Enterprise AI Newsroom, Editorial Co-Pilot & Multimedia Publishing Pipeline for WordPress.
PressHub AI Co-Pilot transforms WordPress into an autonomous, AI-driven newsroom and multimedia production suite. Spanning from multi-source web/RSS news gathering to editorial quality scoring, bilingual daily briefings, multi-host podcast script generation, and broadcast-quality text-to-speech synthesis, PressHub AI provides news organizations and content creators with an enterprise-grade automated editorial workflow.
Built with a modular Provider Store (supporting OpenAI, Anthropic, Google Gemini, and Google Cloud Vertex AI/TTS), 4-layer enterprise observability, and strict architectural principles (Settings-First, Zero-Silent-Fallback, Clean-Slate), PressHub AI ensures total transparency, deterministic control, and reliability.
┌────────────────────────────────────────────────────────────────────────┐
│ PressHub AI Newsroom │
└──────────────────────────────────┬─────────────────────────────────────┘
│
┌────────────────────────────┼────────────────────────────┐
▼ ▼ ▼
┌──────────────────┐ ┌──────────────────┐ ┌──────────────────┐
│ Daily Briefing │ │ Podcast Producer │ │ In-Editor AI │
│ Pipeline │ │ & TTS Studio │ │ Co-Pilot │
├──────────────────┤ ├──────────────────┤ ├──────────────────┤
│ • RSS & Web Scrape│ │ • Multi-Host Dialogue│ │ • Draft Assistant│
│ • LLM Curation │ │ • Nameless Hosts │ │ • 0-100 Scorecard│
│ • Token Budgets │ │ • Spoken Dates │ │ • Chat Intents │
│ • Multilingual │ │ • 30-Voice TTS │ │ • Imagen Sideload│
│ • Stage Triggers │ │ • Audio Stitching│ │ • Deep Research │
└────────┬─────────┘ └────────┬─────────┘ └────────┬─────────┘
│ │ │
└────────────────────────┼────────────────────────┘
▼
┌────────────────────────────────────────────────────────────────────────┐
│ Decoupled Provider Store & Engine │
│ • OpenAI • Anthropic Claude • Google Gemini • Google Cloud │
├────────────────────────────────────────────────────────────────────────┤
│ 4-Layer Observability & CLI Diagnostics │
│ Layer 1: Structured │ Layer 2: Token Log │ Layer 3: Audit │ L4: Core │
└────────────────────────────────────────────────────────────────────────┘
- Autonomous News Harvester (
PressHub_AI_News_Harvester): Concurrently scrapes and processes articles from configured RSS feeds and direct web URLs. Features domain politeness delays, per-domain concurrency limits, total execution time budgets, lead-article extraction, and HTML sanitization. - Intelligent News Curator (
PressHub_AI_News_Curator): Evaluates harvested candidate stories using LLM analysis. Employs semantic article deduplication to prevent repetitive coverage, enforces strict token context limits, and produces comprehensive bilingual (Greek and English) daily briefings. - Interactive Briefing Admin (
PressHub_AI_Briefing_Admin): A dedicated WordPress administrative control center providing visibility into each pipeline stage (Harvest → Curate → Audio/Podcast → Publish), stage event timelines, manual triggers, single-stage regeneration, and full timezone-aware scheduling.
- Multi-Host Script Engine (
PressHub_AI_Podcast_Producer): Automatically transforms curated news briefings into dynamic, conversational podcast scripts between multiple hosts:- Nameless Speaker Invariant: Presenters never awkwardly refer to each other or themselves by name (eliminating robotic "Thanks, John" conversational tropes); dialogue progresses naturally through ideas.
- Spoken Date Formatting: All dates are converted to natural spoken-language phrasing (e.g. "3rd of May 2026" or "3 Μαΐου 2026") rather than raw numerical digits or slashed formats that confuse TTS models.
- Presenter Persona Alignment: Distinct host roles (anchor, analyst, commentator) maintain consistent editorial voices and perspectives.
- Broadcast Audio Synthesizer (
PressHub_AI_Audio_Synthesizer): Synthesizes broadcast-quality audio via Google Cloud Text-to-Speech:- Supports modern speech engines including Journey, Chirp, Neural2, Studio, Standard, and Wavenet.
- 30-Voice Catalog across multilingual accents with granular speaking rate and pitch adjustment.
