Skip to main content
What can we help you with?
< All Topics
Print

Tavily & Pinecone Integrations

Posted
Updated
ByPeter Westerman

Optional API Integrations – Tavily & Pinecone

Version: 1.1.0
Date: February 2, 2026
Status: Optional Enhancement Features

## Overview

Two optional third-party services can be integrated to enhance the AI Assistant and search capabilities:

1. Tavily Search API – Enhanced web search for real-time information

2. Pinecone.io Vector Database – Semantic search and long-term conversational memory

Key Points:

– ✅ Completely Optional – All features work without these services

– ✅ User-Controlled – Enable/disable in Settings

– ✅ Privacy-Focused – User consent required for personal data

– ✅ Graceful Fallbacks – App uses local alternatives if disabled

1. Tavily Search API Integration

Purpose

Enhance the AI Assistant with real-time web search capabilities for current information not available in the local knowledgebase.

Use Cases

Research & Information:

  • “Find latest research on companion planting for tomatoes”
  • “What’s the best organic pest control for aphids?”
  • “Find reviews of soil amendments for clay soil”
  • “Research drought-resistant plants for Zone 7b”

Local Resources:

  • “Find nurseries near Atlanta, GA”
  • “Where can I buy heirloom tomato seeds in Atlanta?”
  • “Find landscape fabric suppliers”
  • “Local composting services”

Product Recommendations:

  • “Best grow lights for seedlings 2026”
  • “Compare drip irrigation systems”
  • “Reviews of garden soil test kits”
  • “Recommended rain barrels”

Current Trends:

  • “New vegetable varieties for 2026”
  • “Latest organic gardening techniques”
  • “Current pest outbreaks in Georgia”
  • “New preservation methods”

Features

Real-Time Search:

  • Current information beyond training data
  • Recent research papers and articles
  • Product reviews and comparisons
  • Local business information

Academic Sources:

  • Research papers (Google Scholar)
  • University extension resources
  • Scientific studies
  • Authoritative gardening sites

Source Attribution:

  • Links to original sources
  • Publication dates
  • Author information
  • Credibility indicators

Technical Implementation

API Details:

  • Endpoint: https://api.tavily.com/search
  • Authentication: API Key
  • Rate Limits: Based on plan (free tier available)
  • Response Format: JSON with results and sources

Integration Points:

  • AI Assistant chat interface
  • Triggered when query requires current information
  • Fallback to basic web search if disabled
  • Results formatted and cited inline

Settings:

  • Enable/disable toggle
  • API key management
  • Search result count preference
  • Source filtering options

Privacy:

  • User queries sent to Tavily API
  • No personal garden data sent
  • Search history stored locally only
  • Can be disabled completely

Cost Considerations

Pricing Tiers:

  • Free Tier: Limited searches per month (typically 1,000)
  • Paid Tiers: $10-50/month for higher limits
  • Pay-as-you-go: Option available

Recommendation:

  • Start with free tier
  • Monitor usage in Settings
  • Upgrade if needed
  • Can disable to control costs

2. Pinecone.io Vector Database Integration

Purpose

Enable advanced semantic search across knowledgebase and user content, plus long-term conversational memory for the AI Assistant.

Use Cases

Semantic Search:

  • “Find plants similar to tomatoes”
  • “What pests affect nightshade plants?”
  • “Similar preservation techniques to canning”
  • “Problems related to yellowing leaves”

Conversational Memory:

  • AI remembers past conversations across sessions
  • References previous advice given
  • Builds understanding of user’s garden over time
  • Recalls user preferences and constraints

Related Content Discovery:

  • Find related journal entries
  • Similar problems from past years
  • Related pest/disease profiles
  • Connected preservation methods

Intelligent Suggestions:

  • “You mentioned blight last year – avoid this location”
  • “Based on your past successes with tomatoes…”
  • “Similar to the issue you had in 2024…”

Features

Vector Embeddings:

  • Knowledgebase content embedded:
  • 35+ pest profiles
  • 15+ disease profiles
  • 3,000+ lines preservation methods
  • 2,500+ lines soil strategies
  • User content embedded (with consent):
  • Journal entries
  • Learnings
  • Conversation history

Semantic Similarity Search:

  • Find conceptually similar content
  • Not just keyword matching
  • Understands context and meaning
  • Cross-references related information

Long-Term Memory:

  • Conversations remembered across sessions
  • AI builds user profile over time
  • References past interactions
  • Learns user preferences

