fix(code_generator): emit unique, tutorial-specific output instead of identical boilerplate

[Copilot]

Both code_generator.py files produced the same static boilerplate regardless of which video was processed — extracted tutorial steps, keywords, and features were accepted as parameters but never used in the generated output.

src/youtube_extension/backend/code_generator.py

• _generate_vanilla_main_js — was ignoring both tutorial_steps and features entirely. Now generates a named function per step (executeStep1(), …) wired into runTutorialSteps(), plus optional helpers keyed to detected features (setupFormHandlers, fetchInitialData, setupDarkModeToggle, initSearch).
• _generate_fastapi_main — added extracted_info param; now inlines step comments, exposes /api/tutorial/steps listing the video's extracted steps, and creates one domain route per keyword (e.g. a FastAPI tutorial produces /api/fastapi). Removes the hardcoded generic /api/messages endpoints.
• _generate_python_api — passes extracted_info through to _generate_fastapi_main.

Example diff for a "Build a REST API with FastAPI" video vs. "GraphQL with Python":

# Before — both produced identical:
@app.get("/api/messages")
async def get_messages():
    return [{"message": "Hello from your UVAI generated API!", ...}]

# After — FastAPI tutorial produces:
# Step 1: Install FastAPI and uvicorn
# Step 2: Create a basic FastAPI app
@app.get("/api/tutorial/steps")
async def get_tutorial_steps(): ...

@app.get("/api/fastapi")
async def get_fastapi(): ...

src/agents/specialized/code_generator.py

• execute — propagates top-level video_title / tutorial_steps inputs into context before dispatch.
• _generate_api_endpoint — prepends # Endpoint generated for: <title> + step comments; tutorial steps are reflected in the processing-logic block.
• _generate_generic_api — now accepts intent and prepends a video/intent header, making each generated API unique to its source.

Original prompt

---

This section details on the original issue you should resolve

<issue_title>code_generator.py</issue_title>
<issue_description>fix code_generator.py so each video produces unique, tutorial-specific output instead of identical boilerplate. @claude </issue_description>

<agent_instructions>fix the file as stated
</agent_instructions>

Comments on the Issue (you are @copilot in this section)

Custom agent used: Mastering-Any-Task AI
A hybrid agentic framework for research-led planning and execution.

• Fixes code_generator.py #77

---

⚡ Quickly spin up Copilot coding agent tasks from anywhere on your macOS or Windows machine with Raycast.

[vercel]

The latest updates on your projects. Learn more about Vercel for GitHub.

Project	Deployment	Actions	Updated (UTC)
v0-uvai	Ready	Preview, Comment, Open in v0	Mar 20, 2026 10:19am

[vercel]

Additional Suggestions:

1. Three methods in code_generator.py were partially fixed; two still ignore important parameters for feature-specific code generation

2. User-derived content from video analysis (tutorial steps, technology names, features, and project titles) was directly inserted into generated JSX/HTML code without HTML escaping - FIXED

3. Title and URL parameters are directly inserted into markdown without escaping, allowing markdown special characters to be interpreted as formatting syntax instead of literal text

[groupthinking]

Closing: superseded by PR #113 which ported the UUID suffix fix + _build_title helper from YOUTUBE-EXTENSION PR #706. The root cause (timestamp % 10000 collision + generic fallback title) is fixed.