Add 122 Python example scripts + .env.example

Each of the 122 code slides now has a standalone .py script in
examples/slide-NNN/script.py with the LLM provider pattern from
bro-js/agents/teacher/assistant/src/angry_teacher/llm.py:
  - load_dotenv() + ChatOpenAI with base_url=https://llm.brojs.ru/v1
  - OPENAI_API_KEY from .env (with safe (or '' or None) pattern)
  - Same comment block with alternative model/base_url choices

Layout:
  - sec1/sec5: code extracted from slide-NN.js (addCodeBlock)
  - sec2/sec3/sec4: code extracted from sectionN.pptx via python-pptx
    (JetBrains Mono font shapes)

100 scripts contain runnable code, 22 are visual-only (dividers,
intro, recap) and just print the slide title.

Setup:
  cp .env.example .env       # fill OPENAI_API_KEY (or symlink to your
                              # teacher/assistant/.env)
  pip install -r requirements.txt
  python examples/slide-005/script.py

examples/slide-122/script.py

@@ -0,0 +1,55 @@
+"""
+Slide 122: TypeScript SDK и плюсы/минусы
+Section 5: Экосистема
+Source: slides/section5-ecosystem/slide-12.js
+
+Сгенерировано автоматически из slide-NN.js / sectionN.pptx.
+Паттерн провайдера скопирован из
+bro-js/agents/teacher/assistant/src/angry_teacher/llm.py.
+
+Запуск:
+  cd langchain-evolution-deck
+  cp .env.example .env   # заполни OPENAI_API_KEY
+  python examples/slide-122/script.py
+"""
+import os
+from dotenv import load_dotenv
+from langchain_openai import ChatOpenAI
+
+load_dotenv()
+
+llm = ChatOpenAI(
+    # model="openrouter/free",
+    # model="qwen/qwen3.5-35b-a3b",
+    model="openai/gpt-oss-20b",
+    # model="nvidia/nemotron-3-nano",
+    # model="qwen/qwen3.5-9b",
+    base_url="https://llm.brojs.ru/v1",
+    # base_url="https://api.minimax.io/v1",
+    # base_url="http://0.0.0.0:8090/v1",
+    # base_url="https://openrouter.ai/api/v1",
+    # api_key=os.getenv("MINIMAX_API_KEY"),
+    api_key=(os.getenv("OPENAI_API_KEY") or "").strip() or None,
+    temperature=0.7,
+    stream_usage=True,
+)
+
+
+# === Code from slide 122 ===
+import { Client, traceable } from "langsmith";
+
+const client = new Client();
+
+// @traceable decorator -- точно как в Python
+const retrieve = traceable(
+  async function retrieve(query: string) {
+    return vectorStore.similaritySearch(query, 4);
+  },
+  { name: "retrieve_docs", runType: "retriever" }
+);
+
+// run_on_dataset для eval -- тоже есть
+await client.runOnDataset(
+  "rag-qa-eval-v1", target,
+  { evaluators: [answerMatch] }
+);