Gemini Enterprise Agent Platform Docs

Gemini Enterprise Agent Platform Docs

Official Google Cloud Gemini Enterprise Agent Platform documentation synced from docs.cloud.google.com/gemini-enterprise-agent-platform.

Use this skill for Google Cloud Gemini Enterprise Agent Platform (GEAP): building, deploying, governing, scaling, and optimizing agents and model workloads on Google Cloud. This is distinct from the consumer/developer Gemini API docs at ai.google.dev; use google-ai-docs / google-ai-api for Gemini API-only behavior.

Content Scope

This skill syncs product-local .md.txt pages discovered under /gemini-enterprise-agent-platform, excluding generated REST/RPC API reference, legacy machine-learning/, and notebooks/ subtrees. Current coverage is 426 Markdown pages:

• references/index.md, overview.md, build.md, scale.md, govern.md, models.md, agent-studio.md, agents.md — product overviews and section entrypoints.
• references/build/ — Agent Development Kit, Agent Runtime, Managed Agents API, RAG Engine, Vector Search, Skill Registry, custom/ADK/LangChain/LangGraph/LlamaIndex/A2A/AG2 agents.
• references/scale/ — Agent Runtime deployment, Memory Bank, sessions, sandbox/code execution, monitoring, tracing, private connectivity, access, and scaling.
• references/govern/ — Agent Gateway, IAM/policies, semantic governance, agent identity, agent registry, content security, security findings, and sharing.
• references/optimize/ — agent evaluation, simulated evaluation, optimization, observability, traces, topology, and example store.
• references/models/ — Gemini model usage on GEAP, model versions, Gemini 2.5/3.x, Imagen/Veo/Lyria/open/partner models, grounding, tuning, embeddings, context cache, Live API, Model Garden, provisioned throughput, MaaS, safety, quotas, and OpenAI migration.
• references/resources/, agent-studio/, agents/, troubleshooting/, tutorials/ — locations, quotas, support, quickstarts, Agent Studio, Deep Research, ADK quickstarts, and troubleshooting.

Excluded by design:

• reference/rest, reference/rpc, and reference/python generated API reference pages.
• machine-learning/ and notebooks/ legacy or adjacent Vertex AI/Workbench docs.
• HTML pages, static assets, notebook files, images, and non-product links.

Hard Rules

• MUST search references/ before answering GEAP-specific questions about Agent Runtime, Managed Agents API, Agent Studio, RAG Engine, Vector Search, Agent Gateway, Memory Bank, model deployment, tuning, grounding, evaluation, governance, quotas, or locations.
• MUST distinguish Google Cloud GEAP from ai.google.dev Gemini API. API-key-only Gemini API behavior belongs in google-ai-docs / google-ai-api; Google Cloud IAM/project/location/Agent Platform behavior belongs here.
• MUST call out launch-stage and data-sensitivity warnings when docs mention Pre-GA, preview, trusted tester, Agentic AI Services terms, or restrictions on confidential/proprietary data.
• MUST distinguish build/control-plane concepts such as Managed Agents API, Agent Runtime, Agent Gateway, RAG Engine, Vector Search, Skill Registry, and Memory Bank; do not conflate them.
• NEVER invent Google Cloud resource names, IAM roles, REST fields, model IDs, regions, quota limits, pricing, or API behavior without checking references or the dedicated generated API reference.
• NEVER use this skill as the source of truth for generated REST/RPC reference details, Vertex AI Workbench notebooks, or legacy machine-learning/ docs; fetch/use those dedicated docs when needed.

Fast Lookup

rg -n "Managed Agents|Agents API|Interactions API|sandbox|Antigravity|agent skills" skills/gemini-enterprise-agent-platform-docs/references/build
rg -n "Agent Runtime|deploy|ReasoningEngine|ADK|LangChain|LangGraph|LlamaIndex|A2A|AG2" skills/gemini-enterprise-agent-platform-docs/references/build skills/gemini-enterprise-agent-platform-docs/references/scale
rg -n "RAG Engine|Vector Search|corpus|retrieval|ranking|hybrid search|index" skills/gemini-enterprise-agent-platform-docs/references/build
rg -n "Agent Gateway|IAM|policy|semantic governance|agent identity|agent registry|security findings" skills/gemini-enterprise-agent-platform-docs/references/govern
rg -n "Memory Bank|session|sandbox|code execution|tracing|monitoring|private service connect" skills/gemini-enterprise-agent-platform-docs/references/scale
rg -n "Gemini|Claude|Grok|Llama|Veo|Imagen|Lyria|MaaS|Model Garden|provisioned throughput|tuning|grounding|embeddings" skills/gemini-enterprise-agent-platform-docs/references/models
rg -n "quota|location|region|data residency|support|pricing|Pre-GA|confidential" skills/gemini-enterprise-agent-platform-docs/references

Reference Map

• Start here: references/index.md, overview.md, build.md, scale.md, govern.md, models.md.
• Agent build paths: references/build/adk.md, build/runtime/quickstart.md, build/managed-agents.md, build/rag-engine/rag-overview.md, build/vector-search/overview.md, build/skill-registry.md.
• Runtime and scaling: references/scale/runtime/deploy-an-agent.md, scale/runtime/use-an-adk-agent.md, scale/memory-bank.md, scale/sessions.md, scale/sandbox/code-execution-overview.md.
• Governance: references/govern/gateways/agent-gateway-overview.md, govern/policies/overview.md, govern/agent-identity-overview.md, govern/agent-registry.md.
• Models: references/models/start.md, models/gemini/, models/partner-models/, models/open-models/, models/grounding/, models/tuning/, models/embeddings/, models/context-cache/.
• Optimization and operations: references/optimize/evaluation/, optimize/observability/, resources/locations.md, resources/agent-quotas.md, troubleshooting/.

Workflow

1. Identify whether the question is about build, scale, govern, optimize, models, resources/quotas/locations, or troubleshooting.
2. Search the relevant subtree first, then broader references/ for cross-cutting warnings or prerequisites.
3. State Google Cloud prerequisites explicitly: project, region/location, IAM, APIs, service accounts, billing/quotas, and launch stage where relevant.
4. If the answer depends on generated REST/RPC schema or a machine-learning//notebooks page excluded from this skill, say so and consult the dedicated Google Cloud reference page.