AGENT HANDOFF

2 Squirrels AI

Technology Stack

Agent-to-Agent (A2A) Protocol implementation demonstrating standardized communication between autonomous AI agents across multiple frameworks including LangGraph, CrewAI, LlamaIndex, and Semantic Kernel.

Protocol A2A (JSON-RPC 2.0)
LLM Google Gemini
Framework LangGraph
Framework CrewAI
Framework LlamaIndex
RAG LlamaParse

Multi-Framework Agent Ecosystem

🔄
LangGraph Agent
ReAct Pattern + Memory
🎨
CrewAI Agent
Image Generation
📄
LlamaIndex Agent
Document RAG
🏠
Host Agent
Task Orchestration
🔌
Semantic Kernel
Plugin Architecture

A2A Protocol: Enables secure, standardized communication between independent agents. Agents treat external entities as untrusted, supporting streaming responses, push notifications, and multi-turn conversations.

Workflow 1: LangGraph Currency Conversion Agent

%%{init: {'theme': 'dark', 'themeVariables': { 'primaryColor': '#C17852', 'primaryTextColor': '#F0F6FC', 'primaryBorderColor': '#4A5E32', 'lineColor': '#E6C98F', 'secondaryColor': '#161B22', 'tertiaryColor': '#0D1117', 'background': '#0D1117', 'mainBkg': '#161B22', 'nodeBorder': '#4A5E32', 'clusterBkg': '#161B22', 'clusterBorder': '#4A5E32', 'titleColor': '#E6C98F', 'edgeLabelBackground': '#161B22'}}}%%
flowchart TB
    subgraph Input["📥 CLIENT REQUEST"]
        REQ[/"💱 Currency Query"/]
    end

    subgraph Memory["🧠 STATE MANAGEMENT"]
        CHKPT[("💾 Memory Checkpoint
Multi-turn Context")]
    end

    subgraph LLM["🤖 GEMINI REASONING"]
        PARSE["📝 Parse Request"]
        CHECK{"🔍 Info Complete?"}
    end

    subgraph Conditional["🔀 CONDITIONAL ROUTING"]
        ASK["❓ Request Clarification
state: input-required"]
        TOOL["🔧 Invoke Tool"]
    end

    subgraph ToolCall["🌐 EXTERNAL API"]
        FX["💹 Frankfurter API
get_exchange_rate()"]
    end

    subgraph Stream["📡 STREAMING"]
        S1["Looking up rates..."]
        S2["Processing..."]
    end

    subgraph Output["📤 RESPONSE"]
        RES["✅ Exchange Rate Result
state: completed"]
    end

    REQ --> CHKPT
    CHKPT --> PARSE
    PARSE --> CHECK
    CHECK -->|"No"| ASK
    ASK -->|"User Response"| CHKPT
    CHECK -->|"Yes"| TOOL
    TOOL --> FX
    FX --> S1
    S1 --> S2
    S2 --> RES

Workflow 2: CrewAI Image Generation Agent

%%{init: {'theme': 'dark', 'themeVariables': { 'primaryColor': '#C17852', 'primaryTextColor': '#F0F6FC', 'primaryBorderColor': '#4A5E32', 'lineColor': '#E6C98F', 'secondaryColor': '#161B22', 'tertiaryColor': '#0D1117', 'background': '#0D1117', 'mainBkg': '#161B22', 'nodeBorder': '#4A5E32', 'clusterBkg': '#161B22', 'clusterBorder': '#4A5E32', 'titleColor': '#E6C98F', 'edgeLabelBackground': '#161B22'}}}%%
flowchart LR
    subgraph Input["📥 INPUT"]
        PROMPT["🖼️ Image Prompt"]
        REF[/"📎 Reference Image
(Optional)"/]
    end

    subgraph Agent["🤖 CREWAI AGENT"]
        PARSE["📝 Parse Request"]
        INIT["🎭 Image Creation Expert"]
        TASK["📋 Define Task"]
    end

    subgraph Tool["🔧 TOOL EXECUTION"]
        GEN["🎨 Gemini Image API"]
        RETRY["🔄 Retry Logic"]
    end

    subgraph Cache["💾 SESSION CACHE"]
        STORE[("🗄️ In-Memory Cache
Unique ID")]
    end

    subgraph Output["📤 OUTPUT"]
        B64["📦 Base64 Encode"]
        ART["🖼️ Image Artifact
+ Cache Reference"]
    end

    PROMPT --> PARSE
    REF --> PARSE
    PARSE --> INIT
    INIT --> TASK
    TASK --> GEN
    GEN -->|"Failure"| RETRY
    RETRY --> GEN
    GEN -->|"Success"| STORE
    STORE --> B64
    B64 --> ART

