Descripción

KimaAI brings modern AI features to WordPress by connecting your site to AI providers (like OpenAI) and exposing them through an easy-to-configure chatbot and a growing set of AI tools.

With KimaAI you can:

Build a customizable chatbot with model/provider selection, content awareness, quick actions, and an optional attention popover.
Add a vector database search tool and multilingual embeddings to power smarter, language-aware answers.
Collect chatbot feedback to understand response quality and improve your on-site experience over time.
Use a setup wizard to get started quickly, plus a settings system designed to be maintainable and extensible.
Extend KimaAI via developer hooks/filters to register and control custom tools and tool execution.

KimaAI is built to stay flexible: it works for non-technical users who want a powerful chatbot, and it also provides hooks and internal APIs for developers who want to integrate AI deeper into their workflows.

Please make sure to carefully read the Disclaimer before using the plugin, check the KimaAI website for more information. Thank you!

Features

AI Chatbot: provider/model selection, content awareness, quick actions, and UI customization.
Setup Wizard: guided first-time configuration.
Vector Search: vector DB search tool + embeddings to improve answers.
Multilingual-ready: multilingual embeddings and translatable dynamic settings (Polylang/WPML).
Feedback: collect user feedback on chatbot responses.
Developer hooks: register custom tools and control/observe tool calls.
More features coming soon!

My Vision for AI

AI is incredible, no doubt about it. It’s changing everything—from the way we work to the way we create. But for me, AI isn’t just about churning out endless content or automating everything for the sake of it. I want to see AI being used thoughtfully—as a tool that actually helps people.
I dream of an AI-driven world where creators can focus on their ideas instead of struggling with words, where language barriers don’t hold anyone back, and where tedious work is handled for us so we can spend more time on what really matters—our passions, our families, and just enjoying life.

I’ll always push for AI to be used in a way that empowers rather than overwhelms. Hope you’re with me on this! 💡💙

Disclaimer

KimaAI is a plugin that helps you to connect your websites to AI services. You need your own API keys and must follow the rules set by the AI service you choose. For OpenAI, check their Terms of Service and Privacy Policy. It is also important to check your usage on the OpenAI website for accurate information. Please do so with other services as well.

The developer of KimaAI and related parties are not responsible for any issues or losses caused by using the plugin or AI-generated content. You should talk to a legal expert and follow the laws and regulations of your country. KimaAI does only store data on your own server, and it is your responsibility to keep it safe. KimaAI’s full disclaimer is here.

Capturas

KimaAI chatbot widget on the website – light and dark mode side by side.

Chatbot live on your site: visitor getting an instant answer after connecting your API key and completing the setup wizard. — KimaAI chatbot widget on the website – light and dark mode side by side.

Instalación

Upload the plugin files to the /wp-content/plugins/kimaai directory, or install the plugin through the WordPress plugins screen directly.
Activate the plugin through the ‘Plugins’ screen in WordPress.
Run the Setup Wizard — it guides you through connecting your AI provider and API key.
Place the chatbot on your site and start answering visitor questions.

FAQ

How do I get an API key?: You can get an API key from OpenAI by creating an account on their website. Check the OpenAI website for more information.

Reseñas

zedkima 17 de abril de 2025

I installed KimaAI mainly to test out the chatbot feature, and I was genuinely surprised by how well it works. Setting it up was super easy, and within minutes I had an AI chatbot live on my site, answering visitor questions smoothly and naturally.

Leer la 1 reseña

Colaboradores y desarrolladores

«KimaAI – AI Chatbot & Assistant for WordPress» es un software de código abierto. Las siguientes personas han colaborado con este plugin.

nabiabdi

Traduce «KimaAI – AI Chatbot & Assistant for WordPress» a tu idioma.

¿Interesado en el desarrollo?

Revisa el código , echa un vistazo al repositorio SVN o suscríbete al registro de desarrollo por RSS.