- Audio Delimiter Handling & Stitching: Sound effect tags (
[SFX: ...]) and topic transition markers are isolated and processed, automatically stitching synthesized host segments into unified broadcast audio files.
- AI Draft Generation: Accelerates content creation by synthesizing comprehensive drafts from raw source URLs, interview transcripts, or bullet notes.
- 0–100 Editorial Scorecard Gate: Evaluates drafts against an editorial rubric (readiness score, depth, factual grounding, tone). High-scoring drafts (score ≥ 80) can automatically transition to
pending review, while published posts are safeguarded against accidental demotion. - Sidebar Chat Intent Router: Automatically classifies author prompts into specialized handlers:
chat(editorial conversation),research(deep web synthesis),image(visual prompt generation), orreport(audio script generation). - Google Imagen Generation & Media Sideloading: Creates editorial imagery via Google Imagen on Vertex AI, automatically sideloading assets into the WordPress Media Library and setting them as featured images.
- Scheduled Deep-Research Jobs: Runs long-horizon research queries asynchronously in the background as
presshub_researchcustom posts, protected by concurrency locking and automated retention sweeps.
- Multi-Tier Presets: Admin-curated global defaults, author-specific profiles, author defaults, and per-request metabox dropdown selection. Presets append specialized instructions to base system prompts without overwriting foundational editorial safety constraints.
- Externalized Prompt Assets: Prompts are stored as modular text files under
presshub-ai-editor/assets/prompts/(e.g.default_greek_briefing.txt,podcast_style_two_hosts_greek.txt) and dynamically managed viaPressHub_AI_Prompt_Loader. - Admin Prompt Studio: Inspect, customize, and reset prompt templates directly within WordPress settings without editing source code.
PressHub AI adheres strictly to enterprise software engineering standards:
No operational limit, threshold, cap, time budget, batch size, retry count, or rate limit that affects user-visible behavior may be hardcoded inside business logic.
Every parameter is registered in the WordPress Options API with strict sanitization bounds, exposed via PressHub_AI_Settings_Storage, and rendered in the admin settings UI.
No business logic, API client, prompt resolver, or pipeline stage may silently fall back to a hardcoded model, prompt, or operational parameter when configuration is missing or failing.
Missing keys, unconfigured models, or failed API calls halt immediately with explicit WP_Error objects or descriptive exceptions, logging full diagnostic details to structured logs. Hidden code fallbacks and silent multi-model retry loops are prohibited.
The Provider Store is the sole source of truth. No legacy option shadowing or backward-compatibility baggage.
All provider parameters and default models resolve through PressHub_AI_Provider_Store (wp_presshub_ai_providers). Deprecated options are actively purged via idempotent database migrations rather than lingering in resolution cascades.
PressHub AI implements a 4-layer logging architecture to provide total transparency into every automated decision, token cost, and administrative action:
| Layer | Purpose | Target / Storage | Inspection Tool |
|---|---|---|---|
| Layer 1: Structured Logs | Subsystem operations, harvesting steps, TTS synthesis, HTTP events | wp-content/presshub-debug.log |
php dev-env/scripts/tail-logs.php --source=presshub |
| Layer 2: Token Telemetry | Model IDs, prompt/completion tokens, latency, cost, and JSON metadata | wp_presshub_ai_token_logs |
php dev-env/scripts/view-token-logs.php |
| Layer 3: Security & Audit | Settings updates, provider credential changes, and admin actions | wp_presshub_ai_audit_log |
php dev-env/scripts/query-db.php "SELECT * FROM wp_presshub_ai_audit_log" |
| Layer 4: WordPress Core | PHP runtime warnings, fatal tracebacks, cURL/environment issues | wp-content/debug.log |
php dev-env/scripts/tail-logs.php --source=wp |
# Stream all PressHub logs in real time
php dev-env/scripts/tail-logs.php --follow
# Filter logs by error level
php dev-env/scripts/tail-logs.php --level=ERROR
# Inspect aggregate LLM token usage, latencies, and provider breakdowns
php dev-env/scripts/view-token-logs.php --stats
# View recent token logs with detailed JSON request/response metadata
php dev-env/scripts/view-token-logs.php --detail=1
# Query the SQLite database directly
php dev-env/scripts/query-db.php "SELECT provider, model, total_tokens, duration_ms FROM wp_presshub_ai_token_logs ORDER BY id DESC LIMIT 5"Navigate to PressHub AI → Settings to configure the platform across responsive, per-tab AJAX-saving sections:
- Providers: Register API keys, organization IDs, and default models for OpenAI, Anthropic, Google Gemini, and Google Cloud (Vertex AI/TTS). Masked secrets prevent accidental credential exposure.