Workflow 3: LlamaIndex Document RAG Pipeline

%%{init: {'theme': 'dark', 'themeVariables': { 'primaryColor': '#C17852', 'primaryTextColor': '#F0F6FC', 'primaryBorderColor': '#4A5E32', 'lineColor': '#E6C98F', 'secondaryColor': '#161B22', 'tertiaryColor': '#0D1117', 'background': '#0D1117', 'mainBkg': '#161B22', 'nodeBorder': '#4A5E32', 'clusterBkg': '#161B22', 'clusterBorder': '#4A5E32', 'titleColor': '#E6C98F', 'edgeLabelBackground': '#161B22'}}}%%
flowchart TB
    subgraph Input["📥 CLIENT MESSAGE"]
        MSG["💬 Query"]
        FILE[/"📎 Document
(PDF, DOCX)"/]
    end

    subgraph Parse["📄 DOCUMENT PROCESSING"]
        STREAM1["📡 Parsing document..."]
        LLAMA["🦙 LlamaParse
Extract + Line Numbers"]
        STREAM2["✅ Document parsed"]
    end

    subgraph Context["📝 CONTEXT BUILDING"]
        SYS["🔧 System Prompt
+ Document Context"]
        HIST[("💬 ChatHistory
Multi-turn Memory")]
    end

    subgraph LLM["🤖 LLM QUERY"]
        GEMINI["🧠 Gemini API
Query + Context"]
        CITE["📌 Extract Citations"]
    end

    subgraph Output["📤 RESPONSE"]
        MAP["🔗 Map Citation Numbers
to Line Numbers"]
        ART["📄 Response Artifact
+ Citation Metadata"]
    end

    MSG --> Context
    FILE --> STREAM1
    STREAM1 --> LLAMA
    LLAMA --> STREAM2
    STREAM2 --> SYS
    SYS --> HIST
    HIST --> GEMINI
    GEMINI --> CITE
    CITE --> MAP
    MAP --> ART
    ART -->|"Follow-up"| HIST

Workflow 4: Multi-Agent Trip Planner Orchestration

%%{init: {'theme': 'dark', 'themeVariables': { 'primaryColor': '#C17852', 'primaryTextColor': '#F0F6FC', 'primaryBorderColor': '#4A5E32', 'lineColor': '#E6C98F', 'secondaryColor': '#161B22', 'tertiaryColor': '#0D1117', 'background': '#0D1117', 'mainBkg': '#161B22', 'nodeBorder': '#4A5E32', 'clusterBkg': '#161B22', 'clusterBorder': '#4A5E32', 'titleColor': '#E6C98F', 'edgeLabelBackground': '#161B22'}}}%%
flowchart TB
    subgraph Input["📥 USER REQUEST"]
        REQ[/"🗺️ Complex Trip Query
Location + Dates + Budget"/]
    end

    subgraph Host["🎯 HOST AGENT"]
        PARSE["📝 Parse Request"]
        ANALYZE["🔍 Identify Subtasks"]
    end

    subgraph Parallel["⚡ PARALLEL A2A CALLS"]
        subgraph Weather["☁️ WEATHER AGENT"]
            W_RECV["📨 Receive Query"]
            W_API["🌤️ Weather API"]
            W_RES["📤 Conditions"]
        end
        subgraph Booking["🏠 AIRBNB AGENT"]
            B_RECV["📨 Receive Query"]
            B_MCP["🔌 MCP Airbnb Server"]
            B_RES["📤 Availability"]
        end
    end

    subgraph Stream["📡 STREAMING"]
        S1["Checking weather..."]
        S2["Finding accommodations..."]
    end

    subgraph Aggregate["🔗 AGGREGATION"]
        COLLECT["📥 Collect Responses"]
        SYNTH["✨ Synthesize Plan"]
    end

    subgraph Output["📤 FINAL OUTPUT"]
        ITIN["📋 Trip Itinerary
Weather + Booking Details"]
    end

    REQ --> PARSE
    PARSE --> ANALYZE
    ANALYZE --> W_RECV
    ANALYZE --> B_RECV
    W_RECV --> W_API
    W_API --> W_RES
    B_RECV --> B_MCP
    B_MCP --> B_RES
    W_RES --> COLLECT
    B_RES --> COLLECT
    ANALYZE --> S1
    S1 --> S2
    COLLECT --> SYNTH
    SYNTH --> ITIN

