parad0x-growth-engine

Mission

Build a local-first autonomous growth engine inspired by the OpenClaw/Larry loop and tuned for Parad0x Labs: sovereign AI infrastructure, open-source, and developer-first. The system orchestrates research, angle selection, visual asset generation, caption drafting, approval-gated publishing, and performance-driven learning.

Constraints

• Never mention tokens or cryptocurrency language in marketing copy.
• Prioritize GitHub traction (stars/forks/clones), downloads, and developer adoption.
• Start local-first with explicit approval before automating publishing.
• Keep every subsystem replaceable via clear interfaces.
• Use globally portable paths; no machine-specific hardcoding.
• Document branch/commit/push workflows where relevant.
• Assume GitHub access is available for future automation.
• Apply a PLANS-driven execution model for longer efforts.

Architecture

The repo follows a clean monorepo layout with focused packages hosting domain models, research/strategy/playbooks, creative tooling, publishing pipelines, metrics/memory stores, storage adapters, and shared utilities. Worker and dashboard apps live in apps/*, each backed by tsconfig project references so pnpm run typecheck validates the whole graph. Docs sit in docs/, scripts in scripts/, and automation skills under skills/.

Core Loop

1. Research findings flow from packages/research into structured ResearchFinding objects.
2. ContentAngle and ContentBrief definitions from packages/strategy feed creative SlideDeckSpec and SlideAsset generation in packages/creative.
3. packages/publishing manages CaptionDraft, approvals, and PublishingJob execution, with PostPerformanceSnapshot tracking in packages/metrics.
4. Classifications (PerformanceClassification) determine winners/neutral/losers, stored as RunArtifactBundle artifacts and MemoryInsight learnings in packages/memory, while packages/storage handles persistence.
5. Everything routes through packages/shared helpers and explicit interfaces so each subsystem can be swapped.

Getting Started

1. Install dependencies: pnpm run bootstrap.
2. Type-check the workspace: pnpm run typecheck.
3. Run lint and tests: pnpm run lint and pnpm run test.
4. Start the worker in dev mode: pnpm run dev.
5. Explore research data with pnpm research:run, pnpm research:list, and pnpm research:show --id <finding-id>.
6. Read docs/RESEARCH-PROVIDERS.md once you are ready to wire real data sources safely.
7. Consult docs/ARCHITECTURE.md and docs/ROADMAP.md for deeper context and next steps.
8. Review the operational docs for day-to-day guidance: docs/LOCAL_SETUP.md, docs/END_TO_END_FLOW.md, docs/OPERATIONS.md, docs/SECURITY.md, and docs/RELEASE_AND_PACKAGING.md.
9. Follow the skill-led guardrails in skills/ when reasoning about architecture, strategy, compliance, execution, or testing.

Workflows

• Branches must use the codex/ prefix; commits should be focused and atomic.
• Follow the staged milestones in PLANS.md for longitudinal work. Update PLANS before starting multi-step initiatives.
• Push to the remote after each milestone for review and gating.
• Consult AGENTS.md for Codex-specific guardrails on architecture, branch policy, quality gates, and testing expectations.

Symbols

• ResearchFinding, ContentAngle, ContentBrief, SlideDeckSpec, SlideAsset, CaptionDraft, PublishingJob, PostPerformanceSnapshot, PerformanceClassification, RunArtifactBundle, MemoryInsight, ExperimentTemplate, RepoTarget, RepoMetricSnapshot, and ApprovalDecision all live in packages/domain for shared reference.