feat(plugins): add Lecture Notes plugin for academic conversations

[Iansabia]

Summary

• Adds a Lecture Notes webhook plugin (plugins/lecture-notes/) — a standalone FastAPI service that transforms recorded lectures into structured academic notes
• Adds a Study Companion prompt-based plugin to community-plugins.json — an interactive study buddy that generates flashcard-style Q&A pairs and quizzes users via chat
• Students are a core Omi demographic — this gives them both passive note extraction and active recall study tools

Relationship to Existing class-notes Plugin

The existing class-notes plugin is prompt-only with memories capability and a basic 4-section outline format. This PR adds two complementary tools that don't overlap:

	class-notes (existing)	lecture-notes (this PR)	study-companion (this PR)
Type	Prompt-based	Webhook FastAPI service	Prompt-based
Capabilities	memories	webhook integration	memories + chat
Output	Generic outline, facts, homework, other	Structured key concepts w/ definitions, topics, summary, questions, action items, study tips	Flashcard Q&A pairs, practice questions, concept connections
Academic detection	None (runs on everything)	Heuristic NLP pre-filter (keyword density, speaker asymmetry, duration)	None
Chat companion	No	No	Yes — quizzes users using active recall

class-notes = passive outline. lecture-notes = smart webhook with structured extraction. study-companion = interactive chat that tests you on the material.

How the Webhook Plugin Works

Two-stage pipeline that avoids wasting API tokens on casual conversations:

1. Heuristic pre-filter — Scores conversations using keyword density (60+ academic terms), speaker asymmetry (lectures have one dominant speaker), duration, and Omi's own category classification. Casual conversations are filtered out before any LLM call.
2. Structured LLM extraction — GPT-4o with with_structured_output extracts key concepts with definitions, topics covered, summary, questions to explore, action items (homework/readings), and study tips.

Files Changed

File	Description
plugins/lecture-notes/main.py	FastAPI service — models, heuristic scoring, LLM extraction, formatting
plugins/lecture-notes/requirements.txt	Dependencies (fastapi, langchain-openai, tiktoken)
plugins/lecture-notes/.env.template	Environment variable template
plugins/lecture-notes/Dockerfile	Docker deployment
plugins/lecture-notes/README.md	Setup, example output, test payload, deployment guide
community-plugins.json	Study Companion prompt-based plugin entry

Test Plan

• Syntax validation: python -c "import py_compile; py_compile.compile('plugins/lecture-notes/main.py')"
• Black formatting: black --check --line-length 120 --skip-string-normalization plugins/lecture-notes/main.py
• JSON validation: python -c "import json; json.load(open('community-plugins.json'))"
• 23 automated tests with FastAPI TestClient — endpoints, heuristic scoring, formatting, edge cases (discarded, empty, casual, extra Omi payload fields)
• Test payload included in README — curl command to POST a sample OS lecture transcript

Example Webhook Output

LECTURE NOTES | Operating Systems

Summary
Introduction to process scheduling, covering the key algorithms
used to manage CPU time allocation among competing processes.

Key Concepts
  - Process: A program in execution with its own address space and resources
  - Context Switch: Saving and restoring process state when switching CPU allocation
  - Round Robin: Scheduling algorithm assigning equal time slices in circular order

Topics Covered
  - Process lifecycle and states
  - CPU scheduling algorithms (FCFS, SJF, Round Robin, Priority)
  - Preemptive vs non-preemptive scheduling

Action Items
  - Read Chapter 5 of Silberschatz
  - Complete Problem Set 3 by Friday

Study Tips
  - Draw state diagrams for each scheduling algorithm with numeric examples
  - Practice Gantt chart problems for exam prep

[gemini-code-assist]

Code Review

This pull request introduces a new 'Lecture Notes' plugin, both as a standalone FastAPI service and a prompt-based version. The implementation is well-structured, with clear data models, a heuristic pre-filter to save costs, and structured LLM extraction. The accompanying documentation, Dockerfile, and configuration files are comprehensive.

My review focuses on improving the robustness and performance of the FastAPI service. I've identified a critical issue where malformed input could crash the server and a high-priority performance issue related to blocking I/O calls. Addressing these will make the plugin more resilient and scalable.