Workflow 5: Multi-Agent Game Loop

%%{init: {'theme': 'dark', 'themeVariables': { 'primaryColor': '#C17852', 'primaryTextColor': '#F0F6FC', 'primaryBorderColor': '#4A5E32', 'lineColor': '#E6C98F', 'secondaryColor': '#161B22', 'tertiaryColor': '#0D1117', 'background': '#0D1117', 'mainBkg': '#161B22', 'nodeBorder': '#4A5E32', 'clusterBkg': '#161B22', 'clusterBorder': '#4A5E32', 'titleColor': '#E6C98F', 'edgeLabelBackground': '#161B22'}}}%%
flowchart TB
    subgraph Init["🎮 GAME INIT"]
        ALICE["🎯 Alice Agent
Picks Secret (1-100)"]
    end

    subgraph UserTurn["👤 USER TURN (BOB)"]
        PROMPT["📡 Stream: Enter guess"]
        GUESS["🔢 Receive Guess"]
    end

    subgraph Evaluate["🤔 EVALUATION (ALICE)"]
        A2A_ALICE["📨 A2A Message/Send"]
        COMPARE["⚖️ Compare to Secret"]
        RESULT{"Result?"}
    end

    subgraph Visualize["📊 VISUALIZATION (CAROL)"]
        A2A_CAROL["📨 Request History"]
        ASCII["🎨 Generate ASCII
Visualization"]
        SORT["🔄 Reshuffle History"]
    end

    subgraph Loop["🔁 GAME LOOP"]
        SHOW["📺 Display to User"]
        CHECK{"Correct?"}
    end

    subgraph End["🏆 COMPLETE"]
        WIN["✅ Correct!
Attempts: N"]
    end

    ALICE --> PROMPT
    PROMPT --> GUESS
    GUESS --> A2A_ALICE
    A2A_ALICE --> COMPARE
    COMPARE --> RESULT
    RESULT -->|"Higher/Lower"| A2A_CAROL
    A2A_CAROL --> ASCII
    ASCII --> SORT
    SORT --> SHOW
    SHOW --> CHECK
    CHECK -->|"No"| PROMPT
    RESULT -->|"Correct"| WIN
    CHECK -->|"Yes"| WIN

Workflow 6: Semantic Kernel Plugin Architecture

%%{init: {'theme': 'dark', 'themeVariables': { 'primaryColor': '#C17852', 'primaryTextColor': '#F0F6FC', 'primaryBorderColor': '#4A5E32', 'lineColor': '#E6C98F', 'secondaryColor': '#161B22', 'tertiaryColor': '#0D1117', 'background': '#0D1117', 'mainBkg': '#161B22', 'nodeBorder': '#4A5E32', 'clusterBkg': '#161B22', 'clusterBorder': '#4A5E32', 'titleColor': '#E6C98F', 'edgeLabelBackground': '#161B22'}}}%%
flowchart TB
    subgraph Input["📥 TRIP REQUEST"]
        REQ[/"🗺️ Trip + Budget Query"/]
    end

    subgraph Manager["🎯 TRAVEL MANAGER AGENT"]
        PARSE["📝 Extract Parameters
Destination, Dates, Budget"]
    end

    subgraph Plugins["🔌 PARALLEL PLUGIN AGENTS"]
        subgraph Currency["💱 CURRENCY AGENT"]
            C_INV["invoke()"]
            C_TOOL["🔧 get_exchange_rate"]
            C_API["💹 Frankfurter API"]
            C_RES["💵 Converted Budget"]
        end
        subgraph Activity["🎯 ACTIVITY AGENT"]
            A_INV["invoke()"]
            A_LLM["🧠 LLM Planning"]
            A_RES["📋 Recommendations"]
        end
    end

    subgraph Stream["📡 STATUS UPDATES"]
        S1["Analyzing requirements..."]
        S2["Converting currency..."]
        S3["Building itinerary..."]
        S4["Finalizing plan..."]
    end

    subgraph Aggregate["🔗 SYNTHESIS"]
        COLLECT["📥 Aggregate Responses"]
        COMBINE["✨ Combine Data"]
    end

    subgraph Output["📤 TRIP PLAN ARTIFACT"]
        PLAN["📋 Complete Itinerary
Budget + Activities + Costs"]
    end

    REQ --> PARSE
    PARSE --> C_INV
    PARSE --> A_INV
    C_INV --> C_TOOL
    C_TOOL --> C_API
    C_API --> C_RES
    A_INV --> A_LLM
    A_LLM --> A_RES
    PARSE --> S1
    S1 --> S2
    S2 --> S3
    S3 --> S4
    C_RES --> COLLECT
    A_RES --> COLLECT
    COLLECT --> COMBINE
    COMBINE --> PLAN