Skip to content
Open
Show file tree
Hide file tree
Changes from all commits
Commits
File filter

Filter by extension

Filter by extension

Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
51 changes: 39 additions & 12 deletions core/skills/rh-qs-architect/SKILL.md
Original file line number Diff line number Diff line change
@@ -1,3 +1,3 @@
---
name: rh-qs-architect
description: Architecture phase for AI Quickstarts. Reads the PRD, maps requirements to OpenShift AI and ai-architecture-charts, presents a bill of materials, generates a Mermaid diagram, and produces an architecture spec. Use when a PRD exists under .rhoai-qs/<slug>/.qs-factory/prds/.
Expand Down Expand Up @@ -111,23 +111,50 @@

#### Step 5: Select ai-architecture-charts

Spawn the **chart-selector** subagent with the refined features:
**Try OpenAI Decision API first:**

```python
Agent(
description="Select ai-architecture-charts for this quickstart",
prompt=f"""
Read and follow instructions from:
core/skills/rh-qs-architect/subagents/chart-selector-prompt.md
Run the chart selector script:

input_features: {input_features}
deployment_questions: {deployment_questions}
slug: {slug}
charts_reference_path: core/skills/rh-qs-architect/references/ai-architecture-charts.md
"""
```python
result = Bash(
description="Select charts using OpenAI Decision API",
command=f"python3 core/skills/rh-qs-architect/scripts/openai_decision_chart_selector.py --slug {slug}"
)
```

If the script succeeds (exit code 0):
1. Parse the JSON output to get `selected` and `uncertain` chart arrays
2. If `uncertain` is non-empty, present each uncertain chart to the user:
```
⚠️ Uncertain chart selections - please review:
- {chart_name}: {probability} ({reason})

Should this chart be included? [y/n]
```
3. Merge user-approved uncertain charts into the selected list
4. Store the combined selection as `{selected_charts}` and continue to Step 6

**Fallback to LLM subagent if script fails:**

If the script exits non-zero (OPENAI_API_KEY missing, API error, refusal, or other error):
1. Log the failure reason from stderr
2. Spawn the **chart-selector** subagent:
```python
Agent(
description="Select ai-architecture-charts for this quickstart",
prompt=f"""
Read and follow instructions from:
core/skills/rh-qs-architect/subagents/chart-selector-prompt.md