- News Harvester: Configure RSS source URLs, fetch timeout budgets, domain polite delays, concurrency limits, and lead article filtering.
- News Curation: Set curation LLM models, curation temperature, maximum article candidate pools, character caps, and language presets.
- Podcast & Audio Studio: Configure default podcast presenter personas, audio synthesis voices (male/female, accents), speaking rate, pitch modulation, and audio segment stitching.
- Prompt Studio: Customize system prompts and editorial guidelines for drafts, research, daily briefings, and podcast scripts.
- Rate Limits & Security: Enforce per-user request rate limiting windows, IP safeguards, and automated research-log retention sweeps.
- Logs & Maintenance: Configure log retention days, log verbosity (
DEBUG,INFO,WARNING,ERROR), prompt debug logging, and token usage analytics.
The repository features a self-contained local WordPress development environment powered by the official WordPress Core SQLite engine (zero external MySQL/MariaDB or Docker requirements).
presshub/
├── presshub-ai-editor/ # Core WordPress plugin (PHP)
│ ├── includes/ # Business logic, modules, Provider Store
│ ├── assets/ # Sidebar UI, admin JS/CSS, prompt assets
│ └── tests/ # 73 isolated PHP unit test suites
├── presshub-workflow/ # TypeScript workflow & MCP server
│ ├── src/ # MCP scorecard server & sync utilities
│ └── tests/ # 6 Jest test suites (81 tests)
├── dev-env/ # Local SQLite WordPress runtime & CLI tools
│ ├── scripts/ # tail-logs, view-token-logs, query-db, setup
│ └── wordpress/ # Local WordPress installation
└── docs/ # Specifications, plans, review records
# Start local WordPress instance at http://127.0.0.1:8888
php dev-env/scripts/server.php- Front-end:
http://127.0.0.1:8888 - WP Admin:
http://127.0.0.1:8888/wp-admin/(User:admin, Password:password123) - Live Code Sync: The plugin directory is dynamically linked via a filesystem junction (
mklink /J) directly topresshub-ai-editor/.
The PHP test suite runs in isolation without external database dependencies:
cd presshub-ai-editor
php tests/run-all-tests.phpExecutes all 73 test suites with 100% pass rate.
The workflow package integrates Model Context Protocol (MCP) servers and editorial automation:
cd presshub-workflow
npm install
npm test
npx tsc --noEmitExecutes all 6 Jest test suites (81 tests) with strict TypeScript type-checking.
# Via Windows batch wrapper
dev-env\bin\wp plugin list
dev-env\bin\wp eval "echo get_option('presshub_ai_log_level');"
# Or via PHP
php dev-env/bin/wp-cli.phar --path=dev-env/wordpress <command>- WordPress:
≥ 6.0(tested up to6.7) - PHP:
≥ 7.4 - API Keys: At least one active provider key (OpenAI, Anthropic Claude, or Google Gemini / Google Cloud Vertex AI & TTS).
- Download the latest release zip (
presshub-ai-editor-x.x.x.zip) from the GitHub Releases tab. - In your WordPress admin dashboard, navigate to Plugins → Add New → Upload Plugin.
- Choose the downloaded zip file and click Install Now.
- Activate PressHub AI Co-Pilot.
- Navigate to PressHub AI → Settings to configure your provider API keys.
The plugin embeds an integrated GitHub update checker. Releases published on the main branch with attached zip assets are automatically detected in WordPress. For private repositories, configure the optional GitHub Personal Access Token under PressHub AI → Settings → Providers.
For production sites, high-reliability background research and automated daily briefing harvesting should be bound to a system cron daemon rather than relying on web visitor traffic:
# In wp-config.php:
define( 'DISABLE_WP_CRON', true );
# In crontab (running every minute):
* * * * * wp cron event run --due-now --path=/var/www/html >/dev/null 2>&1This plugin is licensed under the GPL-2.0-or-later license. See the license declaration in presshub-ai-editor/presshub-ai-editor.php for complete terms.