Additional Foundry Features We Can Leverage

Last Updated: January 2026
Status: Research & Planning
Source: Azure AI Foundry Documentation

Executive Summary

Azure AI Foundry offers tremendous additional capabilities beyond thread management that we can leverage to enhance Engram:

  1. Foundry IQ - Enterprise data grounding via Azure AI Search
  2. Multi-Agent Orchestration - Built-in agent collaboration workflows
  3. Foundry Tools - Prebuilt production-ready capabilities
  4. Agent Catalog - Discovery and management
  5. Microsoft 365 Integration - Teams, Outlook, SharePoint
  6. Fine-Tuning - Custom model training

1. Foundry IQ - Enterprise Data Grounding

What It Is

Foundry IQ is powered by Azure AI Search and provides a smarter way to ground agents in enterprise data. Agents can connect to a single knowledge base to access multiple sources.

Current State

Engram’s Tri-Search:

  • ✅ Keyword Search (Zep)
  • ✅ Vector Search (pgvector)
  • ✅ Graph Search (Knowledge Graph)

How Foundry IQ Could Enhance

Opportunity: Use Foundry IQ as an additional search layer for enterprise documents

Benefits:

  • ✅ Unified knowledge base (single source for multiple data sources)
  • ✅ Azure AI Search integration (managed infrastructure)
  • ✅ Automatic indexing
  • ✅ Better response quality with broader data access

Integration Approach:

  • Keep Engram’s tri-search for episodic memory and conversation history
  • Use Foundry IQ for document-only enterprise knowledge base
  • Combine results using Reciprocal Rank Fusion (RRF)

Implementation:

# Hybrid search: Engram tri-search + Foundry IQ
engram_results = await memory_client.search_memory(query)
foundry_iq_results = await foundry_iq_client.search(query)

# Combine using RRF
combined_results = reciprocal_rank_fusion([
    engram_results,
    foundry_iq_results
])

2. Multi-Agent Orchestration

What It Is

Foundry supports multi-agent workflows where agents can:

  • Have distinct roles
  • Share memory/context
  • Coordinate tasks
  • Hand off to each other

Current State

Engram’s Agent System:

  • ✅ Three agents (Elena, Marcus, Sage)
  • ✅ Agent router with handoff detection
  • ✅ Separate conversation threads per agent
  • ✅ Agents can reference each other’s work via memory search

How Foundry Orchestration Could Enhance

Opportunity: Use Foundry’s built-in multi-agent orchestration

Benefits:

  • ✅ Built-in agent coordination
  • ✅ Shared context management
  • ✅ Automatic handoff handling
  • ✅ Workflow visualization

Integration Approach:

  • Create Foundry workflow with Elena, Marcus, and Sage
  • Define handoff rules in Foundry
  • Use Foundry’s orchestration for complex multi-agent tasks
  • Keep Engram’s router for simple single-agent requests

Use Cases:

  1. Requirements → Project Planning:
    • Elena creates requirements → Foundry orchestrates handoff → Marcus creates timeline
  2. Project → Story Creation:
    • Marcus identifies need for documentation → Foundry orchestrates → Sage creates story
  3. Complex Multi-Step Tasks:
    • Foundry coordinates all three agents for comprehensive project analysis

3. Foundry Tools - Prebuilt Capabilities

What It Is

Foundry Tools provide ready-to-use APIs for:

  • Content understanding
  • Translation
  • Speech processing
  • Vision analysis
  • Language processing

Current State

Engram’s Tools:

  • ✅ Custom LangChain tools
  • ✅ Microsoft Graph integration
  • ✅ Memory search
  • ✅ GitHub integration

How Foundry Tools Could Enhance

Opportunity: Use Foundry Tools for additional capabilities we don’t have

Potential Tools:

  • Translation - Multi-language support
  • Vision Analysis - Image/document understanding
  • Speech Processing - Enhanced voice capabilities
  • Content Understanding - Better document parsing

Integration Approach:

  • Register Foundry Tools alongside Engram tools
  • Agents can use both Foundry Tools and Engram tools
  • Foundry Tools called via Foundry’s tool execution framework

Example:

# Elena can use both Engram and Foundry tools
tools = [
    # Engram tools
    send_email_tool,
    search_memory_tool,
    # Foundry tools
    foundry_translation_tool,
    foundry_vision_analysis_tool,
]

4. Agent Catalog & Discovery

What It Is

Foundry provides an Agent Catalog for:

  • Discovering available agents
  • Managing agent definitions
  • Versioning agents
  • Sharing agents across projects

Current State

Engram’s Agent Management:

  • ✅ Three agents defined in code
  • ✅ Agent router for selection
  • ✅ Agent info endpoint

How Agent Catalog Could Enhance

Opportunity: Use Foundry’s catalog for agent management

Benefits:

  • ✅ Centralized agent definitions
  • ✅ Version control for agents
  • ✅ Agent discovery across projects
  • ✅ Agent sharing and reuse

Integration Approach:

  • Register Engram agents in Foundry catalog
  • Use catalog for agent discovery
  • Version agents via Foundry
  • Share agents across Engram instances

5. Microsoft 365 Integration

What It Is

Foundry enables publishing agents to:

  • Microsoft Teams - Chat integration
  • Outlook - Email integration
  • SharePoint - Document integration
  • BizChat - Business chat

Current State

Engram’s Microsoft Integration:

  • ✅ Microsoft Graph API (email, OneDrive)
  • ✅ Elena uses elena@zimax.net account
  • ✅ Custom Graph client implementation

How Foundry Integration Could Enhance

Opportunity: Use Foundry’s native Microsoft 365 integration

Benefits:

