Name	Name	Last commit message	Last commit date
Latest commit History 50 Commits
config	config
public	public
scripts	scripts
src	src
types/generated	types/generated
.env.example	.env.example
.gitignore	.gitignore
Dockerfile	Dockerfile
README.md	README.md
favicon.png	favicon.png
package-lock.json	package-lock.json
package.json	package.json
tsconfig.json	tsconfig.json

Colaberry AI CMS

Strapi v5 headless CMS powering colaberry.ai. 6 content types. 2,500+ entries. Zero frontend coupling.
The content backbone for enterprise AI catalog, podcast transcripts, and LLM architecture deep dives.

Frontend Strapi v5 Node.js 20 PostgreSQL 2500+ entries GCP Cloud Run

What It Does

This is the headless CMS for the Colaberry AI platform. It manages all structured content — agents, skills, MCP servers, tools, podcasts, and LLM architectures — and serves them via REST API to the Next.js frontend.

The CMS handles:

Content authoring for 6 primary catalog types
Draft/publish workflow (Strapi v5 draft system)
Media management for cover images, audio files, and transcripts
Global navigation configuration consumed by the frontend header/footer
Distribution channel configuration for automated social posting
Taxonomy relations (tags, companies, departments, categories)

Architecture

graph TB
 subgraph CMS["Strapi v5 CMS"]
 ADMIN["Admin Panel — /admin"]
 REST["REST API — /api"]
 MEDIA["Media Library — /uploads"]
 DRAFT["Draft/Publish — per-entry lifecycle"]
 end
 subgraph ContentTypes["Content Types"]
 AGENTS["Agents — 160+"]
 SKILLS["Skills — 500+"]
 MCPS["MCP Servers — 1,500+"]
 TOOLS["Tools — 200+"]
 PODCASTS["Podcasts — 260+ episodes"]
 LLM["LLM Architectures — 52 models"]
 end
 subgraph Relations["Shared Relations"]
 TAGS["Tags — cross-type labels"]
 COMPANIES["Companies — vendor/provider"]
 DEPARTMENTS["Departments — org units"]
 CATEGORIES["Categories — taxonomy mapping"]
 end
 subgraph Consumers
 FRONTEND["Next.js Frontend — SSG/ISR"]
 SCRIPTS["Deep-Dive Scripts — .mjs authoring"]
 DIST["Distribution Cron — social posting"]
 PREVIEW["Distribution Preview — dry-run API"]
 end
 ADMIN --> ContentTypes
 REST --> ContentTypes
 ContentTypes --> Relations
 FRONTEND -->|fetchCMSJson + bearer auth| REST
 SCRIPTS -->|PUT /api + ?status=published| REST
 DIST -->|source + log| REST
 PREVIEW -->|read-only| REST

Data Flow

sequenceDiagram
 participant Author as Content Author
 participant Admin as Strapi Admin Panel
 participant DB as PostgreSQL
 participant API as Strapi REST API
 participant FE as Next.js Frontend
 Author->>Admin: Create/edit content entry
 Admin->>DB: Save draft
 Author->>Admin: Click "Publish"
 Admin->>DB: Set status = published
 Note over DB: Entry now visible via API
 FE->>API: GET /api/agents?filters[visibility]=public
 API->>DB: Query published + public entries
 DB-->>API: Results with relations populated
 API-->>FE: JSON response
 FE->>FE: ISR revalidate (600s) or on-demand

Content Types

Primary Catalog Types

Content Type	API Endpoint	Key Fields	Entries
Agent	`/api/agents`	name, slug, description, longDescription, keyBenefits, useCases, limitations, whatItDoes, outcomes, coreTasks, inputs, outputs, tools, executionModes, orchestration, securityCompliance, industry, status, visibility, rating, source, verified	160+
Skill	`/api/skills`	name, slug, category, skillType, description, provider, prerequisites, linkedMCP, visibility	500+
MCP Server	`/api/mcp-servers`	name, slug, description, serverType, primaryFunction, capabilities, authentication, hostingOptions, visibility	1,500+
Tool	`/api/tools`	name, slug, toolCategory, description, webLink, pricing, visibility	200+
Podcast Episode	`/api/podcast-episodes`	title, slug, publishedDate, transcript, transcriptSegments, audioUrl, buzzsproutId, podcastStatus, tags, companies	260+
LLM Architecture	`/api/llm-architectures`	name, slug, organization, parameters, activeParameters, contextWindow, vocabSize, numLayers, hiddenSize, releaseDate, decoderType, attention, keyFeatures, deepDive (Dynamic Zone)	52

