A colleague at Nordlicht Reisen GmbH has built a small prototype in Python for destination ideas, ideenlabor, with LangGraph. It is not part of the group-travel team (M-03) and is not in the channel plan. The management's question: take it over, connect it or replace it? For that, someone has to be able to read it. This lesson places LangChain, LangGraph and Hugging Face Transformers and says openly what it does not do: this course teaches .NET. According to the pages of Learn, the exam expects Python knowledge and familiarity with the Agent Framework, MCP, RAG and LangGraph. Whoever wants to take it practices Python on their own in addition; a bridge course is not planned.
Prior knowledge. "AI-103, lesson 0.3" shows how to read Python. "AI track, lesson T7.4" shows workflows of executors and edges, "T7.5" memory and the session, "T7.3" approvals. Lessons 2.8 and 2.9 of this course cover orchestration and the human in the loop, lesson 2.2 fine-tuning. What begins here goes beyond that: reading other frameworks and mapping their terms to one another.
New in this lesson. The prototype ideenlabor (Python, LangGraph, an agent of the Foundry Agent Service as a node). It reads no customer data and writes nothing (assumption of the exercise). The lesson makes no decision on taking it over; it supplies the basis for one.
The requirement
| No. | Requirement | Type | Measure | Weight | Source |
|---|---|---|---|---|---|
| 2.12-01 | A short excerpt in LangGraph or Hugging Face Transformers is read and named. | F | all parts of the excerpt named | Must | Exam list |
| 2.12-02 | Every mapping between frameworks says whether Learn names it or whether it is a reading. | Q | 0 mappings without a mark | Must | Course rule |
| 2.12-03 | We write no Python for the case. | R | 0 Python files of our own | Must | Stack: .NET |
| 2.12-04 | Models from Hugging Face run only if Foundry deploys them by its criteria. | R | criteria of Learn met | Should | M-08, R-11 |