WP Manifestindependent plugin directory
manifest / unclassified / zeko-ai

Zeko AI

Zeko zeko-ai - WordPress plugin/theme (GPL-2.0-or-later)

by Zeko Team · github.com/zeko-io/zeko-ai · website

★ 0stars
0forks

Install

No release zip yet. The repository archive installs, but the folder name will carry the branch suffix and updates will not flow:

wp plugin install https://github.com/zeko-io/zeko-ai/archive/refs/heads/main.zip

AI layer for the Zeko ecosystem: assistant/chatbot, unified search, recommendations, content moderation, usage analytics, and a personal profile hub. Provider-agnostic.

Description

Zeko AI brings an AI layer to the Zeko ecosystem. It provides an assistant/chatbot with memory, a unified search across every Zeko module, personalized recommendations, content moderation, and usage analytics — all through a provider-agnostic client that supports Mock, OpenAI, Anthropic, OpenRouter, Gemini, Groq, and DeepSeek.

Features

  • Assistant / Chatbot — Context-aware conversation with the community agent and optional personal memory and recap
  • Personal RAG — Retrieval over the user's own activity, courses, applications, and knowledge base
  • Unified Search — One search endpoint that discovers jobs, QA, courses, freelance projects, mentors, love profiles, rewards, and shop products (zeko_ai_search_sources)
  • Recommendations — Personalized content suggestions per module (zeko_ai_recommendations)
  • Content Moderation — Automated moderation for QA answers, love photo reviews, and shop/product content with configurable thresholds
  • Profile Hub — [zeko_ai_profile] renders a unified profile with checklists, activity, rewards, and module cards (zeko_ai_profile_modules)
  • Usage Analytics — Per-user, per-provider usage and estimated cost tracking
  • Radar Digest — Scheduled weekly digest of relevant ecosystem activity for each user
  • Admin Tools — Provider health, teach/gaps, memory browser, and provider log
  • Provider Agnostic — Mock / OpenAI / Anthropic / OpenRouter / Gemini / Groq / DeepSeek with keyless mock fallback

Shortcodes

  • [zeko_ai_assistant] — Chat with the community agent
  • [zeko_ai_search] — Unified ecosystem search
  • [zeko_ai_recommendations] — Personalized recommendations
  • [zeko_ai_writer] — AI content generation
  • [zeko_ai_profile] — Unified user profile hub
  • [zeko_public_profile] — Public profile view

Installation

  1. Upload the zeko-ai folder to /wp-content/plugins/
  2. Activate the plugin through the 'Plugins' menu in WordPress
  3. Add one or more provider API keys under Zeko AI > Settings (the plugin runs in Mock mode without keys)
  4. Create pages and add the shortcodes above

External Services

Zeko AI is provider-agnostic. No external service is contacted by default: with no API keys configured the plugin runs entirely in a local Mock (keyless) mode. When you add a provider key, data is sent only when a feature you use actually calls that provider.

  • AI model providers (OpenAI, Anthropic, OpenRouter, Gemini, Groq, DeepSeek) — What is sent: your provider API key and the content of the request (assistant/chatbot prompts, your conversation history, and — where you have enabled them — your personal memory/recap context and personal-RAG context from your own activity). What is received: generated text output. When: when a user interacts with the assistant, search, writer, or moderation feature and a provider key is set for the configured provider. Why: to generate responses. Retention: local conversation and memory storage is described in the "Does the assistant remember the conversation?" FAQ; provider-side handling of the content you send is governed by each provider's privacy policy. The configured provider is always selectable per-site (Zeko AI > Settings).
  • Web search providers (keyless DuckDuckGo, or Google/Bing/Brave/Serper with a key) — What is sent: your search query only, when the administrator enables web search and a user request triggers it. When: a "from the web" section is added to answers/site search when turned on. Retention: search results are cached briefly and never stored permanently; provider policies govern the request logs.
  • Wikipedia world knowledge (opt-in) — When the agent's "world knowledge" setting is enabled, an extracted page-title guess from the question is sent to https://en.wikipedia.org/api/rest_v1/page/summary/{title} (and, on a miss, en.wikipedia.org/w/api.php opensearch) to ground factual answers. What is sent: the title string only, no personal data. When: only on enabled, factual agent queries. Retention: cached in a transient for 5 minutes; Wikipedia's own request logs apply.
  • Zeko Cloud LLM (hosted, license-gated) — If you choose the built-in "cloud" provider, prompts are posted to Zeko's license server (https://ozconsultz.com/ by default, overridable via ZEKO_LICENSE_SERVER) together with your license key, site URL, and per-site token, and billed against your license credits. What is sent: the prompt/conversation content, license key, home URL, token, and model. When: only when the cloud provider is selected and a key is configured. Retention: governed by your license agreement with the Zeko license server operator.