Registro de cambios

1.5.4 – 2026-07-14

Added

Added an opt-in «Delete All Data on Uninstall» setting that removes KimaAI tables, options, transients, and scheduled tasks when the plugin is uninstalled.
Added a chatbot evaluation CLI with JSONL test cases, multi-turn conversations, assertions, optional LLM judging, category accuracy, run logs, and non-zero failure exits.
Added terminology glossary, banned terms, current-date injection, and dynamic instruction placeholders for more accurate, on-brand chatbot answers.
Added optional conversation-aware query condensation that rewrites follow-up questions into standalone knowledge-base searches.
Added hybrid keyword and vector retrieval with Reciprocal Rank Fusion and built-in ElasticPress, SearchWP, Relevanssi, and WordPress search providers.
Added pinned answers with an admin management screen, alternative phrasings, similarity thresholds, lazy embedding, and match tracking.
Added deterministic output guardrails that redact banned terms from final chatbot answers and expose violations for observability.
Added per-post-type embedding content templates for titles, permalinks, metadata, taxonomies, and ACF fields.
Added admin-managed evaluation cases, scheduled runs, regression thresholds, email and webhook alerts, stored history, and CLI history reporting.
Added knowledge-gap reporting that clusters recurring or negatively rated questions into candidates for pinned answers and evaluation cases.
Added answer source citations and an optional groundedness check against retrieved knowledge-base content.
Added language-aware knowledge-base retrieval for Polylang, WPML, request languages, and site locales.
Added admin and model-controlled post-type scoping for knowledge-base retrieval.
Added Media Library controls for including or excluding supported documents from the knowledge base.
Added vector index health, freshness reporting, orphan reconciliation, and optional weekly maintenance.
Added configurable human handoff by email, webhook, or action hook after repeated negative feedback or ungrounded answers.
Added daily and per-model token usage tracking, monthly budget alerts, and a WP-CLI usage report.
Added embedding connection diagnostics, sync preflight checks, and deployment-scoped vector environments to prevent cross-environment data mixing.

Improved

Improved embedding cleanup so trashed, unpublished, deleted, or excluded content is purged from its original collection and namespace without deleting retryable local records after remote failures.
Improved queued embedding jobs to skip content that is no longer live and to reschedule documents when they are re-included in the knowledge base.
Improved tool-call round limits so exhausted conversations produce a final answer without executing tools beyond the configured budget.

Security

Sanitized configurable provider endpoint URLs before storage and added REST argument sanitization for chatbot history session identifiers.
Validated feedback targets as exact stored assistant-message IDs before saving feedback or triggering human handoff.
Restricted model-requested retrieval scopes to searchable public post types and enforced admin, language, and post-type restrictions across vector and keyword results.
Excluded password-protected and otherwise non-public content from hybrid keyword retrieval.
Isolated third-party search redirects so rejected keyword sub-queries fall back to vector results instead of terminating chatbot requests.

Fixed

Fixed knowledge-base vector limits and minimum-similarity settings not being honored, and rejected comparisons between incompatible embedding dimensions.
Fixed multilingual retrieval for full language slugs, WPML provider results, and simulated evaluation requests.
Fixed conversation-aware condensation not receiving the session identifier on non-streaming requests.
Fixed utility calls failing with models that do not support reasoning, verbosity, or temperature parameters.
Fixed final-message guardrails, citations, and verification notes being lost from stored history or bypassed by streaming responses.
Fixed streamed chatbot responses not being included in token usage tracking and prevented concurrent usage updates from losing counts.
Fixed request-scoped citation and groundedness evidence leaking between long-running requests or omitting later chunks from the same source.
Fixed pinned-answer and knowledge-gap embeddings changing the shared chatbot model, using provider-incompatible defaults, or retaining cache entries from a previous embedding configuration.
Fixed pinned-answer cold caches blocking frontend requests, persisting incomplete or empty embeddings, and exceeding the call-level warming budget.
Fixed scheduled evaluations exceeding background time limits, leaking their rate-limit bypass, or retaining an outdated schedule frequency.
Fixed evaluation conversations polluting knowledge-gap reports or triggering human handoff notifications.
Fixed human handoff triggers missing real response identifiers, sending incomplete transcripts, or suppressing retries after failed delivery.
Fixed monthly budget alerts being permanently suppressed after failed email delivery.
Fixed index health and freshness reports counting failed uploads, missing partial refresh failures, mishandling unchanged content, or comparing incompatible timestamps.
Fixed orphan reconciliation hiding failed remote deletions and improved scheduled and CLI failure reporting when vectors remain.
Fixed the knowledge-gap look-back window using UTC instead of the WordPress site timezone.
Fixed retired DALL-E integration-test parameters and restored disabled admin-assistant end-to-end coverage without changing production behavior.

Developer

Added JavaScript unit tests and CSS linting to CI.
Raised PHPStan analysis from level 0 to level 1 and added the required project constant stubs and build-artifact exclusions.
Added contributor documentation and enforced Node.js 20 or newer through package metadata.
Added extensive unit, integration, and end-to-end coverage for enterprise knowledge-base, evaluation, guardrail, retrieval, embedding, and admin workflows.
Added filters and actions for tool-round budgets, chatbot rate limits, final messages, streaming eligibility, retrieval providers and scopes, evaluation regressions, handoff delivery, and usage observability.

For the full changelog, visit:
https://kimaai.com/changelog/