diegosouzapw/awesome-omni-skill / skills/ai-agents/adk-rag-agent
adk-rag-agent
VerifiedBuild RAG (Retrieval-Augmented Generation) agents with Google ADK and Vertex AI RAG Engine. Use when implementing document Q&A, knowledge base search, or citation-backed responses. Covers VertexAiRagRetrieval tool, corpus setup, and citation formatting.
| File | Kind | Size |
|---|---|---|
| SKILL.md | skill md | 2.9 KB |
| metadata.json | config | 1.5 KB |
SKILL.md
View on GitHub---
name: adk-rag-agent
description: Build RAG (Retrieval-Augmented Generation) agents with Google ADK and Vertex AI RAG Engine. Use when implementing document Q&A, knowledge base search, or citation-backed responses. Covers VertexAiRagRetrieval tool, corpus setup, and citation formatting.
---
# Google ADK RAG Agent
Build agents that answer questions from document corpora using Vertex AI RAG Engine.
## Requirements
- Vertex AI backend (not Gemini API)
- Google Cloud project with Vertex AI enabled
- RAG corpus created in Vertex AI
## Environment Variables
```bash
GOOGLE_GENAI_USE_VERTEXAI=1
GOOGLE_CLOUD_PROJECT=your-project-id
GOOGLE_CLOUD_LOCATION=us-central1
RAG_CORPUS=projects/{PROJECT_ID}/locations/{LOCATION}/ragCorpora/{CORPUS_ID}
```
## Core Implementation
```python
from google.adk import Agent
from google.adk.tools import VertexAiRagRetrieval
# Configure RAG retrieval tool
rag_tool = VertexAiRagRetrieval(
name="retrieve_docs",
description="Retrieve relevant documentation for the question",
rag_corpus=os.environ["RAG_CORPUS"],
similarity_top_k=10,
vector_distance_threshold=0.6,
)
# Create agent with RAG tool
agent = Agent(
name="rag_agent",
model="gemini-2.0-flash-001",
instruction=INSTRUCTION_PROMPT,
tools=[rag_tool],
)
```
## Instruction Prompt Pattern
```python
INSTRUCTION_PROMPT = """
You are an AI assistant with access to a specialized document corpus.
RETRIEVAL:
- Use retrieve_docs for specific knowledge questions
- Skip retrieval for casual conversation
- Ask clarifying questions when intent is unclear
SCOPE:
- Only answer questions related to the corpus
- Say "I don't have information about that" for out-of-scope queries
CITATIONS:
- Always cite sources at the end of responses
- Format: [Title](url) or [Document Section](url)
- Consolidate multiple citations from the same source
"""
```
## Corpus Setup
Create corpus via Vertex AI Console or SDK:
```python
from vertexai.preview import rag
# Create corpus
corpus = rag.create_corpus(display_name="my-corpus")
# Import documents (PDF, TXT, HTML)
rag.import_files(
corpus_name=corpus.name,
paths=["gs://bucket/doc.pdf"], # or local files
chunk_size=512,
chunk_overlap=100,
)
```
## Key Parameters
| Parameter | Description | Default |
|-----------|-------------|---------|
| `similarity_top_k` | Max chunks to retrieve | 10 |
| `vector_distance_threshold` | Min similarity (0-1, lower=stricter) | 0.6 |
| `chunk_size` | Tokens per chunk at import | 512 |
| `chunk_overlap` | Overlap between chunks | 100 |
## Citation Best Practices
1. Single source → single citation at end
2. Multiple sources → list all citations
3. Same document, multiple chunks → consolidate into one citation
4. Never expose internal chunk IDs to users
## References
- [Corpus setup details](references/corpus-setup.md)
- [Sample repo](https://github.com/google/adk-samples/tree/main/python/agents/RAG)
Related skills
| Skill | Stars | Updated | Works with | Status | Install |
|---|---|---|---|---|---|
| reverse-engineer-anythingmorluto/reaReverse engineer native, managed, Electron/JavaScript, packaged, firmware, and browser targets with REA, and build agent environments (AgentEnv bundles) of web applications from a live instance or a capture, and of Linux desktop apps with a task from a request. Use shipped-artifact or requested runtime evidence to explain features, compare versions, decompile code, or guide a reconstruction. Skip REA for ordinary source-repository architecture analysis. | Stars: 472k | Updated | Claude CodeCodex | Unverified | Install |
| diagnosing-superpowersobra/superpowersUse when a superpowers session went wrong and your human partner wants to know why — repeated work, ignored plans, stumbles, poor results, a skill that didn't fire, "it took too long", "why is it so expensive", "what is it doing" — or wants to build a bug report for the superpowers maintainers, for the current session or a past one identified by id or path, on any harness. | Stars: 322k | Updated | Claude CodeCodex | Verified | Install |
| dispatching-parallel-agentsobra/superpowersUse when facing 2+ independent tasks that can be worked on without shared state or sequential dependencies | Stars: 322k | Updated | Claude CodeCodex | Verified | Install |
| executing-plansobra/superpowersUse when executing an implementation plan in the current session as the implementer yourself — your human partner chose inline execution, or no subagent tool is available | Stars: 322k | Updated | Claude CodeCodex | Verified | Install |