudPlatform samples repository are cited, providing some auditable primary material. The files do not include committed tests, CI results, pinned versions, or third-party execution evidence; key SDK, API, and model claims cannot be independently reproduced from the supplied files alone, so the score is 5.
- Model names, preview status, SDK parameters, and Agent Platform branding/API paths may change; verify current official documentation and stable versions before implementation.
- Examples assume cloud credentials and may use GCS, search, code execution, MCP, caching, batch processing, or video generation, creating data-exposure, privilege, and cost risks; constrain IAM, data scope, budget, and human confirmation first.
- Mainland-China reachability and alternative access paths are not documented; users may be unable to use the core function directly where Google Cloud/Agent Platform services are unreachable.
- Retry, timeout, quota, regional-availability, cancellation, output-validation, and rollback behavior is not specified and should be added before production use.
What does this skill do, and when should you use it?
This skill is for developers using the Gemini API in enterprise environments, especially when a request mentions Vertex AI, Google Cloud, or Agent Platform. It guides SDK usage in Python, JavaScript/TypeScript, Go, Java, and C#, covering text, multimodal inputs, function calling, structured output, embeddings, media generation, caching, batch prediction, and live bidirectional streaming. It requires active Google Cloud credentials or an Agent Platform API key, with the relevant API enabled. It directs users to the newer Google Gen AI SDKs instead of deprecated SDKs.
Maps integration requests to Google Gen AI SDK installation, authentication, client initialization, model calls, and API configuration; provides text-generation examples in multiple languages; explains handling images, audio, video, and documents, invoking functions, producing schema-constrained JSON, generating embeddings, and using caching, batch prediction, media generation, and the Live Realtime API; and points users to official Agent Platform documentation and Python samples when further reference is needed.
- A Google Cloud developer needs to call Gemini for chat, summarization, or generation inside an enterprise project.
- An application team needs Gemini to understand images, audio, video, or documents.
- A backend engineer needs function calling or JSON output that conforms to a defined schema.
- A search application developer needs text embeddings for semantic search.
- A product team needs low-latency bidirectional voice or video interactions.
- A data platform team needs asynchronous batch prediction over a large dataset.
What are this skill's strengths and limitations?
- Covers Python, JavaScript/TypeScript, Go, Java, and C#.
- Spans text generation, multimodal processing, tools, embeddings, media generation, batch prediction, and realtime APIs.
- Clearly standardizes on Google's newer Gen AI SDKs and identifies migration away from legacy SDKs.
- Includes authentication, location selection, model guidance, and official documentation references.
- Requires valid Google Cloud credentials or an API key and an enabled Agent Platform API.
- Depends on external Google Cloud services and network access; pricing, quotas, and service-level details are not provided.
- The source includes no test suite, troubleshooting results, or verification evidence for the language-specific examples.
- Agent Platform was previously called Vertex AI, so legacy branding may create documentation or configuration ambiguity.
How do you install this skill?
In an Agent Skills-compatible client, run npx skills add google/skills, then select skills/cloud/gemini-api from the repository. The README does not document an additional skill-specific installation path for a particular client.
How do you use this skill?
Example trigger: Use the Gemini API on Agent Platform to implement structured JSON output for my Python service. Before use, set GOOGLE_CLOUD_PROJECT, GOOGLE_CLOUD_LOCATION=global, and GOOGLE_GENAI_USE_ENTERPRISE=true, or set GOOGLE_API_KEY and GOOGLE_GENAI_USE_ENTERPRISE=true; then initialize the Gen AI SDK for the chosen language and call a model.
How does this skill compare with similar options?
The source explicitly distinguishes the newer Google Gen AI SDKs from google-cloud-aiplatform, @google-cloud/vertexai, and google-generativeai, and directs users to migrate from those legacy SDKs. It provides no other comparison with competing skills or platforms.