Appendix

Quick-reference tables for file types, keyword lists, environment variables, database paths, ports, and troubleshooting FAQ.

File Types & Upload Limits

Product Specs additionally supports .json, .csv, and .xlsx (Excel) for bulk import. Pricing supports .json and .csv.

Pricing Keywords

When any of these 13 keywords appear in a search query, the pricing database is included in the search pipeline:

harga, price, berapa, biaya, cost, tarif, pricing,
promo, diskon, discount, budget, kisaran, sekitar

Price Range Trigger Words

Indonesian Price Units

Product Spec Keywords

When any of these 36 keywords/phrases appear in a search query, the product specs database is included in the search pipeline:

spesifikasi, spec, fitur, feature, transmisi, transmission,
mesin, engine, cc, tenaga, hp, hybrid, electric, listrik,
bensin, petrol, 4wd, awd, fwd, sunroof, carplay, kamera,
kamera 360, 360 camera, wireless, kursi, kapasitas, seat,
cooling seat, heated seat, cruise control, lane assist,
android auto, keyless, push start, rear camera, electric seat,
punya, ada fitur, apakah ada

Stop Words

Stop words are filtered out in Search Insights analytics and in query cleaning before database searches.

English Stop Words (~96 words)

the, a, an, and, or, but, in, on, at, to, for, of, with, by,
from, as, is, was, are, were, been, be, have, has, had, do, does,
did, will, would, could, should, may, might, can, this, that,
these, those, i, you, he, she, it, we, they, what, which, who,
when, where, why, how, my, your, his, her, its, our, their, me,
him, us, them, am, if, then, than, so, also, just, some, any,
all, each, every, both, few, more, most, other, such, no, nor,
not, only, own, same, too, very, much, many, about, above, after,
again, against, among, another, before, below, between, down,
during, into, off, once, out, over, through, under, until, up,
upon, within, without

Indonesian Stop Words (~65+ words)

yang, dan, untuk, atau, dalam, adalah, dengan, dari, pada, ke,
ini, itu, ada, juga, akan, saya, kamu, kami, mereka, dia, ia,
sudah, belum, masih, oleh, tersebut, dapat, bisa, harus, perlu,
seperti, karena, sebagai, antara, setiap, salah, satu, bagi,
agar, sekitar, hingga, namun, tetapi, bahwa, maka, bila, kalau,
jika, apakah, dimana, bagaimana, mengapa, kapan, siapa, apa,
mana, berapa, telah, sedang, lagi, pula, pun, saja, demi, mau,
biar, kok, sih, dong, lho, tapi, kalo, gak, nggak, engga, tak,
nah, yah, kan, deh, nih, tuh, nya, kah, lah, pasti, mungkin

Pricing & Spec Noise Words

Additional noise words are stripped from queries before searching the pricing and product spec databases. These include common filler words like:

sekarang, terbaru, terkini, saat, ini, itu, yang, dan, atau,
untuk, dari, dengan, ada, apa, mana, bisa, beli, jual, cari,
mau, ingin, tolong, mohon, dong, ya, nih, deh, sih, kan, lah,
the, is, are, what, how, much, does, can, latest, current, now,
today, new, please, mobil, motor, kendaraan, vehicle, car, list,
daftar, info, informasi, detail, type, tipe, model, varian, variant

Environment Variables

All variables are loaded from a .env file in the backend/ directory.

LLM & AI Providers

Authentication

Database

Search & Chunking

Default Port Numbers

Database File Locations

All paths are relative to backend/data/:

Keyboard Shortcuts

Query Topic Categories

The query decomposer maps ~107 keywords to search-hint categories for improved retrieval. Below are the category groupings:

Security Configuration

Troubleshooting FAQ

Search returns no results

• Check that documents have been uploaded and indexed
• Verify the relevance threshold isn't too high (default 0.5)
• Try lowering the similarity threshold in Settings
• Check ChromaDB is running (VectorDB Viewer → collection info)
• Use Debug Search to inspect raw scores

Pricing data not found in search

• Ensure your query contains a pricing keyword (e.g., "harga", "price")
• Verify pricing records exist in the Pricing Database page
• Check that VectorDB sync has been run after adding pricing data
• Use Debug Search to see if pricing pipeline was triggered

Cannot log in

• Default credentials: admin / admin
• Check that the backend is running on port 8000
• Clear localStorage and try again
• Check browser console for CORS errors

LLM response fails or times out

• Verify LLM provider credentials (AWS for Bedrock, Google API key for Gemini)
• Check Settings → LLM Provider is set correctly
• Try switching to a different provider
• Check backend logs for detailed error messages

Widget doesn't appear on page

• Ensure a <div id="kbaas-widget"> element exists
• Verify the script src path to kbaas-widget.js
• Check browser console for initialization errors
• Try calling KBaaSWidget.init() manually in console

VectorDB shows 0 vectors

• Ensure Cohere API key is set in environment variables
• Upload documents and wait for processing to complete
• Run VectorDB Sync from the Pricing or Product Specs page
• Check backend logs for embedding errors