Project Overview
StartupWheel orchestrates a sequence of agents that transform scraped market intel into production-ready artefacts. The pipeline automates research collection, product requirement drafting, design specs, Airtable schemas, and execution backlogs—letting teams spin up idea validation projects in minutes.
Key Features
run_pipeline.pyexecutes the end-to-end workflow: scrape → merge brief → generate PRD, design spec, schema, and task backlog.- Playwright-powered scraper logs into IdeaBrowser, captures curated opportunity feeds, and normalises text/JSON outputs for downstream agents.
- PM pipeline agent enforces strict JSON schemas for PRD, design, and tasks to guarantee structured hand-offs.
- Airtable schema generator converts PRD + design into table/field definitions ready for Meta API provisioning.
Technical Highlights
- Modular agent architecture with reusable helpers, each script callable independently or via orchestrator.
- Strong guardrails around OpenAI responses to ensure deterministic structure; failures trigger retries with tightened prompts.
- Artifacts stored under timestamped directories with zipped bundles for sharing.
- Command-line ergonomics (click/argparse) plus environment-based configuration for secrets and API keys.
Results
- Reduced manual research-to-deliverable workflow from multiple hours to an automated run of ~8 minutes.
- Generated consistent PRD/design/task outputs that teams imported directly into Notion and Jira.
- Airtable schema automation eliminated copy/paste errors and accelerated ops setup.
Next Steps
- Expand scraper coverage beyond IdeaBrowser to include industry-specific RSS feeds.
- Expand on the output from the agentic system to better fit the needs of producing the software.