conversations listlengths 2 2 | metadata dict | id stringlengths 36 36 |
|---|---|---|
[
{
"from": "human",
"value": "What are advanced RAG patterns like HyDE, RAG-Fusion, FLARE, and Self-RAG? Give me code examples for at least two of them."
},
{
"from": "gpt",
"value": "Advanced RAG patterns address the core weaknesses of naive retrieval: query-document mismatch, single-query blind... | {
"category": "rag_architecture",
"context": "advanced RAG patterns"
} | 2515d30a-401a-4156-b174-c10131292f98 |
[
{
"from": "human",
"value": "What are advanced RAG patterns like HyDE, RAG-Fusion, FLARE, and Self-RAG? Give me code examples for at least two of them."
},
{
"from": "gpt",
"value": "Advanced RAG patterns address the core weaknesses of naive retrieval: query-document mismatch, single-query blind... | {
"category": "rag_architecture",
"context": "advanced RAG patterns"
} | a555f125-1071-4d15-af0b-bb73e3694628 |
[
{
"from": "human",
"value": "What is GraphRAG and how do knowledge graphs enhance retrieval compared to standard vector search?"
},
{
"from": "gpt",
"value": "GraphRAG combines vector similarity search with graph traversal, enabling retrieval of information that requires understanding entity rel... | {
"category": "rag_architecture",
"context": "GraphRAG and knowledge graph-enhanced retrieval"
} | 5e2b2cb1-56ef-436a-a1b9-6baa1465b1db |
[
{
"from": "human",
"value": "How do I handle different document types — PDFs, HTML pages, code files, and tables — when ingesting data into a RAG system?"
},
{
"from": "gpt",
"value": "Different document types require specialised loaders and parsers. Using the wrong one loses structure (table ce... | {
"category": "document_ingestion",
"context": "handling PDF, HTML, code, and tables in RAG pipelines"
} | 430db408-7be6-490b-92f4-fa66176cbdb1 |
[
{
"from": "human",
"value": "Explain the main chunking strategies for RAG: fixed-size, semantic chunking, recursive character splitter, and parent-child chunks. When should I use each?"
},
{
"from": "gpt",
"value": "Chunking is one of the highest-leverage decisions in RAG. The wrong strategy cau... | {
"category": "chunking_strategies",
"context": "chunking strategies for RAG"
} | 403edcf7-72f2-4709-b4e4-90ac09a77c3a |
[
{
"from": "human",
"value": "What is contextual compression in RAG and how do I use LLMChainExtractor and EmbeddingsFilter to reduce context window usage?"
},
{
"from": "gpt",
"value": "Contextual compression sits between retrieval and generation. Instead of passing entire chunks verbatim to the... | {
"category": "contextual_compression",
"context": "compressing retrieved context to reduce LLM costs"
} | a983627f-151e-48d4-a5bd-956d77f0c479 |
[
{
"from": "human",
"value": "How do I choose between text-embedding-3-small, BGE embeddings, sentence-transformers, and Cohere Embed for my RAG system?"
},
{
"from": "gpt",
"value": "Embedding model choice affects retrieval quality, latency, and cost more than any other single component. Here is... | {
"category": "embedding_models",
"context": "comparing embedding models for RAG"
} | b5354108-2c34-4d08-a92b-bc95774df0f1 |
[
{
"from": "human",
"value": "How do I optimize a RAG system for production — caching, latency, and cost at scale?"
},
{
"from": "gpt",
"value": "Production RAG at scale has three main cost centres: embedding API calls, vector search latency, and LLM generation tokens. Here is how to tackle each.... | {
"category": "rag_production",
"context": "RAG caching, latency optimization, and cost control"
} | 70f7d86d-feec-4d36-a2b6-86964708d02e |
[
{
"from": "human",
"value": "What is GraphRAG and how do knowledge graphs enhance retrieval compared to standard vector search?"
},
{
"from": "gpt",
"value": "GraphRAG combines vector similarity search with graph traversal, enabling retrieval of information that requires understanding entity rel... | {
"category": "rag_architecture",
"context": "GraphRAG and knowledge graph-enhanced retrieval"
} | 38c1e848-2fb4-41ad-816c-37974e9d91c3 |
[
{
"from": "human",
"value": "Compare Pinecone, Weaviate, ChromaDB, and pgvector for RAG. How do I connect to each from Python?"
