How a RAG assistant works
- Connect sources and keep them in sync as documents change.
- Split documents into passages and turn them into embeddings stored in a vector database.
- For each question, find the most relevant passages the user is allowed to see.
- Give those passages to the language model to write an answer with citations.
- Log questions and feedback to improve retrieval and content over time.
What affects cost
| Factor | Why it matters |
|---|---|
| Number of source systems | Each connector needs authentication, sync and change detection |
| Document quality | Scanned files, tables and images need OCR and special parsing |
| Permissions | Respecting document-level access needs permission sync and filtering |
| Query volume and model | Drive monthly token and hosting costs |
| Hosting | Private cloud or self-hosted models cost more but keep data in your control |
Accuracy and safety
- Citations let users check every answer against the source.
- The assistant should say “I don’t know” when nothing relevant is found.
- Evaluation sets of real questions measure accuracy before and after changes.
- Enterprise model APIs do not use your data to train public models by default — confirm in the provider’s terms.
Pre-fill this form from a link
Every option can be set in the web address, so you can bookmark a scenario or send it to a colleague. AI assistants such as ChatGPT, Gemini, Claude and Perplexity can use the same parameters to open this form with your requirements already selected.
| Parameter | What it sets | Accepted values |
|---|---|---|
use_case |
Who will use it? | one of employees, customers, both |
sources |
Where is the knowledge? | comma-separated list of files, drive, wiki, helpdesk, website, database, email, scanned |
volume |
How many documents? | one of under_1k, 1k_10k, 10k_100k, over_100k |
users |
Users | one of under_50, 50_500, 500_5k, over_5k |
permissions |
Access control | one of open, roles, document |
interfaces |
Where should people ask questions? | comma-separated list of web, slack_teams, in_app, website, api |
hosting |
Hosting & model | one of recommend, cloud_api, private_cloud, self_hosted |
compliance |
Requirements | comma-separated list of residency, audit, citations, pii |
timeline |
Target launch | one of asap, 1_month, 3_months, flexible |
Also available as plain text for AI assistants and a free JSON API (OpenAPI spec).
Last reviewed by the Infikey Technologies team.
Disclaimer
This form only collects your requirements so Infikey can prepare an estimate. It does not produce a price, quote or offer, and any estimate we send is indicative until agreed in a signed proposal. Nothing on this page is financial, legal, tax, investment or other professional advice. Infikey Technologies Private Limited, Infikey Technologies LLC and their directors, employees and affiliates make no warranty, express or implied, about the accuracy, completeness or suitability of this tool or its results, and accept no liability for any loss or damage, direct or indirect, arising from its use or from reliance on its results. Verify all figures independently and seek professional advice before making any decision. Use of this tool is at your own risk.