Chat & Search

How It Works

When you ask a question, Zedly doesn't just send your query to an AI. We use a technique called Retrieval-Augmented Generation (RAG) to find relevant passages from your documents first, then generate an answer based on that context.

Query Processing Flow

1 Your question is analyzed for intent

2 Relevant documents on your Desk are identified

3 Semantic search finds matching passages

4 Results are ranked by relevance

5 AI generates answer from retrieved context

This means answers are grounded in your actual documents, not made up. You can trust the response because it comes directly from your files.

Types of Queries

Zedly automatically detects what type of question you're asking and routes it appropriately:

🔍

Search Queries

Find specific information, clauses, or mentions across your documents.

📊

Analysis Queries

Summarize, compare, or extract patterns from document content.

💬

Conversational

Follow-up questions that build on previous answers in the chat.

📈

Data Queries

For Excel/CSV files: calculations, charts, and data analysis.

Example Queries

For Contracts & Legal Documents

Find specific terms

"What are the termination clauses in this agreement?"

Compare across documents

"Compare the liability caps across all vendor contracts"

Extract key dates

"List all renewal dates and notice periods"

For Financial Documents

Summarize transactions

"What were the largest expenses last month?"

Find patterns

"Show recurring payments over $500"

For Data Files (Excel/CSV)

Create visualizations

"Graph revenue by quarter for 2024"

Calculate metrics

"What's the average order value by region?"

Tips for Better Results

Best Practices

Be specific: "What is the indemnification limit in the Acme contract?" works better than "Tell me about indemnification"
Reference document names: If you have multiple files, mention which one: "In the Q4 report, what was..."
Ask follow-ups: Build on previous answers: "What about the payment terms in that same section?"
Use natural language: No need for keywords—write like you're asking a colleague

Important

AI can make mistakes. Always verify critical information by checking the source references provided in responses. Zedly shows which document and section the answer came from.

Documents vs. Data Files

Zedly handles text documents and data files differently:

📄 Documents (PDF, DOCX, TXT)

Processed via RAG (semantic search)
Good for contracts, reports, manuals
Answers cite specific passages
Uses embedding tokens from your plan

📊 Data Files (XLSX, CSV)

Processed via code execution
Good for calculations, charts, analysis
Can create visualizations
Uses analysis credits from your plan

You can have both types on your Active Desk at the same time. Zedly automatically routes your query to the right processing method.

Chat Limits

Each plan includes a monthly chat message quota:

Free: 100 messages/month
Starter: 1,000 messages/month
Pro: 5,000 messages/month
Business: Unlimited

Data analysis (charts, calculations) uses separate "analysis credits" which are also tracked monthly.