Gorka Molero

Case · story studio

Saga

An agentic story and video creation workflow that combines structured generation, product UI, writing-room agents, script outputs, and media-pipeline direction.

Public on gorkamolero Agentic story studio Two-repo product line

Problem

AI story tools often stop at one-shot generation. A useful creative studio needs a repeatable workflow: brief, research, outline, draft, critique, final script, saved ideas, product state, and a path toward media output.

What I Built

Saga combines two public repos under gorkamolero. saga-ai is the canonical product direction for an agentic video-story creation studio with workflow, AI functions, media pipeline, and product framing. saga-agentic-ai is the earlier CrewAI writing-room prototype: brief, research, outline, draft, critique, final script, and archive flow.

Agent workflow Writing-room roles, structured generation, critique, revision, and final script output.
Product surface Chat workflow, saved scripts or ideas, product UI, and stateful creative work.
Media direction Video-story product framing, media pipeline, and Remotion/editor direction.

System Architecture

Saga story studio diagram with idea intake, writing-room agents, critique pass, product UI, media pipeline, creative output, public repos, and evidence needs.
Public-safe story-studio diagram. It shows the agentic workflow while keeping the screenshot and mock-media cleanup gaps visible.
Saga proof diagram showing the saga-ai product repo, saga-agentic-ai prototype repo, output shape, verification status, and review boundary.
Proof artifact from the local repos. It records the product architecture, the CrewAI prototype outputs, and the current build boundary without overclaiming a clean runtime capture.
Source-grounded Saga mock workflow screen showing chat, saved idea, saved script, voiceover, visual asset map, and editor direction.
Source-grounded workflow screen. It is a public-safe mock based on the app's chat, idea, script, voiceover, visual-asset, and editor components; real runtime screenshots still need env/auth/database cleanup.
Saga local public-safe runtime demo route showing brief intake, research, outline, draft, generated structure, critique pass, and proof boundary.
Runtime capture from /portfolio-demo. It is a deterministic local route with fake story data, zero provider calls, no auth, no database rows, and no private customer material; provider-backed screenshots remain a separate disposable-environment task.
Saga agent prototype output proof showing the writing-room flow from brief to research, outline, generated script, critique loop, and archived markdown artifacts.
Prototype-output proof from saga-agentic-ai. It shows that the earlier CrewAI writing room produced dated, inspectable script artifacts and a critique loop before the fuller product UI, while avoiding provider keys, account data, or private runtime state.

Technical Decisions

  • Represent creative work as a staged workflow instead of a single prompt response.
  • Use the CrewAI prototype as a precursor to the productized Saga application.
  • Make agent roles legible so the output can be inspected, revised, and defended.
  • Tie story generation to product UI and media pipeline direction instead of leaving it as raw text.

What This Proves

Saga shows agentic product thinking in a creative domain: structured multi-step generation, user-facing workflow, reusable story artifacts, and the judgment to turn an agent prototype into a product direction.