Mental Model

Overview

The chat path is easiest to hold as five layers. AI sits in Chat; Ground, Capture, Notify, and Reply are context load, tools, push, and UI return. Anthropic names from Building effective agents (PDF).

Applied workflows

• Prompt Chaining (light). Career Agent builds system context (summary + resume + work samples), then may loop OpenAI → tools → OpenAI until a final text reply.
• Not used here (yet). Routing, Parallelization, Orchestrator-Worker, Evaluator-Optimizer — one Me class owns the turn; no peer agents or judge node.

Layers

• 1 Ground. Me.__init__ loads summary.txt, resume PDF, and work-sample PDF into prompt material.
• 2 Chat. Gradio ChatInterface sends message + history into Me.chat; OpenAI gpt-4o-mini runs with the two record tools.
• 3 Capture. On finish_reason == tool_calls, handle_tool_call runs record_user_details or record_unknown_question.
• 4 Notify. Each record tool calls push() → Pushover.
• 5 Reply. When the model returns text (no more tool calls), the assistant message goes back to Gradio.

Layer map

Layer stack for one user turn. Diagram key: ✦ AI agent · second line = Anthropic workflow · purple = actor · white = service.

---
config:
  theme: default
  themeVariables:
    background: "#ffffff"
    fontSize: "14px"
---
flowchart TB
    person[Person]
    ground["1 Ground
resume summary samples"]
    chat["✦ 2 Chat
Prompt Chaining"]
    capture["3 Capture
record tools"]
    notify["4 Notify
Pushover"]
    reply["5 Reply
Gradio"]

    person -->|ask or leave email| chat
    ground --> chat
    chat -->|tool calls| capture --> notify
    capture --> chat
    chat --> reply --> person

    classDef agent fill:#e8f0ee,stroke:#0f3d36,stroke-width:2px,color:#0d0d0d
    classDef service fill:#fafaf8,stroke:#c8c8c2,stroke-width:1px,color:#0d0d0d
    classDef actor fill:#f3eef8,stroke:#6b4f9a,stroke-width:2px,color:#0d0d0d
    class chat agent
    class ground,capture,notify,reply service
    class person actor

Read layers top to bottom; drill into Career Agent for tool schemas and prompt rules.