Supporting Types

Content Type	Purpose
Tag	Cross-type labels (many-to-many)
Company	Vendor/provider references
Department	Organizational units with category nesting
Category	Taxonomy parent grouping
Global Navigation	CMS-driven header/footer link structure
Distribution Channel	Social platform config (name, platform, template, credentials ref)
Distribution Log	Per-dispatch audit trail (platform, entry ID, status, response)

Dynamic Zone: Deep Dive Blocks

LLM Architecture entries include a deepDive Dynamic Zone with these components:

Component	Fields	Purpose
`deep.heading`	level (h2/h3/h4), text, anchor	Section headings
`deep.paragraph`	body (rich text)	Prose content
`deep.callout`	variant (insight/warning/note), title, body	Highlighted blocks
`deep.code-block`	language, code, caption	Code examples
`deep.table`	caption, headers[], rows[][]	Data tables
`deep.list`	style (bullet/numbered), items[]	Lists
`deep.image`	url, alt, caption	Figures
`deep.references`	items[] (title, url, authors, year)	Citations

Quick Start

Prerequisites

Node.js 20+
PostgreSQL 16+ (or use Docker)

Local Development

git clone https://github.com/colaberry/colaberry-ai-cms.git
cd colaberry-ai-cms
npm install
cp .env.example .env # Configure DATABASE_URL and secrets
npm run develop

Open http://localhost:1337/admin to access the admin panel.

Docker

docker compose up

Environment Variables

Variable	Required	Description
`DATABASE_URL`	Yes	PostgreSQL connection string
`APP_KEYS`	Yes	Strapi app keys (comma-separated)
`API_TOKEN_SALT`	Yes	Salt for API token hashing
`ADMIN_JWT_SECRET`	Yes	JWT secret for admin panel
`JWT_SECRET`	Yes	JWT secret for API auth
`TRANSFER_TOKEN_SALT`	Yes	Salt for transfer tokens

Build

npm run build # Build admin panel
npm run start # Start production server

API Usage

Authentication

All API requests from the frontend use a bearer token:

curl -H "Authorization: Bearer <CMS_API_TOKEN>" \
 "http://localhost:1337/api/agents?filters[visibility]=public&populate=*"

Common Query Patterns

# List all published agents (public visibility)
GET /api/agents?filters[visibility]=public&populate=tags,companies,department,coverImage
# Single agent by slug
GET /api/agents?filters[slug]=crop-stress-detection&populate=*
# Paginated skills with category filter
GET /api/skills?filters[visibility]=public&filters[category]=AI&pagination[page]=1&pagination[pageSize]=25
# Podcast episodes with audio
GET /api/podcast-episodes?filters[podcastStatus]=published&populate=tags,companies&sort=publishedDate:desc
# LLM architecture with deep dive blocks
GET /api/llm-architectures?filters[slug]=deepseek-v3&populate[deepDive][populate]=*
# Catalog counts (lightweight)
GET /api/agents?filters[visibility]=public&pagination[pageSize]=1&pagination[withCount]=true
GET /api/skills?filters[visibility]=public&pagination[pageSize]=1&pagination[withCount]=true
GET /api/mcp-servers?filters[visibility]=public&pagination[pageSize]=1&pagination[withCount]=true
# All tags with usage counts
GET /api/tags?populate=agents,skills,mcp_servers&pagination[pageSize]=100

Deep Dive Authoring (Script -> CMS)

The author-llm-deep-dive.mjs script PUTs dynamic zone blocks directly:

# Publish a deep dive (appends ?status=published for Strapi v5 draft system)
node scripts/author-llm-deep-dive.mjs --slug deepseek-v3
# Dry run
node scripts/author-llm-deep-dive.mjs --slug deepseek-v3 --dry-run
# Publish all 52 flagships
node scripts/author-llm-deep-dive.mjs --all

Critical: The ?status=published query param is required. Without it, Strapi v5 defaults PUT requests to draft status and SSR never sees the content.