},
{
"from": "gpt",
"value": "Each vector database targets a different part of the spectrum from developer convenience to enterprise scale. Here is a practical comparis... | {
"category": "vector_databases",
"context": "choosing and using vector databases"
} | 5957cd78-838a-4abe-b2a3-7db6de2f0102 |
[
{
"from": "human",
"value": "Compare Pinecone, Weaviate, ChromaDB, and pgvector for RAG. How do I connect to each from Python?"
},
{
"from": "gpt",
"value": "Each vector database targets a different part of the spectrum from developer convenience to enterprise scale. Here is a practical comparis... | {
"category": "vector_databases",
"context": "choosing and using vector databases"
} | 6f5109db-1260-4544-bbc7-2388cb3c6e86 |
[
{
"from": "human",
"value": "Compare Pinecone, Weaviate, ChromaDB, and pgvector for RAG. How do I connect to each from Python?"
},
{
"from": "gpt",
"value": "Each vector database targets a different part of the spectrum from developer convenience to enterprise scale. Here is a practical comparis... | {
"category": "vector_databases",
"context": "choosing and using vector databases"
} | 0e4d6091-67f7-4caf-a37e-a98376809594 |
[
{
"from": "human",
"value": "Explain the main chunking strategies for RAG: fixed-size, semantic chunking, recursive character splitter, and parent-child chunks. When should I use each?"
},
{
"from": "gpt",
"value": "Chunking is one of the highest-leverage decisions in RAG. The wrong strategy cau... | {
"category": "chunking_strategies",
"context": "chunking strategies for RAG"
} | f52a8db2-f54f-434f-bc64-c6d1e678223a |
[
{
"from": "human",
"value": "How do I evaluate my RAG system using RAGAS? Walk me through faithfulness, answer relevancy, and context recall metrics with code."
},
{
"from": "gpt",
"value": "RAGAS (Retrieval Augmented Generation Assessment) provides reference-free and reference-based metrics tha... | {
"category": "retrieval_evaluation",
"context": "RAGAS evaluation pipeline for RAG"
} | 748c75d6-e754-48f7-b645-ea34ad9ddbab |
[
{
"from": "human",
"value": "How do I implement multi-hop RAG for questions that require chaining multiple retrieval steps? Give me a concrete example."
},
{
"from": "gpt",
"value": "Multi-hop RAG answers questions like \"What is the population of the country where the CEO of OpenAI was born?\" ... | {
"category": "multi_hop_rag",
"context": "multi-hop retrieval for complex questions"
} | ffa819c0-5987-4ef2-a75c-7846edb26a0c |
[
{
"from": "human",
"value": "How does hybrid search work in RAG? Show me how to combine BM25 sparse retrieval with dense embeddings using Reciprocal Rank Fusion."
},
{
"from": "gpt",
"value": "Hybrid search combines keyword-based BM25 (sparse) with semantic embedding search (dense). Each method ... | {
"category": "hybrid_search",
"context": "combining BM25 sparse search with dense embeddings"
} | e8079953-f2fd-4a96-9f71-6fa1e16fb5ab |
[
{
"from": "human",
"value": "Compare Pinecone, Weaviate, ChromaDB, and pgvector for RAG. How do I connect to each from Python?"
},
{
"from": "gpt",
"value": "Each vector database targets a different part of the spectrum from developer convenience to enterprise scale. Here is a practical comparis... | {
"category": "vector_databases",
"context": "choosing and using vector databases"
} | ecd4c40e-38d1-470c-bc48-a7a72123a23b |
[
{
"from": "human",
"value": "How do I handle different document types — PDFs, HTML pages, code files, and tables — when ingesting data into a RAG system?"
},
{
"from": "gpt",
"value": "Different document types require specialised loaders and parsers. Using the wrong one loses structure (table ce... | {
"category": "document_ingestion",
"context": "handling PDF, HTML, code, and tables in RAG pipelines"
} | 4073fb44-95e6-437a-9a13-d497bbcfed24 |
[
{
"from": "human",
"value": "What are advanced RAG patterns like HyDE, RAG-Fusion, FLARE, and Self-RAG? Give me code examples for at least two of them."
},
{
"from": "gpt",
"value": "Advanced RAG patterns address the core weaknesses of naive retrieval: query-document mismatch, single-query blind... | {
"category": "rag_architecture",
"context": "advanced RAG patterns"
} | 2049aabb-16c3-4776-92ad-c325a916c83b |
[
{
"from": "human",
"value": "What are advanced RAG patterns like HyDE, RAG-Fusion, FLARE, and Self-RAG? Give me code examples for at least two of them."
