# Gen AI

Create Authentication and Autorization layer for GenAI platform. Use Keycloak with MySQL. Keycloak is multi tenant. Our following components should also be multi tenant. So there will be customer CRUD somewhere and users belonging to particular customer.

### Connectors for RAG
> We will keep on adding connectors in future
Simple authentication for OAuth for following services. Use a simple config page specific to client that allow authentication all below api's. For google drive authentication look at this video (https://www.youtube.com/watch?v=nE7XATKx6iQ) from 2.35 min. Try using Mayank. And put these things in JIRA

1. Ingestion Connector for Google Drive
2. Ingestion Connector for Microsoft One Drive
3. Ingestion Connector for S3 (Use secret key and access key)

> Multi tenant - customer id below represent tenent

| Customer Id | Connectors | Authentication Json |
| ----------- | ----------- |----------- |
| indigo | google drive | {bearer: xxxx} |
| indigo | s3 | {access_key : xxxx, secret_key : yyyy} |
| itq | one drive | {bearer: zzzz} |

#### Ingestion support
One time ingestion - batch ingestion
Increment knowledge ingestion based on triggers

### RAG accelerator
- Chunking - Apply agentic chunking for above
- Use - Langchain, Qdrant
- Use - Spring boot (Mongo DB or MySQl) for other master data
- Use open source model all three public clouds (Configurable) as well as ChatGPT

### Simple chat interface
For customer specific users on that customer documents.

## Connectors for Marketing Agents (Multi tenant).
> Table simlar to above. Use same approach. Look at how n8n simplify it.

### Able to create audiences and run marketing campaign using simple UI
- Meta (Instagram and Facebook)
- Google
- Twitter
- Linked In

#### Low priority
- Whatsapp (Via Gupshup, and WABA)
- Email (Use SES)
- SMS (Use some hosted solution)

### Agentic AI components
- TBD