Privacy Controls:

  • Knowledgebase: Auto-enabled (public content)
  • User content: Requires explicit consent
  • Clear what data is uploaded
  • One-click delete all data
  • Embeddings generated locally first

Technical Implementation

API Details:

  • Endpoint: https://api.pinecone.io
  • Authentication: API Key
  • Storage: Vector embeddings (not raw text)
  • Dimensions: 1536 (OpenAI ada-002 model)

Database Schema: New table: vector_embeddings

  • Tracks what’s synced to Pinecone
  • Maps local content to vector IDs
  • Enables deletion
  • Records user consent

Embedding Generation:

  • Generated locally (or via OpenAI API)
  • Only vectors uploaded to Pinecone
  • Original text stays local
  • Privacy-preserving

Namespaces:

  • knowledgebase – Public garden knowledge
  • user_journal – User’s journal entries (consent required)
  • user_learnings – User’s learnings (consent required)
  • conversations – Chat history (consent required)

Integration Points:

  1. Search Enhancement:
  • AI Assistant uses semantic search first
  • Falls back to local SQLite if disabled
  • Results ranked by relevance
  1. Conversational Memory:
  • AI references past conversations
  • Builds long-term context
  • Personalizes advice over time
  1. Content Discovery:
  • “Related content” suggestions
  • Automatic linking
  • Cross-references

Settings:

  • Enable/disable toggle
  • API key management
  • Data consent checkboxes:
  • ☐ Upload journal entries
  • ☐ Upload learnings
  • ☐ Upload conversations
  • ☑ Knowledgebase (always enabled)
  • View synced data
  • Delete all data button
  • Usage statistics

Privacy Guarantees:

  1. Consent Required – No user data uploaded without explicit consent
  2. Local First – Embeddings generated locally when possible
  3. Vectors Only – Only mathematical vectors uploaded, not text
  4. User Control – Can delete all data anytime
  5. Transparent – Clear what is stored where
  6. Optional – Entire feature can be disabled

Cost Considerations

Pricing Tiers:

  • Free Tier: 100,000 vectors, 1 index, limited queries
  • Starter: $70/month for 5M vectors
  • Standard: $0.096 per GB-hour (usage-based)

Recommendation:

  • Free tier sufficient for most users
  • Knowledgebase: ~500 vectors
  • User content: ~1,000-5,000 vectors (if opted in)
  • Total: Well within free tier

Cost Control:

  • Monitor vector count in Settings
  • Disable user content embedding if needed
  • Can delete old conversation embeddings
  • Falls back to free local search

Fallback Behavior

When Tavily is Disabled

Fallback: Basic web search using URLSession

  • Search engines via URL parameters
  • Parse HTML results
  • Less comprehensive than Tavily
  • No academic source filtering
  • Still functional, just less powerful

Alternative: Local-only mode

  • Only use knowledgebase content
  • No web search at all
  • Most reliable for privacy
  • Works completely offline

When Pinecone is Disabled

Fallback: Local SQLite full-text search

  • Standard SQL LIKE queries
  • Keyword-based search
  • No semantic understanding
  • Still fast and functional
  • No external dependencies

User Impact:

  • Slightly less intelligent search
  • No long-term memory across sessions
  • No “similar content” discovery
  • All features still work

Settings UI Design

API Configuration Section


┌─ Optional Enhancements ──────────────────────┐
│                                              │
│ These optional services enhance search and   │
│ AI capabilities. All features work without   │
│ them using local alternatives.               │
│                                              │
│ ☐ Enable Tavily Web Search                   │
│   Enhanced web search for current info       │
│   [Test Connection]  [Usage Stats]           │
│   API Key: ••••••••••••••••                  │
│                                              │
│ ☐ Enable Pinecone Semantic Search            │
│   Advanced search and AI memory              │
│   [Test Connection]  [Usage Stats]           │
│   API Key: ••••••••••••••••                  │
│                                              │
│   Data Consent:                              │
│   ☑ Knowledgebase (public content)           │
│   ☐ My journal entries                       │
│   ☐ My learnings                             │
│   ☐ Conversation history                     │
│                                              │
│   [View Synced Data]  [Delete All Data]      │
│                                              │
└──────────────────────────────────────────────┘