input_features: {input_features}
deployment_questions: {deployment_questions}
slug: {slug}
charts_reference_path: core/skills/rh-qs-architect/references/ai-architecture-charts.json
"""
)
```
3. Store the subagent's returned charts as `{selected_charts}` and continue to Step 6

#### Step 6: Map leftover features to OpenShift AI

Prefer the charts selected in Step 5. For each refined PRD feature:
Expand Down
119 changes: 119 additions & 0 deletions core/skills/rh-qs-architect/references/ai-architecture-charts.json
Original file line number Diff line number Diff line change
@@ -0,0 +1,119 @@
{
"llama-stack": {
"description": "Comprehensive AI orchestration platform that provides a unified API for multiple model providers, safety shields, and AI agent capabilities.",
"provides": [
"Multi-provider model support (local, remote, VertexAI)",
"Safety shields with Llama Guard and other safety models",
"AI agent capabilities with persistent memory",
"Automatic model discovery and URL generation"
],
"use_when": "Comprehensive AI orchestration platform needed"
},
"ogx-ai": {
"description": "Successor chart providing orchestration with OpenAI-compatible APIs, multi-provider model support, and MCP-style tool integration.",
"provides": [
"OpenAI-compatible agentic API server",
"Multi-provider model support (local, remote, VertexAI)",
"Content moderation via OpenAI-compatible /v1/moderations endpoint",
"AI agent capabilities with persistent memory",
"Automatic model discovery and URL generation",
"Supersedes llama-stack for new deployments"
],
"use_when": "Agentic orchestration with OpenAI compatibility needed"
},
"llm-service": {
"description": "High-performance model serving using vLLM runtime with OpenShift AI/KServe integration supporting GPU and CPU deployment modes.",
"provides": [
"vLLM-based model serving with OpenAI-compatible API",
"Support for any vLLM-compatible models and sizes",
"GPU/CPU deployment flexibility",
"Tool calling and function execution support",
"OpenShift AI/KServe integration"
],
"use_when": "On-cluster GPU model serving required"
},
"pgvector": {
"description": "PostgreSQL with pgvector extension enabling high-performance vector database capabilities for embeddings in AI/ML applications.",
"provides": [
"Vector similarity search (cosine, L2, inner product)",
"Multiple index types (IVFFlat, HNSW)",
"Support for various embedding dimensions",
"ACID compliance with PostgreSQL reliability"
],
"use_when": "Vector database for RAG/semantic retrieval needed"
},
"minio": {
"description": "S3-compatible object storage server for documents, models, and data in AI/ML pipelines with scalable storage and web console management.",
"provides": [
"S3-compatible API",
"Web-based management console",
"Bucket policies and lifecycle management",
"Sample file upload functionality"
],
"use_when": "Object storage for documents/models/data needed"
},
"oracle-db": {
"description": "Oracle Database Free 23ai with AI Vector features providing enterprise-grade database capabilities with built-in vector operations.",
"provides": [
"Native vector operations and similarity search",
"JSON duality and graph analytics",
"Enterprise database reliability",
"AI-optimized storage and indexing",
"TPC-DS automated data population via Kubernetes Job",
"Configurable scale factors and schema management",
"Automated security configuration for AI MCP server integration"
],
"use_when": "Enterprise database with AI vector features needed"
},
"ingestion-pipeline": {
"description": "Comprehensive data ingestion pipeline processing documents from various sources and storing vector embeddings for semantic search and RAG.",
"provides": [
"Multi-source data ingestion (S3/MinIO, GitHub, URLs)",
"Document chunking and embedding generation",
"REST API for pipeline management",
"Integration with vector databases"
],
"use_when": "Automated document ingestion and embedding pipeline needed"
},
"configure-pipeline": {
"description": "Jupyter notebook environment for RAG configuration and pipeline setup with interactive tools and MinIO integration for data access.",
"provides": [
"Pre-configured Jupyter environment",
"RAG pipeline configuration tools",
"MinIO integration for data access",
"Template and configuration management"
],
"use_when": "Interactive RAG configuration and pipeline setup needed"
},
"mcp-servers": {
"description": "Model Context Protocol servers providing external tools and capabilities to AI models for interacting with external systems and APIs.",
"provides": [
"Weather information services",
"Server-Sent Events (SSE) endpoints",
"Custom tool development framework",
"Integration with LlamaStack and OGX agents"
],
"use_when": "External tools and capabilities for AI agents needed"
},
"oracle-sqlcl": {
"description": "MCP server exposing Oracle SQLcl capabilities to AI agents via Toolhive, enabling database tooling and interactions from AI platforms.",
"provides": [
"Execute SQL/PLSQL against Oracle databases via Model Context Protocol",
"Integrates with Toolhive Operator (CRDs and operator managed as chart dependencies)",
"Oracle connection managed via Kubernetes secrets and configurable service"
],
"use_when": "Oracle database tooling for AI agents needed"
},
"keycloak": {
"description": "Identity and access management solution providing authentication and authorization for AI applications via OAuth2, OpenID Connect, and SAML.",
"provides": [
"Official Keycloak container images",
"OAuth2, OpenID Connect, and SAML protocols",
"Built-in PostgreSQL database (via pgvector subchart) or external database support",
"OpenShift Routes and Kubernetes Ingress support",
"Production-ready configuration with health checks and metrics",
"Flexible authentication and authorization policies"
],
"use_when": "Identity and access management for AI applications needed"
}
}
130 changes: 0 additions & 130 deletions core/skills/rh-qs-architect/references/ai-architecture-charts.md

This file was deleted.

Loading
Loading