},
{
"from": "gpt",
"value": "Advanced RAG patterns address the core weaknesses of naive retrieval: query-document mismatch, single-query blind... | {
"category": "rag_architecture",
"context": "advanced RAG patterns"
} | 0eeddcfc-f37a-4bed-b510-c4f703ee037e |
[
{
"from": "human",
"value": "What is GraphRAG and how do knowledge graphs enhance retrieval compared to standard vector search?"
},
{
"from": "gpt",
"value": "GraphRAG combines vector similarity search with graph traversal, enabling retrieval of information that requires understanding entity rel... | {
"category": "rag_architecture",
"context": "GraphRAG and knowledge graph-enhanced retrieval"
} | 5aa05cb5-b67d-4d0d-8f47-747db40dab69 |
[
{
"from": "human",
"value": "How do I choose between text-embedding-3-small, BGE embeddings, sentence-transformers, and Cohere Embed for my RAG system?"
},
{
"from": "gpt",
"value": "Embedding model choice affects retrieval quality, latency, and cost more than any other single component. Here is... | {
"category": "embedding_models",
"context": "comparing embedding models for RAG"
} | 1dcfee20-5626-4f7d-bb02-f0b06baffe0c |
[
{
"from": "human",
"value": "Walk me through designing a basic RAG pipeline from scratch. What are the core components and how do they fit together?"
},
{
"from": "gpt",
"value": "A Retrieval-Augmented Generation (RAG) pipeline has five core stages: document loading, chunking, embedding, retriev... | {
"category": "rag_architecture",
"context": "basic RAG pipeline design"
} | eaaeb2a1-9b6c-4e5f-b58a-6bf808ea1b34 |
[
{
"from": "human",
"value": "My RAG system has poor answer quality. How do I diagnose whether the problem is low recall, hallucinations, or context stuffing, and fix each?"
},
{
"from": "gpt",
"value": "Poor RAG quality always traces back to one of three failure modes: the right document was nev... | {
"category": "retrieval_evaluation",
"context": "debugging RAG failures — low recall, hallucinations, context stuffing"
} | 2e5d1df3-e9b7-4b15-849f-2bf181a8c4d2 |
[
{
"from": "human",
"value": "Explain the main chunking strategies for RAG: fixed-size, semantic chunking, recursive character splitter, and parent-child chunks. When should I use each?"
},
{
"from": "gpt",
"value": "Chunking is one of the highest-leverage decisions in RAG. The wrong strategy cau... | {
"category": "chunking_strategies",
"context": "chunking strategies for RAG"
} | 3796ed6e-748f-4333-9bb5-6a846f1c0bd6 |
[
{
"from": "human",
"value": "What are advanced RAG patterns like HyDE, RAG-Fusion, FLARE, and Self-RAG? Give me code examples for at least two of them."
},
{
"from": "gpt",
"value": "Advanced RAG patterns address the core weaknesses of naive retrieval: query-document mismatch, single-query blind... | {
"category": "rag_architecture",
"context": "advanced RAG patterns"
} | bff458d0-7018-42a3-af19-80458541f160 |
[
{
"from": "human",
"value": "What is GraphRAG and how do knowledge graphs enhance retrieval compared to standard vector search?"
},
{
"from": "gpt",
"value": "GraphRAG combines vector similarity search with graph traversal, enabling retrieval of information that requires understanding entity rel... | {
"category": "rag_architecture",
"context": "GraphRAG and knowledge graph-enhanced retrieval"
} | abe2cd4e-79fd-4ec1-8465-c2c2fefc24ca |
[
{
"from": "human",
"value": "Walk me through designing a basic RAG pipeline from scratch. What are the core components and how do they fit together?"
},
{
"from": "gpt",
"value": "A Retrieval-Augmented Generation (RAG) pipeline has five core stages: document loading, chunking, embedding, retriev... | {
"category": "rag_architecture",
"context": "basic RAG pipeline design"
} | 1e57ea61-a39a-4764-8bc9-fe777bb8f8bf |
[
{
"from": "human",
"value": "How do I evaluate my RAG system using RAGAS? Walk me through faithfulness, answer relevancy, and context recall metrics with code."
},
{
"from": "gpt",
"value": "RAGAS (Retrieval Augmented Generation Assessment) provides reference-free and reference-based metrics tha... | {
"category": "retrieval_evaluation",
"context": "RAGAS evaluation pipeline for RAG"
} | a2cbb900-d148-445e-98de-855d0945d17f |
[
{
"from": "human",
"value": "What is GraphRAG and how do knowledge graphs enhance retrieval compared to standard vector search?"
