Cloud Drive Sync & RAG Pipeline
Retrieval-Augmented Generation (RAG) ensures your chatbot delivers accurate, hallucination-free answers grounded in your company's actual internal documentation. Chatty automatically indexes and synchronizes files from Google Drive, Microsoft OneDrive, Notion, and live web sitemaps directly into high-dimensional vector embeddings.
Supported Knowledge Connectors
Google Drive
Connect Google Workspace or personal Drive folders. Continuously synchronizes Google Docs, Sheets, Slides, and uploaded PDFs.
Microsoft OneDrive / SharePoint
Index enterprise SharePoint libraries and OneDrive folders with real-time Microsoft Graph delta updates.
Notion Knowledge Bases
Sync Notion team workspaces, documentation wikis, and databases. Honors page nesting hierarchies.
Web Sitemap Crawler
Supply an XML sitemap or root URL. Chatty crawls child links, strips HTML boilerplate using Jina Reader, and indexes clean content.
The RAG Ingestion Pipeline
Chunking & Embedding Specifications
To preserve semantic context without exceeding LLM attention spans, documents are processed according to strict indexing parameters:
Continuous Synchronization
Documents evolve constantly. Chatty prevents stale information through automated sync mechanisms:
- Delta Sync Webhooks: For Google Drive and OneDrive, webhooks alert Chatty whenever a file is updated or deleted. Only modified documents are re-chunked.
- SHA-256 Content Hashing: Before computing new embeddings, Chatty checks document content hashes. If a file was saved without text changes, vector recalculation is skipped to save compute quotas.
- Scheduled Sitemap Polling: Web sources are recrawled on a configurable cadence (Daily, Weekly, or Monthly) to pick up newly published blog posts or documentation updates.
Citations & Source Attribution
When Chatty answers using RAG knowledge, it extracts metadata (document title, source URL, page number) and appends formatted citation chips:
Based on our enterprise service agreement, customers are entitled to 99.9% uptime
with 24/7 priority phone support [1]. Migration assistance is included at no extra charge [2].
Sources:
[1] Enterprise SLA Guidelines.pdf (Page 4)
[2] https://docs.example.com/onboarding/migration
You can manage, re-index, or delete individual knowledge sources programmatically using the Add Knowledge API and Delete Knowledge API.