All provider endpoints are HTTPS and filterable (zeko_ai_*_endpoint) so the request destination is always visible and auditable. Endpoint/payload detail with file:line citations is in REMOTE-SERVICES.txt in this plugin folder.

Third-party resources

Zeko AI bundles no third-party runtime assets — all CSS, JS, icons (inline SVG), and images (the screenshot only) are original Zeko work under GPL-2.0-or-later. Its interactions with external services are outbound API calls only, documented in REMOTE-SERVICES.txt.

Frequently Asked Questions

Do I need an AI provider API key?

No. Without keys the plugin falls back to a Mock provider so every flow can be tested. Add an OpenAI, Anthropic, or other provider key to enable real generation.

Which modules does search cover?

Search is extensible via zeko_ai_search_sources. Out of the box it discovers jobs, QA, learn, freelance, mentor, love, rewards, and shop content.

Does the assistant remember the conversation?

Yes. Each conversation is persisted, and users can opt into personal memory so future sessions recall prior context.

Dependencies

Zeko AI runs standalone (Mock provider needs no plugins and no keys). Zeko Core (when active) provides shared helpers and page lookups; content moderation and search read module content only from the modules that are installed.

Privacy

Data stored includes conversation transcripts, assistant memory (opt-in), personal-RAG context drawn from the user's own activity, search results cache, usage analytics (per-user, per-provider), and moderation records. AI requests send the content and any enabled personal memory/context to the configured provider — see External Services. Moderation thresholds are configurable in Zeko AI > Settings.

Uninstall

Deleting the plugin drops all Zeko AI tables, clears its crons (maintenance, radar digest), removes the pages it created, and deletes its options.

Troubleshooting

  • No answer? Without an API key the Mock provider runs; add a provider key under Zeko AI > Settings for real generation.
  • Provider errors? Check the provider log and health screen in the admin.
  • Search misses content? Search uses zeko_ai_search_sources; confirm the target module is active and indexed.

Screenshots

  1. AI dashboard - cloud composer with personal memory sources

Changelog

0.5.1

  • Community agent keeps full command of every module: the knowledge paraphrase recall now always runs a strict token-AND net (>= 2 meaningful words), so a stale/missing InnoDB FULLTEXT index can no longer silently wipe "learned answer" retrieval
  • Corpus paraphrase ranking matches MySQL's accent-insensitive behaviour (Zeko_AI_Search::fold folds Latin diacritics: "cafe" surfaces a business named "Café"), and an item whose title carries at least half of a query's distinctive terms (e.g. "Zeko Central Café & Bistro" for "find me a cafe or restaurant") beats loose rows that only mention the words
  • Business Directory agent intent broadened (barber, plumber, hotel, clinic, dentist, mechanic, spa, bakery, near me, opening hours, booking/appointment, …), with the business fallback message listing concrete service categories
  • Web-search answers now render clickable [url](url) source links in the chat UI and expose the raw results via raw['web_results']
  • Community-agent seed knowledge grown to 168 FAQ rows (version-gated reseed): business listing management, business-account on-ramp, services and bookings, shop cancellations, job re-open/delete, quiz retakes, freelance cancellations/disputes, mentoring, dating privacy, wallets/payouts, rewards tiers, scam protection and more
  • Direct "how do I create a business account" answer added to the seeds; business intent fallback now also links the real Add-your-business form (/add-business/)
  • Intent detection broadened for most modules (salary/applications, tutorials/study, forums, pharmacies/bistros/open-now, portfolios/invoices, career advice, deposits/refunds/payouts, achievements/cashback, flagging) and made fully overridable via the zeko_ai_agent_intents filter
  • Auto-learning hardened: bare "good"/"nice"/"help" no longer counts as affirmation, greetings are never learned, one-word questions are skipped, and logged-in users are capped at a few learned rows per day (zeko_ai_agent_auto_learn_daily_cap) so junk never floods the knowledge base
  • Tests: 265 tests / 728 assertions — all green

0.5.0

  • Profile hub, personal RAG, recap memory, provider health, memory browser, radar digest
  • Provider-agnostic client (Mock / OpenAI / Anthropic / OpenRouter / Gemini / Groq / DeepSeek)
  • Unified search + recommendations across all modules
  • Content moderation with configurable thresholds
  • Usage analytics with estimated cost

Upgrade Notice

0.5.0

Initial release of Zeko AI.