},
{
"from": "gpt",
"value": "GraphRAG combines vector similarity search with graph traversal, enabling retrieval of information that requires understanding entity rel... | {
"category": "rag_architecture",
"context": "GraphRAG and knowledge graph-enhanced retrieval"
} | 070aad5e-b07c-4f65-ab5d-93092e773ddf |
[
{
"from": "human",
"value": "How do I optimize a RAG system for production — caching, latency, and cost at scale?"
},
{
"from": "gpt",
"value": "Production RAG at scale has three main cost centres: embedding API calls, vector search latency, and LLM generation tokens. Here is how to tackle each.... | {
"category": "rag_production",
"context": "RAG caching, latency optimization, and cost control"
} | 3652aa95-e2ef-47ff-814e-daeef83e0320 |
[
{
"from": "human",
"value": "Walk me through designing a basic RAG pipeline from scratch. What are the core components and how do they fit together?"
},
{
"from": "gpt",
"value": "A Retrieval-Augmented Generation (RAG) pipeline has five core stages: document loading, chunking, embedding, retriev... | {
"category": "rag_architecture",
"context": "basic RAG pipeline design"
} | 4ebfec67-b85d-4c44-b371-e873e272cda8 |
[
{
"from": "human",
"value": "What is GraphRAG and how do knowledge graphs enhance retrieval compared to standard vector search?"
},
{
"from": "gpt",
"value": "GraphRAG combines vector similarity search with graph traversal, enabling retrieval of information that requires understanding entity rel... | {
"category": "rag_architecture",
"context": "GraphRAG and knowledge graph-enhanced retrieval"
} | ba99b1b9-f3bc-4d92-bcd9-96f1c564734c |
[
{
"from": "human",
"value": "What is GraphRAG and how do knowledge graphs enhance retrieval compared to standard vector search?"
},
{
"from": "gpt",
"value": "GraphRAG combines vector similarity search with graph traversal, enabling retrieval of information that requires understanding entity rel... | {
"category": "rag_architecture",
"context": "GraphRAG and knowledge graph-enhanced retrieval"
} | 494f227d-7782-4eeb-ab3d-dda20331a980 |
[
{
"from": "human",
"value": "My RAG system has poor answer quality. How do I diagnose whether the problem is low recall, hallucinations, or context stuffing, and fix each?"
},
{
"from": "gpt",
"value": "Poor RAG quality always traces back to one of three failure modes: the right document was nev... | {
"category": "retrieval_evaluation",
"context": "debugging RAG failures — low recall, hallucinations, context stuffing"
} | 3414c3f8-f924-460c-a8e8-5ca11374aac2 |
[
{
"from": "human",
"value": "How do I evaluate my RAG system using RAGAS? Walk me through faithfulness, answer relevancy, and context recall metrics with code."
},
{
"from": "gpt",
"value": "RAGAS (Retrieval Augmented Generation Assessment) provides reference-free and reference-based metrics tha... | {
"category": "retrieval_evaluation",
"context": "RAGAS evaluation pipeline for RAG"
} | 761ebcd0-dd7b-4aa2-94ac-8cf45c54bce7 |
[
{
"from": "human",
"value": "How do I choose between text-embedding-3-small, BGE embeddings, sentence-transformers, and Cohere Embed for my RAG system?"
},
{
"from": "gpt",
"value": "Embedding model choice affects retrieval quality, latency, and cost more than any other single component. Here is... | {
"category": "embedding_models",
"context": "comparing embedding models for RAG"
} | 2cf053c3-ec3a-4215-97a3-411f7edf160c |
[
{
"from": "human",
"value": "How do I optimize a RAG system for production — caching, latency, and cost at scale?"
},
{
"from": "gpt",
"value": "Production RAG at scale has three main cost centres: embedding API calls, vector search latency, and LLM generation tokens. Here is how to tackle each.... | {
"category": "rag_production",
"context": "RAG caching, latency optimization, and cost control"
} | 5abbdf4e-7d96-4a1d-998c-f659eec07047 |
[
{
"from": "human",
"value": "Compare Pinecone, Weaviate, ChromaDB, and pgvector for RAG. How do I connect to each from Python?"
},
{
"from": "gpt",
"value": "Each vector database targets a different part of the spectrum from developer convenience to enterprise scale. Here is a practical comparis... | {
"category": "vector_databases",
"context": "choosing and using vector databases"
} | 3fe523fb-e835-4701-81a3-231956bc1465 |
[
{
"from": "human",
"value": "How do I evaluate my RAG system using RAGAS? Walk me through faithfulness, answer relevancy, and context recall metrics with code."
