thearnavrustagi/marketmenow

1. Overall shape

MarketMeNow is a Python 3.12+ monorepo built around a central orchestration core plus platform-specific adapters. The system is designed to generate, transform, review, and publish marketing content across multiple channels from shared project config and reusable “capsules.”

2. Core boundary: src/marketmenow

The marketmenow package is the architectural center. It contains the domain models, workflow registry, pipeline/orchestrator logic, campaign manager/planner, scheduling, project management, prompt building, normalisation, and shared utilities like text sanitisation and embedding-based selection.

3. Ports-and-adapters contract

The core talks to the outside world through explicit ports in src/marketmenow/ports for renderer, uploader, platform adapter, and analytics concerns. This is the key boundary to show in a diagram: core logic never depends directly on platform APIs, browsers, or SDKs.

4. Platform adapters

Each platform lives under src/adapters/<platform> and owns its own CLI, settings, renderer, uploader, and any browser/API client code. Supported integration surfaces include Instagram, Twitter/X, Reddit, LinkedIn, YouTube, TikTok, Facebook, and Email, with each adapter encapsulating the platform-specific publishing mechanics.

5. Content generation flow

The typical flow is: load active project config and prompts, build content through a workflow, render it into platform-specific assets/text, sanitise the final text, and upload/publish via the adapter. This pipeline appears repeatedly across reels, posts, threads, replies, comments, outreach messages, and email sends.

6. Project configuration layer

Brand and strategy are not hardcoded; they live under projects/<slug>/ and global prompts/ plus platform-specific prompt YAMLs and templates. Each project carries product description, personas, targets, campaigns, feedback assets, and generation configs, so workflows resolve content from the active project first and then fall back to global defaults.

7. Workflow system

src/marketmenow/workflows contains the named end-to-end jobs such as Instagram reel, Instagram carousel, Twitter thread/engage/outreach, Reddit launch/engage/outreach, LinkedIn post, YouTube short, TikTok reel, email outreach, and capsule repurposing/posting. The workflow registry and step registry provide the composition layer that maps these jobs to reusable pipeline steps.

8. Reusable “content capsule” model

A major architectural concept is the capsule: generated content is packaged so it can be reposted or repurposed without regeneration. The capsule-related steps and workflows support cross-platform reuse, such as turning a reel into a YouTube short or a post/thread into another format.

9. Instagram-specific subsystem

Instagram is the most elaborate adapter. It includes carousel generation, grading, reel scripting/TTS, Figma export, and a substantial Remotion-based video rendering pipeline under src/adapters/instagram/reels/remotion. That subtree is a separate Node/TypeScript video composition app embedded inside the Python repo, used to render scenes, transitions, and assets for reels.

10. Twitter/X subsystem

Twitter has both content publishing and outreach automation. In addition to thread/reply generation and posting, there is a dedicated outreach area for profile scraping, vector-style discovery of prospects, DM sending, and performance tracking, which makes Twitter a richer “engage” pipeline than a simple post publisher.

11. Feedback and learning loop

The feedback subsystem under src/marketmenow/core/feedback stores and classifies feedback, generates guidelines, and tracks sentiment. Instagram has its own grading service as well. These pieces feed the system’s in-context-learning and example-selection behavior so future generations can adapt to prior performance.

12. Campaigns and scheduling

The repo supports campaign-level planning and daemonized execution through campaign_* core modules and YAML campaign definitions in projects/*/campaigns and top-level campaigns/. This is the layer that turns one-off content generation into scheduled, multi-step marketing runs.

13. Web dashboard

src/web is a separate FastAPI-style web application layer with routes, templates, static assets, queue worker, credential handling, and progress/events support. It is the human control plane for reviewing, generating, managing flows, monitoring queues, and submitting work to the same core services the CLI uses.

14. External dependencies and infrastructure

The repo integrates with LLM providers via src/marketmenow/integrations/providers (OpenAI, Anthropic, Gemini) and uses browser automation, SMTP, platform APIs, Google/YouTube auth, and Figma export where needed. Operationally it expects uv, Docker/PostgreSQL for persistence, Playwright/Chromium for browser-driven flows, and Node.js for Remotion-based reel rendering.

15. What to draw as the main diagram

The clean diagram boundary is: CLI/Web/Queue Worker at the top, then the marketmenow core orchestrator and workflow registry in the middle, then platform adapters and external services at the edge. Around that, show project YAML/prompts/templates as a configuration/data source, and feedback/analytics as the closed loop that informs future generations.