  • ✅ Native Teams integration
  • ✅ Outlook add-in support
  • ✅ SharePoint connector
  • ✅ Simplified Graph API usage

Integration Approach:

  • Publish Elena to Microsoft Teams
  • Use Foundry’s Outlook integration
  • Leverage SharePoint connectors
  • Keep custom Graph client for advanced features

Use Cases:

  1. Teams Bot: Elena available as Teams bot
  2. Outlook Add-in: Elena helps with email composition
  3. SharePoint: Elena can access SharePoint documents directly

6. Fine-Tuning Capabilities

What It Is

Foundry offers fine-tuning with:

  • Expanded regional support
  • Developer Tier (cost-effective)
  • Custom model training
  • Model versioning

Current State

Engram’s Models:

  • ✅ Uses pre-trained models (GPT-5.2-chat, Claude, Gemini)
  • ✅ No fine-tuning currently

How Fine-Tuning Could Enhance

Opportunity: Fine-tune models for Engram-specific tasks

Potential Use Cases:

  • Fine-tune for requirements analysis (Elena)
  • Fine-tune for project management (Marcus)
  • Fine-tune for technical documentation (Sage)

Integration Approach:

  • Use Foundry’s fine-tuning for specialized models
  • Deploy fine-tuned models in Foundry
  • Reference fine-tuned models in agent definitions

7. Vector Stores (Managed)

What It Is

Foundry provides managed vector stores using Azure AI Search:

  • Automatic embedding generation
  • Managed infrastructure
  • Project-scoped stores

Current State

Engram’s Vector Storage:

  • ✅ Zep with pgvector
  • ✅ Custom embedding client
  • ✅ Tri-search with vector component

How Foundry Vectors Could Enhance

Opportunity: Use Foundry vectors for document-only semantic search

Benefits:

  • ✅ Managed infrastructure
  • ✅ Automatic embedding generation
  • ✅ Project-based isolation
  • ✅ Less operational overhead

Integration Approach:

  • Keep Zep vectors for episodic memory
  • Use Foundry vectors for enterprise documents
  • Combine results using RRF

Decision: Don’t use initially - Zep vectors work well. Consider for document-only use case later.

Implementation Roadmap

Phase 1: Current (Thread Management) ✅

  • ✅ Foundry thread management
  • ✅ Elena migration to Foundry
  • ✅ Tool endpoints

Phase 2: Foundry IQ (Next)

Goal: Add Foundry IQ for enterprise document search

Tasks:

  1. Create Foundry IQ knowledge base
  2. Connect to Azure AI Search
  3. Integrate with Engram’s search
  4. Combine results using RRF

Timeline: 2-3 weeks

Phase 3: Multi-Agent Orchestration

Goal: Use Foundry workflows for complex multi-agent tasks

Tasks:

  1. Create Foundry workflow with Elena, Marcus, Sage
  2. Define handoff rules
  3. Integrate with Engram router
  4. Test complex workflows

Timeline: 3-4 weeks

Phase 4: Foundry Tools

Goal: Add Foundry Tools for additional capabilities

Tasks:

  1. Identify useful Foundry Tools
  2. Register with agents
  3. Integrate tool execution
  4. Test new capabilities

Timeline: 2-3 weeks

Phase 5: Microsoft 365 Integration

Goal: Publish agents to Teams/Outlook

Tasks:

  1. Publish Elena to Teams
  2. Create Outlook add-in
  3. Test integration
  4. Document usage

Timeline: 3-4 weeks

Feature Comparison Matrix

Feature Engram Current Foundry Alternative Recommendation
Thread Management In-memory Foundry threads ✅ Use Foundry
Vector Search Zep pgvector Foundry vectors ⚠️ Keep Zep (works well)
Multi-Agent Custom router Foundry orchestration ✅ Use Foundry for complex workflows
Enterprise Search Tri-search Foundry IQ ✅ Use Foundry IQ for documents
Tools Custom LangChain Foundry Tools ✅ Use both (hybrid)
Microsoft 365 Graph API Foundry integration ✅ Use Foundry for Teams/Outlook
Fine-Tuning None Foundry fine-tuning ⚠️ Consider for specialized models

Immediate (This Week)

  1. ✅ Complete Elena migration to Foundry
  2. ✅ Store Foundry configuration in Key Vault
  3. ✅ Test Foundry Elena with Microsoft Graph tools

Short Term (Next 2-4 Weeks)

  1. Research Foundry IQ:
    • Evaluate Azure AI Search integration
    • Test with enterprise documents
    • Compare with Engram tri-search
  2. Explore Multi-Agent Orchestration:
    • Create Foundry workflow
    • Test agent handoffs
    • Compare with Engram router

Medium Term (Next 1-2 Months)

  1. Foundry Tools Integration:
    • Identify useful tools
    • Register with agents
    • Test capabilities
  2. Microsoft 365 Integration:
    • Publish to Teams
    • Create Outlook add-in
    • Test user experience

Summary

Foundry Features to Leverage:

  1. ✅ Thread Management - Already implementing
  2. 🎯 Foundry IQ - Enterprise document search (high value)
  3. 🎯 Multi-Agent Orchestration - Complex workflows (high value)
  4. 🎯 Foundry Tools - Additional capabilities (medium value)
  5. 🎯 Microsoft 365 Integration - Teams/Outlook (high value)
  6. ⚠️ Vector Stores - Keep Zep for now (low priority)
  7. ⚠️ Fine-Tuning - Consider later (low priority)

Strategy: Hybrid Approach

  • Use Foundry for infrastructure and orchestration
  • Keep Engram’s unique capabilities (tri-search, custom tools)
  • Leverage Foundry’s strengths (IQ, orchestration, M365 integration)

Last Updated: January 2026