Feature

Train Fiko AI with your real business knowledge

FAQs, prompts, products, website pages, documents — all become searchable knowledge for the AI. Powered by RAG (Retrieval-Augmented Generation) for accurate, grounded answers.

Start FreeWhat is RAG?
Knowledge sources

5 ways to teach the AI

FAQs / Q&A pairs

Direct question-answer pairs. The fastest way to get common answers right.

Manual business prompt

Free-form text describing tone, policies, services, and rules. Chunked for retrieval.

Products & services

Catalog entries with prices, descriptions, features. Manual or WooCommerce sync.

Website crawl

Crawl your site, clean HTML, generate Q&A pairs automatically with AI.

Documents (PDF, DOCX, TXT)

Upload policy docs, manuals, terms. Extracted and chunked with overlap.

Like/Dislike corrections

Your corrections to AI replies become training examples for next time.

How RAG works

From customer question to grounded answer

1

Customer asks a question

"What are your prices for hair coloring?"
2

Question gets embedded

The question is converted into a vector (a list of numbers representing meaning).
3

Similarity search

Fiko searches all your knowledge chunks for the most similar vectors. Top 3–10 chunks are selected.
4

Context assembly

The selected chunks become the AI's context: your actual price list, service descriptions, and FAQs.
5

Answer generation

The AI generates an answer grounded in this specific context — not from its general training data.
Website crawl

Turn your website into a knowledge base

Configurable crawl

Set max pages, crawl depth, and which external links to follow or skip.

Auto Q&A generation

AI generates question-answer pairs from each page, scored by confidence.

Clean text extraction

HTML is cleaned, metadata extracted (headings, links). No menu/footer noise.

FAQ

Common questions about the knowledge base

PDF, DOCX, and TXT. Fiko extracts text, normalizes it (including Persian/Arabic-aware text handling), splits it into overlapping chunks, embeds each chunk, and stores them for retrieval.
Every customer question is embedded (turned into a vector) and compared to all stored knowledge chunks using similarity search. The top-matching chunks are passed to the AI as context. This is the 'retrieval' in Retrieval-Augmented Generation.
Yes. Add your website as a source, configure max pages, crawl depth, and external link rules. Fiko crawls asynchronously, cleans HTML, extracts metadata, and can auto-generate Q&A pairs from each page.
Connect your WooCommerce store. Fiko syncs products automatically, including title, description, price, sale price, currency, images, category, and stock status. Updates sync via webhooks.
Yes. Each AI reply has a debug view showing which knowledge chunks were retrieved and their similarity scores. This helps you debug bad answers and add missing knowledge.
Three ways: re-crawl websites on a schedule, re-sync product catalogs via webhooks, and manually edit FAQs or upload new documents anytime. Fiko re-embeds changed content automatically.
Yes. Add content in any language. The embeddings work multilingually, so a Turkish customer question can retrieve relevant English content if needed (Fiko then translates the answer back to Turkish).

Train your AI in under 30 minutes

Upload a few FAQs, connect your website, and watch the AI come alive with accurate answers.

Start Free Trial