},
{
"from": "gpt",
"value": "RAGAS (Retrieval Augmented Generation Assessment) provides reference-free and reference-based metrics tha... | {
"category": "retrieval_evaluation",
"context": "RAGAS evaluation pipeline for RAG"
} | 804a1d36-d29d-46ba-93a1-71333f04ea91 |
[
{
"from": "human",
"value": "My RAG system has poor answer quality. How do I diagnose whether the problem is low recall, hallucinations, or context stuffing, and fix each?"
},
{
"from": "gpt",
"value": "Poor RAG quality always traces back to one of three failure modes: the right document was nev... | {
"category": "retrieval_evaluation",
"context": "debugging RAG failures — low recall, hallucinations, context stuffing"
} | 79300a15-7d0d-43ef-b91f-29d0193c917a |
[
{
"from": "human",
"value": "What are advanced RAG patterns like HyDE, RAG-Fusion, FLARE, and Self-RAG? Give me code examples for at least two of them."
},
{
"from": "gpt",
"value": "Advanced RAG patterns address the core weaknesses of naive retrieval: query-document mismatch, single-query blind... | {
"category": "rag_architecture",
"context": "advanced RAG patterns"
} | d2cade3a-e613-44e3-bf92-d06f4f64e991 |
[
{
"from": "human",
"value": "How do I choose between text-embedding-3-small, BGE embeddings, sentence-transformers, and Cohere Embed for my RAG system?"
},
{
"from": "gpt",
"value": "Embedding model choice affects retrieval quality, latency, and cost more than any other single component. Here is... | {
"category": "embedding_models",
"context": "comparing embedding models for RAG"
} | 23178623-6a0b-4afa-a7de-232d0b6dcc61 |
[
{
"from": "human",
"value": "How do I build a RAG system that generates responses with inline citations linking back to source documents?"
},
{
"from": "gpt",
"value": "Citation generation ensures every claim in the answer is traceable to a specific source, which is critical for legal, medical, ... | {
"category": "citation_generation",
"context": "generating cited responses linked to source documents"
} | eb954c12-f82b-49ee-8251-4e9cfeba4044 |
End of preview. Expand in Data Studio
RAG Systems SFT 100K
A synthetic supervised fine-tuning dataset of 100,000 high-quality conversations covering Retrieval-Augmented Generation (RAG) systems — from basic pipelines to advanced multi-hop retrieval, evaluation, and production optimization. Designed to train AI assistants that can help engineers build, debug, and scale RAG applications.
Dataset Description
This dataset covers the full spectrum of RAG system development across 12 specialized categories. Each record follows the ShareGPT format with a practitioner-level question and a detailed, implementation-ready response with working Python code.
Categories
| Category | Description |
|---|---|
rag_architecture |
Basic pipeline, advanced patterns (HyDE, RAG-Fusion, FLARE, GraphRAG) |
chunking_strategies |
Fixed-size, semantic, recursive, parent-child chunking |
vector_databases |
Pinecone vs Weaviate vs ChromaDB vs pgvector |
embedding_models |
text-embedding-3-small, BGE, sentence-transformers comparison |
retrieval_evaluation |
RAGAS metrics, debugging RAG failures |
hybrid_search |
BM25 + dense embeddings, RRF reranking |
reranking |
Cross-encoder reranking with Cohere, FlashRank |
contextual_compression |
LLMChainExtractor, EmbeddingsFilter, context window optimization |
citation_generation |
Generating cited responses linked to source documents |
multi_hop_rag |
Multi-step chained retrieval for complex questions |
rag_production |
Caching, latency optimization, cost control |
document_ingestion |
PDF, HTML, code, table parsing and loading |
Format
ShareGPT format:
{
"conversations": [
{"from": "human", "value": "...RAG question..."},
{"from": "gpt", "value": "...implementation-ready response with code..."}
],
"metadata": {"category": "...", "context": "..."},
"id": "uuid"
}
Use Cases
- Fine-tuning AI assistants for RAG system design and debugging
- Training models to reason about retrieval quality and evaluation
- Building AI-assisted RAG development tooling
- Educating teams on production RAG patterns
Quality Notes
All responses include working Python code examples using LangChain, LlamaIndex, and direct API integrations. Covers RAGAS evaluation, hybrid search implementation, and production deployment patterns.
- Downloads last month
- 74