Skip to content

ai_providers_data_flow.md

Latest commit

 

History

History
151 lines (115 loc) · 4.04 KB

ai_providers_data_flow.md

File metadata and controls

151 lines (115 loc) · 4.04 KB

AI Providers Data Flow

This document describes how the AI Workflows system interacts with AI model providers for automated decision-making and content generation.

AI Provider Architecture

graph TD
    subgraph "Agent Framework"
        BEEAI["BeeAI Framework<br/>(Agent Orchestration)"]
        AGENTS["AI Agents<br/>(Rebase, Backport, MR, Triage)"]
    end

    subgraph "AI Model Access Layer"
        MCP["MCP Server<br/>(Gateway)"]
        CHAT["Chat Service<br/>(chat.py)"]
    end

    subgraph "Google Cloud Platform"
        VERTEX["Vertex AI<br/>(Gemini Models)"]
        GCP_AUTH["Service Account<br/>jotnar-vertex-prod.json"]
    end

    PHOENIX["Phoenix<br/>(Observability & Tracing)"]

    BEEAI -->|"Orchestrates"| AGENTS
    AGENTS -->|"Agent Requests"| MCP
    MCP -->|"Model Calls"| CHAT
    CHAT -->|"Authenticated API Calls"| GCP_AUTH
    GCP_AUTH -->|"API Request"| VERTEX
    VERTEX -->|"Model Response"| CHAT
    CHAT -->|"Response"| MCP
    MCP -->|"Tool Results"| AGENTS

    CHAT -.->|"Trace API Calls"| PHOENIX
    VERTEX -.->|"Log Responses"| PHOENIX

    style BEEAI fill:#f3e5f5,stroke:#7b1fa2,stroke-width:3px
    style AGENTS fill:#e8f5e9,stroke:#388e3c,stroke-width:3px
    style MCP fill:#fff4e6,stroke:#f57c00,stroke-width:3px
    style CHAT fill:#fff4e6,stroke:#f57c00,stroke-width:3px
    style VERTEX fill:#e3f2fd,stroke:#1976d2,stroke-width:3px
    style GCP_AUTH fill:#fff9c4,stroke:#f57f17,stroke-width:3px
    style PHOENIX fill:#fce4ec,stroke:#c2185b,stroke-width:3px
Loading

Google Cloud Platform Integration

Service Accounts

Project Purpose API Key Storage
jotnar-bot Production deployment Bitwarden: jotnar-vertex-prod.json
packit-automated-packaging Development and testing GCP Console

Authentication Flow

sequenceDiagram
    participant Agent as AI Agent
    participant Chat as Chat Service
    participant GCP as Google Cloud
    participant Vertex as Vertex AI API

    Agent->>Chat: Request model inference
    Chat->>Chat: Load service account key<br/>(jotnar-vertex-prod.json)
    Chat->>GCP: Authenticate with service account
    GCP-->>Chat: OAuth2 token
    Chat->>Vertex: API request with token
    Vertex->>Vertex: Execute model inference
    Vertex-->>Chat: Model response
    Chat-->>Agent: Processed response
Loading

AI Model Usage

Models in Use

Primary Model: Gemini (Google Vertex AI)

Use Cases:

  • Spec file analysis and modification
  • Patch backporting and application
  • Build failure diagnosis and fixing
  • Test result analysis
  • Jira issue triage
  • Merge request review

Agent-to-Model Communication

flowchart LR
    TRIAGE["Triage Agent"]
    BACKPORT["Backport Agent"]
    REBASE["Rebase Agent"]
    MR_AGENT["MR Agent"]
    TESTING["Testing Analyst"]

    CHAT["Chat Service<br/>(Gemini)"]

    TRIAGE -->|"Analyze Jira issue<br/>Determine workflow"| CHAT
    BACKPORT -->|"Review patches<br/>Apply to spec"| CHAT
    REBASE -->|"Update version<br/>Resolve conflicts"| CHAT
    MR_AGENT -->|"Review MR comments<br/>Address feedback"| CHAT
    TESTING -->|"Analyze test results<br/>Triage failures"| CHAT

    CHAT -->|"Structured responses"| TRIAGE
    CHAT -->|"Patch recommendations"| BACKPORT
    CHAT -->|"Conflict resolution"| REBASE
    CHAT -->|"Action decisions"| MR_AGENT
    CHAT -->|"Test analysis"| TESTING

    style CHAT fill:#e3f2fd,stroke:#1976d2,stroke-width:3px
Loading

Configuration

Environment Variables

Chat Service Configuration:

# Model selection
CHAT_MODEL=gemini-1.5-pro

# API credentials
GOOGLE_APPLICATION_CREDENTIALS=/etc/secrets/jotnar-vertex-prod.json

# Model parameters
MAX_RETRIES=3
TEMPERATURE=0.7

Agent Configuration:

# BeeAI framework settings
BEEAI_MAX_ITERATIONS=10
BEEAI_TIMEOUT=300

OpenShift Configuration

Secrets:

  • vertex-key - Contains jotnar-vertex-prod.json service account key

ConfigMaps:

  • chat-env - Chat model configuration
  • agents-env - Agent-specific model parameters

Last Updated: 2026-03-03