Distribution System

The CMS stores distribution channel configuration and dispatch logs.

Distribution Channel (CMS Collection)

Field	Type	Description
`name`	String	Channel display name (e.g. "X/Twitter Main")
`platform`	Enum	`x`, `moltbook`, `huggingface`, etc.
`template`	Text	Mustache-style post template
`credentialRef`	String	Env var name (CMS never holds secrets)
`active`	Boolean	Enable/disable channel

Distribution Log (CMS Collection)

Field	Type	Description
`platform`	String	Target platform
`entryId`	String	Source content entry ID
`entryType`	String	Content type (agent, skill, etc.)
`status`	Enum	`success`, `failed`, `skipped`, `dry-run`
`response`	JSON	Platform API response
`idempotencyKey`	String	`${platform}:${id}:${updatedAt}`

Seeding Channels

# Idempotent — re-running updates existing rows by name
node scripts/seed-distribution-channels.mjs

Templates live in scripts/distribution-templates/{x,moltbook,huggingface}.md.

Deployment

GCP Cloud Run

Service	Description
`colaberry-ai-cms-prod`	Production CMS instance

The CMS runs as a stateless container on Cloud Run with a managed PostgreSQL backend.

Important Notes

Media storage: Configure a cloud storage provider (GCS, S3) for production media uploads
API tokens: Generate via Strapi admin panel → Settings → API Tokens
Draft/Publish: All content types use Strapi v5's built-in draft/publish system. Only published entries are visible via the API with default filters
ISR sync: The frontend uses ISR with 600s revalidation. Content changes appear on the live site within 10 minutes of publishing

Project Structure

colaberry-ai-cms/
├── config/
│ ├── api.js # API configuration
│ ├── admin.js # Admin panel configuration
│ ├── database.js # Database connection (PostgreSQL)
│ ├── middlewares.js # CORS, security, body parser
│ ├── plugins.js # Plugin configuration
│ └── server.js # Server host/port settings
├── src/
│ ├── api/ # Content type definitions
│ │ ├── agent/ # Schema, controllers, routes, services
│ │ ├── skill/
│ │ ├── mcp-server/
│ │ ├── tool/
│ │ ├── podcast-episode/
│ │ ├── llm-architecture/
│ │ ├── tag/
│ │ ├── company/
│ │ ├── department/
│ │ ├── category/
│ │ ├── global-navigation/
│ │ ├── distribution-channel/
│ │ └── distribution-log/
│ ├── components/ # Reusable Dynamic Zone components
│ │ └── deep/ # Deep dive block types (heading, paragraph, etc.)
│ └── extensions/ # Strapi lifecycle hooks
├── public/ # Static assets
├── database/ # Migrations
└── types/ # Generated TypeScript types

Related Repositories

Repository	Purpose
colaberry/colaberry-ai	Next.js frontend (consumer of this CMS)
colaberry/WorldOfTaxonomy	Taxonomy classification platform

Colaberry AI CMS — Structured content for the enterprise AI catalog.

Folders and files

Latest commit

History

Repository files navigation

Colaberry AI CMS

What It Does

Architecture

Data Flow

Content Types

Primary Catalog Types

Supporting Types

Dynamic Zone: Deep Dive Blocks

Quick Start

Prerequisites

Local Development

Docker

Environment Variables

Build

API Usage

Authentication

Common Query Patterns

Deep Dive Authoring (Script -> CMS)

Distribution System

Distribution Channel (CMS Collection)

Distribution Log (CMS Collection)

Seeding Channels

Deployment

GCP Cloud Run

Important Notes

Project Structure

Related Repositories

About

Resources

Uh oh!

Stars

Watchers

Forks

Releases

Packages 0

Uh oh!

Contributors

Uh oh!

Languages

Packages