talk-pipeline

Conference talk preparation can be orchestrated end-to-end by running staged tasks that transform raw materials into finalized revision sheets. This pipeline enforces checkpoints and file-naming conventions across extraction, research, scripting, and revision stages.

Stage isolation allows running a single stage like scripting or concept generation independently for fast iteration. You can also run the full pipeline sequentially or resume from any enforced human-in-the-loop checkpoint.

REX mode handles repository archaeology for code-backed presentations using existing source material. Concept mode supports idea-driven talks, adapting the extraction, research, and scripting stages to generate concepts from raw ideas.

You must provide context including slug, event, date, duration, and audience to start the talk preparation workflow. Mode selection between REX and Concept is also required to properly route the staged execution.

The pipeline enforces upstream file checks and naming conventions before generating artifacts. It produces a clear summary of outputs and finalized revision sheets for downstream workflows, ensuring traceability across all stages.

Speaker preparation is managed by orchestrating six staged tasks with human-in-the-loop checkpoints to prevent premature progression. This approach ensures consistent file naming, validation, and traceability across extraction, positioning, and revision.