Custom Lecture Generator

Generate a single lecture using direct API calls to Gemini, Claude, Perplexity, and VectorShift KB. This bypasses the VectorShift pipeline orchestration and uses the custom pipeline at lib/lecture_pipeline/.

Slash Command Usage

/generate-lecture-custom <outline_path> --target "Lecture 1 - Title"

Examples:

• /generate-lecture-custom outline.md --target "Lecture 1"
• /generate-lecture-custom "course materials/outline.md" --target "GLP-1 Agonists"
• /generate-lecture-custom outline.md --target "Lecture 3" --output ./my-outputs/

When to Use This Skill

Use this skill when:

• You want to generate a single lecture from an outline
• You're experiencing issues with VectorShift polling
• You want more control over the generation process
• You need detailed progress logging

Architecture

This skill uses the custom pipeline at lib/lecture_pipeline/ which orchestrates:

Required Environment Variables

export GOOGLE_API_KEY="..."        # Gemini API
export ANTHROPIC_API_KEY="..."     # Claude API
export PERPLEXITY_API_KEY="..."    # Perplexity API
export VECTORSHIFT_API_KEY="..."   # VectorShift KB
export UNSTRUCTURED_API_KEY="..."  # Unstructured.io (PDF parsing)

Execution Instructions

When this skill is invoked:

Step 1: Verify Environment

First, check that API keys are configured:

python -m lib.lecture_pipeline.cli check

If any keys are missing, inform the user which environment variables need to be set.

Step 2: Parse Arguments

Extract from the skill invocation:

• outline_path: Path to course outline markdown
• --target: Which lecture to generate (required)
• --output: Output directory (optional, default: ./outputs)
• --dossier: Path to research dossier (optional)

Step 3: Run the Generator

cd "/Users/anantvinjamoori/NGM Site NextJS V1/ngm-website-official"

python -m lib.lecture_pipeline.cli generate \
    --outline "{outline_path}" \
    --target "{target_lecture}" \
    --output "{output_path}"

Add --dossier "{dossier_path}" if a dossier was provided.

Step 4: Report Results

After completion, show the user:

• List of generated files
• Location of output directory
• Any errors that occurred

Output Files

For each lecture, the following files are saved:

Example Workflow

User: /generate-lecture-custom outline.md --target "Lecture 2 - Metabolic Health"

Claude:
1. Checks API keys with `python -m lib.lecture_pipeline.cli check`
2. Runs the custom pipeline
3. Reports generated files

Output:
  - lecture_2_-_metabolic_health_slides.json
  - lecture_2_-_metabolic_health_transcript.md
  - lecture_2_-_metabolic_health_blueprint.md
  - ...

Comparison to VectorShift Pipeline

Troubleshooting

Missing API keys:

export GOOGLE_API_KEY="your_key"
export PERPLEXITY_API_KEY="your_key"

Import errors:

pip install google-generativeai httpx tenacity click

Timeout on long lectures: The pipeline has built-in timeouts. For very long content, consider breaking into smaller lectures.