Usage Statistics


┌─ API Usage This Month ───────────────────────┐
│                                              │
│ Tavily Search:                               │
│   Searches: 23 / 1,000 (2.3%)                │
│   Estimated Cost: $0.00 (free tier)          │
│                                              │
│ Pinecone Vector DB:                          │
│   Vectors Stored: 1,247 / 100,000 (1.2%)    │
│   Queries: 156 this month                    │
│   Estimated Cost: $0.00 (free tier)          │
│                                              │
└──────────────────────────────────────────────┘

Implementation Checklist

Tavily Integration

Phase 1: Basic Integration

  • [ ] Add Tavily API client class
  • [ ] Implement search request/response
  • [ ] Parse and format results
  • [ ] Add Settings UI controls
  • [ ] Test connection functionality
  • [ ] Error handling

Phase 2: AI Integration

  • [ ] Integrate with AI Assistant
  • [ ] Determine when to trigger search
  • [ ] Format results for AI context
  • [ ] Display sources to user
  • [ ] Cache search results locally

Phase 3: Polish

  • [ ] Usage tracking
  • [ ] Rate limit handling
  • [ ] Fallback to basic search
  • [ ] User documentation

Pinecone Integration

Phase 1: Basic Integration

  • [ ] Add Pinecone API client class
  • [ ] Create vector_embeddings table
  • [ ] Implement embedding generation
  • [ ] Add Settings UI controls
  • [ ] User consent flow
  • [ ] Test connection functionality

Phase 2: Knowledgebase Sync

  • [ ] Embed pest profiles
  • [ ] Embed disease profiles
  • [ ] Embed preservation methods
  • [ ] Embed soil strategies
  • [ ] Sync to Pinecone
  • [ ] Track in database

Phase 3: Search Implementation

  • [ ] Semantic search queries
  • [ ] Result ranking
  • [ ] Integrate with AI Assistant
  • [ ] Display related content
  • [ ] Fallback to local search

Phase 4: User Content (Optional)

  • [ ] Embed journal entries (with consent)
  • [ ] Embed learnings (with consent)
  • [ ] Embed conversations (with consent)
  • [ ] Update on content changes
  • [ ] Delete on content deletion

Phase 5: Polish

  • [ ] Usage tracking
  • [ ] Data management UI
  • [ ] Delete all data feature
  • [ ] User documentation
  • [ ] Privacy policy update

Privacy & Security

Data Handling

What Goes to Tavily:

  • ✅ User’s search queries only
  • ❌ No garden data
  • ❌ No personal information
  • ❌ No plant data

What Goes to Pinecone:

  • ✅ Vector embeddings (numbers only)
  • ✅ Public knowledgebase content
  • ⚠️ User content (only with consent):
  • Journal entries
  • Learnings
  • Conversations
  • ❌ Plant data
  • ❌ Harvest data
  • ❌ Location data

User Rights

Transparency:

  • Clear explanation of what’s shared
  • Before enabling feature
  • In Settings documentation
  • On consent screens

Control:

  • Enable/disable anytime
  • Granular consent (per data type)
  • Delete all data button
  • View what’s synced

Data Deletion:

  • Delete from Pinecone via API
  • Remove from local database
  • Clear tracking records
  • Confirm deletion to user

Benefits Summary

With Tavily Enabled

✅ Better AI Responses:

  • Current information (2026 research)
  • Local resources (Atlanta nurseries)
  • Product recommendations
  • Latest techniques

✅ Enhanced Research:

  • Academic papers
  • Extension resources
  • Expert articles
  • Community knowledge

With Pinecone Enabled

✅ Smarter Search:

  • Find similar plants/problems
  • Conceptual matching
  • Better recommendations
  • Cross-references

✅ Long-Term Memory:

  • AI remembers past discussions
  • Learns your preferences
  • References previous advice
  • Builds garden understanding

Cost: Free Tier Sufficient

Both services offer free tiers that are more than adequate for typical use:

  • Tavily: 1,000 searches/month (plenty for most users)
  • Pinecone: 100K vectors (knowledgebase + user data fits easily)

Conclusion

Tavily and Pinecone are optional enhancements that make the AI Assistant significantly more powerful and useful, while maintaining:

✅ Privacy – User controls what’s shared
✅ Cost – Free tiers sufficient for most users
✅ Functionality – All features work without them
✅ Transparency – Clear what data goes where

Users can start with local-only features and enable these services later if they want enhanced capabilities.

Document Version: 1.1.0
Last Updated: February 2, 2026
Author: IT Influentials
Status: Optional Feature Documentation Complete ✅

Table of Contents
Privacy Policy Terms and Conditions Cookie Policy