How discovery makes your AI smarter
Discovery analyzes your Odoo in 90 seconds to build a knowledge base. The AI uses this to give context-aware answers about your specific business, not generic responses.
What is deep discovery?
AI-powered RAG system analyzes your Odoo installation
Deep discovery is an AI-powered RAG (Retrieval-Augmented Generation) system that analyzes your Odoo installation in five phases:
- Analyzes your Odoo modules and configuration
- Discovers relevant content (products, docs, discussions, published content)
- Embeds this content into a semantic search database using vector representations
- Generates an AI intelligence summary of your business
- Enhances chat responses by automatically retrieving relevant context
Think of discovery as giving your AI assistant comprehensive knowledge of your specific Odoo installation and business context.
Why use deep discovery?
Compare AI responses with and without business context
Without discovery
- •Generic, unhelpful responses with no business context
- •Multiple slow tool calls required
- •Raw data without interpretation
With discovery
- •Contextual, intelligent responses with business insights
- •Organized by relevance
- •Faster responses, fewer database queries
How discovery works
Five-phase pipeline from analysis to query-time retrieval
┌─────────────────────────────────────────────┐
│ 1. AI planning (Sonnet 4.5) │
│ Analyzes installed modules │
│ Decides which phases to run │
└─────────────────────────────────────────────┘
↓
┌─────────────────────────────────────────────┐
│ 2. Parallel discovery (supervisor pattern) │
│ ├─ Schema analysis │
│ ├─ Product catalog │
│ ├─ Customer insights │
│ └─ Additional phases │
└─────────────────────────────────────────────┘
↓
┌─────────────────────────────────────────────┐
│ 3. Embedding (Voyage AI) │
│ Converts text → 1024-dimension vectors │
└─────────────────────────────────────────────┘
↓
┌─────────────────────────────────────────────┐
│ 4. Intelligence summary (Sonnet 4.5) │
│ Generates comprehensive business report │
└─────────────────────────────────────────────┘
↓
┌─────────────────────────────────────────────┐
│ 5. Query-time RAG │
│ Semantic search finds relevant content │
└─────────────────────────────────────────────┘Privacy and security
What data gets analyzed and what stays private
What gets discovered
Public and published content:
- Product catalogs (prices, descriptions)
- Published knowledge articles
- Public website pages
- Public team discussion channels
- Aggregate statistics (example: "127 customers in tech industry")
What does not get discovered
Private and sensitive data:
- Individual customer names, emails, addresses
- Employee personal information, salaries, reviews
- Private messages or internal emails
- Financial details (specific invoices, payments, balances)
- Passwords, API keys, credentials
- Draft or unpublished content
Performance
Typical processing times by business size
| Business size | Modules | Items | Tokens | Time |
|---|---|---|---|---|
| Small | 5-8 | 50-80 | 15K | 25s |
| Medium | 8-12 | 80-150 | 25K | 45s |
| Large | 12-16 | 150-300 | 40K | 75s |
Query-time benefits
Token efficiency: Up to 60% reduction in tokens per conversation with discovery enabled
When to run discovery
First-time setup and regular update schedules
First-time setup
Run discovery immediately after connecting Odoo for instant business context and intelligent responses.
Regular updates
Re-run discovery when you:
- Add new products (major catalog changes)
- Install new Odoo modules
- Publish new knowledge articles or website content
- Want quarterly or monthly refresh (for active businesses)
Do not re-run for daily data changes, individual customer updates, or stock level changes. Discovery captures structure and knowledge, not transactional data.
Shared instances (teams)
Efficient for teams
If multiple team members use the same Odoo instance, run discovery once and everyone benefits from the shared knowledge base.
Multi-layer knowledge system
Advanced Odoo intelligence that continuously learns and updates
The multi-layer knowledge system combines multiple intelligence sources to give you the most accurate and up-to-date Odoo assistance:
Layer 1: Version-Aware Official Documentation
Automatically fetches the correct Odoo documentation for your specific version (v14-19 supported).
- Module-specific guides and API references
- Version differences automatically handled
- Community and Enterprise edition documentation
Layer 2: Your Instance Discovery Data
Contextual knowledge about your specific Odoo setup and business data.
- Installed modules and configurations
- Product catalogs and business structure
- Custom fields and workflows
Layer 3: Community Intelligence
Aggregated knowledge from Odoo community forums, best practices, and common patterns.
- Common issues and solutions
- Best practices for Odoo operations
- Module compatibility insights
Automatic Knowledge Gap Detection
The system automatically detects when it encounters unknown models, fields, or patterns and flags them for learning. Knowledge gaps are tracked and filled over time through discovery updates and community contributions.
Knowledge transparency
See exactly what the AI knows and doesn't know
Every AI response includes transparency indicators showing knowledge freshness, confidence levels, and data sources. This helps you understand when to trust AI responses and when to verify independently.
Visual indicators you'll see
Knowledge Freshness Badge
Shows how recently the AI's knowledge about your Odoo instance was updated.
Confidence Level Display
Each response shows how confident the AI is based on available knowledge sources.
Knowledge Source Citations
Responses include references showing which knowledge layers were used.
When to verify independently
Always verify AI responses for critical business decisions, especially when confidence is below 70% or knowledge freshness shows "Stale". The transparency indicators help you know when independent verification is recommended.
Best practices
Maximize discovery effectiveness with these recommendations
Run discovery on clean data: Publish important content before running discovery
Schedule regular updates: Monthly for active e-commerce, quarterly for stable catalogs
Monitor progress: Watch for completion or connectivity issues
Leverage in prompts: Reference discovery explicitly in your questions
Check transparency indicators: Use knowledge freshness and confidence levels to gauge response reliability