skills.rest
Aadi-110i
Aadi-110iCOMMUNITY
Explore all skills by this author →
WebsiteGitHub

ai-rag-pipeline

Builds RAG pipelines combining web search tools with LLMs via the inference.sh CLI.

Updated Jul 28, 2026
Search Tags:#fact-checking#llm#rag#web search#tavily#retrieval augmented generation#research pipeline
One-click install
npx skills add https://github.com/Aadi-110i/PEP-PROJECT --skill ai-rag-pipeline-aadi-110i
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: ai-rag-pipeline
Source: https://github.com/Aadi-110i/PEP-PROJECT/tree/main/skills/ai-rag-pipeline
Command: npx skills add https://github.com/Aadi-110i/PEP-PROJECT --skill ai-rag-pipeline-aadi-110i

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve? LLMs lack access to current information and cannot cite sources, producing outdated or unverifiable answers. This Skill provides ready-made pipeline patterns that combine web search, content extraction, and LLM generation into grounded, citation-backed responses. ## Core Features & Use Cases - Search + LLM Pipelines: Chain Tavily or Exa search results into Claude, GPT-4o, or Gemini prompts through the inference.sh belt CLI. - Multiple Pipeline Patterns: Includes templates for simple search-and-answer, multi-source research, URL content extraction, fact-checking, and full research report generation. - Use Case: A research assistant agent receives a claim to verify, searches for supporting and contradicting evidence via Tavily, then asks Claude Sonnet to produce a verdict with cited sources. ## Quick Start Run a Tavily search on your topic with the belt CLI, then pass the results into a Claude prompt asking for a cited summary.

Frequently Asked Questions about ai-rag-pipeline

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I build a RAG pipeline with web search and an LLM?▼

Run a search with `belt app run tavily/search-assistant` using your query, capture the output, then pass it into an LLM call like `openrouter/claude-sonnet-45` with a prompt instructing it to answer based on the search results and cite sources.

Tavily vs Exa search: which should I use for RAG?▼

Tavily search-assistant provides AI-powered search with ready answers, making it good for general queries. Exa search offers neural semantic matching for finding conceptually similar content, while Exa Answer gives direct factual responses.

How do I fact-check a claim using search and an LLM?▼

Search for evidence about the claim with Tavily, then prompt Claude with the claim and evidence, asking for a verdict (True/False/Partially True/Unverified), supporting evidence, contradicting evidence, and sources.

What LLM models are available through OpenRouter in this pipeline?▼

Available models include Claude Sonnet 4.5 for complex analysis, Claude Haiku 4.5 for fast processing, GPT-4o for general purpose tasks, and Gemini 2.5 Pro for long context documents.

What are the limitations of search-based RAG pipelines?▼

Search-based RAG depends on the quality and freshness of web results, and long search outputs can exceed context limits. The skill recommends summarizing long results with a fast model like Claude Haiku before deeper analysis.

Related Skills

Explore More Software Engineering Skills→
inference-shinference-sh
723

gpt-image

Generate and edit images with OpenAI GPT-Image-2 via the inference.sh CLI.

Official
Basic
emilneuraz-aiemilneuraz-ai

rag-implementation

Implement retrieval-augmented generation pipelines with vector databases and reranking.

Community
Advanced
Menschlichkeit-OsterreichMenschlichkeit-Osterreich

agent-tools

Execute over 150 AI applications via the inference.sh CLI.

Official
Intermediate
patronus-aipatronus-ai
2

rag-implementation

Build RAG systems using vector databases and semantic search.

Official
Intermediate