PROC_DB_GEN

v1.0.0

[SYS] Procedural knowledge database generator: Formula creation + Steps vectorization

$ cat description.txt Advanced procedural document generation system for Analogy Augmented Generation (AAG). Transforms workflows and technical processes into structured, vectorized procedural knowledge. Creates standardized procDoc format with name, formula (procFormula), and detailed step-by-step descriptions for analogous inference in technical domains.

CORE_FEATURES:

Formula Generation

> procFormula creation
> Component identification
> Step sequence mapping

Procedural Structuring

> Detailed step breakdown
> Sequential processing
> Technical clarity enforcement

Document Processing

> Plain-text formatting
> Consistent document structure
> Embedding-ready output

Analogy Enhancement

> Knowledge transfer preparation
> Vector database integration
> Inference-optimized structure

EXAMPLE_OUTPUT:

Workflow Input:

"Procedure Text": "A hybrid RAG system description with web search fallback...",

"Format Type": "n8n workflow",

"procDoc Generation": "enabled",

"Vector Embedding": "enabled"

RESULTING PROC DOC

Hybrid RAG System Documentation (procDoc Format)

ℹ️
This document demonstrates the standardized procDoc format created by the Procedural Database Generator workflow. It transforms informal workflow descriptions into structured, vectorizable knowledge.

Document Metadata

Source: n8n Workflow Description
Format: procDoc with procFormula
Tags: #HybridRAG #VectorSearch #WebSearch #Qdrant #Gemini

Hybrid RAG System with Vector Store and Web Search

procFormula

                        Hybrid RAG Query Processing(Chat Input, Qdrant Vector Store, OpenAI Embeddings, Aggregate Node, Gemini Text Classifier, Gemini LLM, Perplexity AI LLM --> Context-aware LLM Response(Receive chat input, Search Qdrant vector store using OpenAI embeddings, Aggregate search results, Classify relevance of results using Gemini, If relevant route to Gemini LLM with context, If not relevant route to Perplexity AI for web search, Generate final response))
                    

Description

This procedure describes a workflow designed to answer user queries by first attempting to retrieve relevant information from a vector database (Qdrant) and, if the retrieved information is deemed irrelevant, falling back to a web search using Perplexity AI.

🎯
INTENTION: To provide accurate and contextually appropriate responses to user chat inputs by leveraging both a private vector knowledge base and public web search capabilities. The system prioritizes information from the vector store but uses web search as a fallback to ensure comprehensive answers.

PROCEDURE STEPS:

RECEIVE INPUT: The workflow is initiated when a message is received through the 'When chat message received' node (@n8n/n8n-nodes-langchain.chatTrigger).
EMBED QUERY: The input message ('chatInput') is implicitly used by the 'Qdrant Vector Store' node, which requires embeddings. The 'Embeddings OpenAI' node (@n8n/n8n-nodes-langchain.embeddingsOpenAi), connected via the 'ai_embedding' connection, generates these embeddings using an OpenAI model.
SEARCH VECTOR STORE: The 'Qdrant Vector Store' node (@n8n/n8n-nodes-langchain.vectorStoreQdrant) performs a similarity search in the specified Qdrant collection ('YOUR COLLECTION NAME') using the embeddings of the 'chatInput'. It retrieves the top 2 ('topK': 2) most relevant documents.

                        ⋮ 

                        7 more steps available (hidden for brevity)

Vector Database Schema

{
  "id": "hybrid_rag_001",
  "procedureName": "Hybrid RAG System with Vector Store and Web Search",
  "procFormula": "Hybrid RAG Query Processing(Chat Input, Qdrant Vector Store, OpenAI Embeddings, Aggregate Node, Gemini Text Classifier, Gemini LLM, Perplexity AI LLM --> Context-aware LLM Response(Receive chat input, Search Qdrant vector store using OpenAI embeddings, Aggregate search results, Classify relevance of results using Gemini, If relevant route to Gemini LLM with context, If not relevant route to Perplexity AI for web search, Generate final response))",
  "metadata": {
    "tags": ["HybridRAG", "VectorSearch", "WebSearch", "Qdrant", "Gemini", "PerplexityAI"],
    "domain": "AI/LLM Workflows",
    "components": ["chat", "vectordb", "classifier", "routing", "llm", "websearch"]
  },
  "embedding": [0.023, -0.112, 0.043, ...] // Vector representation (1536 dimensions)
}

💡
This documentation was automatically generated using the PROC_DB_GEN workflow, which transforms workflow descriptions into structured procedural knowledge suitable for vector embedding and analogous inference.

This is an example of a procedural document created with our template

$ system_requirements

MODELS: gemini 2.5, gpt text-embedding-3-small
STORAGE: qdrant
SERVICES: none required
OUTPUT: vector database of embeddings
PRICING: gemini - per token,
         gpt - per token, 
         qdrant - free tier         
EST. PER RUN COST: less than €0.01

PROCESS_FLOW:

AUTOMATION_BENEFITS:

> Transform complex workflows into analogous learning material
> Enhance LLM responses with procedural knowledge
> Enable cross-domain technical understanding
> Standardize procedural documentation format
> Automate the creation of procedural knowledge bases

€149

PURCHASE_TEMPLATE

* Compatible with all n8n installations v1.0.0+

*Superflowz is a subsidiary of CARDUME ESBELTO UNIP. LDA. Your purchase will be from, and your receipt will list, CARDUME ESBELTO