What is Context?
Context is the data layer that powers AI decision-making. It provides agents with:- Business data
- Historical records
- Real-time signals
- Structured knowledge
With context, AI generates accurate, actionable, and domain-specific insights.
Types of Context in Avigrah
Avigrah supports multiple context sources, each designed for a specific purpose.1. Integration Context (Live Systems)
Integration context connects AI directly to external systems. Examples include:- CRM systems (e.g., Salesforce)
- E-commerce platforms
- Internal business tools
How it works
- Data is fetched directly from connected systems
- AI agents can query live operational data
- Context reflects the current state of the business
Use Cases
- Customer activity analysis
- Sales pipeline insights
- Operational decision-making
2. Standard RAG Context (Retrieval Layer)
This is the retrieval-based context system used for unstructured and semi-structured data. It works by:- Storing vectorized data in a vector database
- Retrieving relevant information during AI execution
- Providing contextual grounding for responses
Use Cases
- Document understanding
- Knowledge base querying
- Historical data reasoning
3. Lakehouse Context (Structured Data)
Lakehouse context provides AI with access to structured, stored datasets. This includes:- Processed pipeline data
- Tabular datasets
- Historical records
How it works
- Data is stored in a structured format
- AI agents query this data directly
- Enables precise and deterministic analysis
Use Cases
- Analytics and reporting
- Aggregations and trends
- Financial and operational insights
4. Redis Cache Context (Fast Access Layer)
Cache context is used for low-latency, high-frequency data access. It stores:- Recently accessed data
- Frequently used signals
- Intermediate results
How it works
- Data is temporarily stored in cache
- AI agents can access it instantly
- Reduces repeated computation and latency
Use Cases
- Real-time recommendations
- Session-based interactions
- High-speed decision systems
How Context Works Together
Avigrah does not rely on a single context source. Instead, it combines multiple layers:- Integrations → Live operational data
- RAG → Retrieved knowledge
- Lakehouse → Structured datasets
- Cache → Fast access memory
Context Selection (Agent-Level)
Each AI agent is configured to use specific context sources. This ensures:- Relevant data is used for each task
- No unnecessary data exposure
- Better performance and accuracy