#1.Rag-friendly Document

Convert Tech Radar Gsheet into GDoc. Read each rows and cols data then transformed it into simple paragraph rows so that it will be easy to convert into vector database.

You may use appscript optionally to do this transformation.

Set up steps

1. Google Cloud Project and Vertex AI API:

   • Create a Google Cloud project.
   • Enable the Vertex AI API for your project.

2. Google AI API Key:

   • Obtain a Google AI API key from Google AI Studio.

3. Groq AI API Key:

   • Obtain a Groq AI API key from Groq.

4. Pinecone Account:

   • Create a free account on the Pinecone website.
   • Obtain your API key from your Pinecone dashboard.
   • Create an index named seanrag or any other name in your Pinecone project.

5. Google Drive:

   • Create a dedicated folder in your Google Drive to store company documents.

6. Credentials in n8n:

   • Configure the following credentials in your n8n environment:
     • Google Drive OAuth2
     • Google Gemini (PaLM) API (using your Google AI API key)
     • Pinecone API (using your Pinecone API key)

7. Import the Workflow:

   • Import this workflow into your n8n instance.

8. Configure the Workflow:

   • Update both Google Drive Trigger nodes to watch the specific folder you created in Google Drive.
   • Configure the Pinecone Vector Store nodes to use your company-files index.

9. Optional

   • Set up NocoDB and create a table with the same fields. Map the fields exactly or as preferred. ConversationHistory - user,email,ai,sessionid,date,datetime

• Remember to map the table name and fields according to your customizations.

#2. Convert Document into Vector database (RAG ingestion)

Listen for any file changes and update the vector database. The goal is that the llm agent can interact and retrieve information from it later.

#3. Convert Gsheet into MYSQL database

Periodically sync data from gsheet tech radar into mysql database. The goal is so that the llm sql agent can interact with it for certain scenario.

SETUP

Chatting Stage : CHAT ENDPOINT

Purpose

This endpoint api allows you to chat with the ai agent. The ai agent input router will determine if the nature of question best answered with RAG or SQL agent. Then it invokes the workflow rag or workflow db agent accordingly. When the answer comes back , there is an AI agent output guardrail that uses the answer and validates before responding.

How to integrate

1. Connect your frontend interface to this api below. You may change the base endpoint to webhook or webhook-test depending on your environment.

You can also change the based the endpoint 'https://n8n.io' to your own hosted domain like 'https://mycustomdomain.io/'

curl -X POST 'https://n8n.jom.lol/webhook-test/radar-rag' -H 'Content-Type: application/json' -d '{
  "chatInput": "i wanna write agentic ai code, which strategic direction "
}'

curl -X POST 'https://n8n.jom.lol/webhook-test/radar-rag' -H 'Content-Type: application/json' -d '{
  "chatInput": "is backstage used by company2 "
}'

2. You will see a sample output response:

[
    {
        "output": "Based on the researched answer provided, the tools that are considered strategic directions but are not used by company3 are:\n\n* Automatically generate Backstage entity descriptors (technique)\n* Arm in the cloud (platform)\n* Azure OpenAI Service (platform)\n* Infrastructure orchestration platforms (platform)\n* Rancher Desktop (platform)\n* Conan (tool)\n* Karpenter (tool)\n* GitHub Copilot (tool)\n* Renovate (tool)\n* Velero (tool)\n* Continue (tool)\n* LLaVA (tool)\n* Ollama (tool)\n* DataComPy (language and framework)\n* Ray (language and framework)\n* Concrete ML (language and framework)\n* Crux (language and framework)\n* Electric (language and framework)\n* Mojo (language and framework)\n\nThese tools and technologies are considered strategically important for the organization, but for various reasons, they haven't been adopted by company3 yet."
    }
]

[
    {
        "output": "Based on the researched answer provided, the answer to your question is:\n\nAccording to our records, the \"Automatically generate Backstage entity descriptors\" feature, which is related to Backstage, is not currently being used by Company 2. However, it does not necessarily mean that Backstage as a whole platform is not used by Company 2 in some other capacity. If you need more detailed information about Backstage's overall usage in Company 2, please feel free to ask."
    }
]```

CHAT

Setup Stage : Storing into vector and structured sql database

Purpose

This setup is important to ensure that the tech radar google sheets are stored and transformed into the mysql database so that the sql ai agent can interact with it.

For the RAG ai agent to interact with vector database we need ensure we have converted from gsheet into google docs and subsequently into vector database.

Example

Gsheet https://docs.google.com/spreadsheets/d/REPLACE_WITH_DOCUMENT_ID/edit?gid=0#gid=0

Gdoc https://docs.google.com/spreadsheets/d/REPLACE_WITH_DOCUMENT_ID/edit?tab=t.0

Github frontend code

https://github.com/dragonjump/techconstellation/tree/gh-pages

Example Demo

https://raw.githubusercontent.com/dragonjump/techconstellation/refs/heads/gh-pages/build-your-own-radar-master/src/images/image.png

Subworkflow 1: DB SQL Agent

1.Copy and paste into another workflow. 2. Activate it. 3. Link it back

Subworkflow 2: RAG Agent

1.Copy and paste into another workflow. 2. Activate it. 3. Link it back

Tech Radar

Purpose and structure

This folder contains the Tech Radar workflow. Its executable nodes use agent, chainLlm, code, cron, documentDefaultDataLoader, embeddingsGoogleGemini, executeWorkflow, executeWorkflowTrigger, googleDocs, googleDrive, googleDriveTrigger, googleSheets, if, lmChatAnthropic, lmChatGoogleGemini, lmChatGroq, memoryBufferWindow, mySql, mySqlTool, respondToWebhook, textSplitterRecursiveCharacterTextSplitter, toolVectorStore, vectorStorePinecone, webhook. The table and connection map below are derived from the JSON, not an inferred guarantee of behavior.

Setup

Import inactive. Configure your own credentials and replace REPLACE_WITH_* values. Review external destinations and Code nodes before running.

Entry points

Cron, Google Drive - Doc File Updated, When Executed by Another Workflow, API Request - Webhook

Verification

Sanitized source candidate; not runtime-tested. This overview is generated from node metadata. Read the folder README for the connection map, credential types, and license.

Attribution

https://github.com/Zie619/n8n-workflows/blob/94007c1445d9258a7da116646b79473e7c7c3282/workflows/Code/1836_Code_Googledocs_Automation_Webhook.json MIT; preserve source notices.