🌊v2.0 — now with streaming answers

Answers from your documents — with citations, not guesses.

Dolpix is a RAG platform that turns your PDFs, runbooks, and policy docs into an AI assistant that cites every answer. One project per knowledge base. Full tenant isolation.

Get started — it's freeSee a live demo
· 50 MB per file· PDF · MD · DOCX · TXT· 340 ms median reply· 99.98% uptime
HR Assistant
What's our parental leave policy?
Primary caregivers receive 16 weeks of fully paid leave. Secondary caregivers receive 4 weeks. Leave may begin up to 4 weeks before the due date.
policy-benefits.pdf · p.14
Can I take it in parts?
Yes — leave can be split into two blocks, provided HR is notified 8 weeks in advance.
policy-benefits.pdf · p.15

Knowledge base

policy-benefits.pdfREADY
employee-handbook.mdREADY
expense-guide.pdfPROCESSING

3 docs · 1 indexing…

Trusted by teams building AI on their own knowledgeNorthwindHelix LabsMonoformVeridianAcme Co.Nori Studio

How it works

Up and running in minutes

Four steps from signup to production queries — no SDK, no boilerplate.

01

Upload

Drop PDF, MD, DOCX, TXT

Up to 50 MB per file

onboarding-guide.pdf

PROCESSING

Drop PDFs, policy docs, runbooks — one project per knowledge base.

02

Index

policy-benefits.pdfREADY
employee-handbook.mdREADY
expense-guide.docxPROCESSING
live feed — SSE stream

We chunk, embed, and index in seconds. Watch live status via Server-Sent Events.

03

Configure persona & rules

Pick a voice. Turn on the guardrails. One click each — zero re-indexing.

04

Embed or query

Support Bot

Hi! Ask me anything about your benefits.

policy-benefits.pdf · p.14
<script src="https://cdn.dolpix.com/widget.js"
data-app-key="YOUR_APP_KEY"></script>

Paste one <script> tag or call our REST API directly.

Live chat example

What's our remote work allowance?
Employees are entitled to a one-time $600 home-office setup allowance, plus $50/month for internet reimbursement.
remote-work-policy.pdf · p.3
Is there a deadline to claim the setup allowance?
Claims must be submitted within 90 days of your start date or the policy effective date, whichever is later.
remote-work-policy.pdf · p.4

Grounded answers

Grounded. Every. Single. Answer.

Dolpix uses vector similarity retrieval to surface the most relevant chunks from your documents, then constrains the LLM to answer only from those chunks — so every answer is traceable to a source paragraph.

  • RAG pipeline with text-embedding-3-small for high-quality retrieval
  • Citations enforced at the prompt level — not optional post-processing
  • Structured refusal when nothing relevant is found — no hallucinations
  • Top-k chunks configurable per request (1–20)

Persona + Rules

A pod, not a prompt. Tune your voice.

Every project gets its own persona and a set of guardrail toggles. The system prompt is compiled on the fly — change a toggle, the next query reflects it immediately.

  • Pick from curated presets (Friendly Support, Technical Expert, Concise Helper)
  • Write your own system prompt — up to 4,000 characters
  • Toggle 7 guardrails: grounding, citations, speculation, competitors, and more
  • Free-text fields for topic scope, forbidden topics, and support contact

Pick a preset

Compiled system prompt

You are a warm, patient support assistant. Use a casual tone and contractions. If the user seems frustrated, acknowledge their feeling before answering. Always cite the source file and page.

Embeddable widget

Embed anywhere. Theme it in seconds.

Dolpix ships a ready-made chat widget you can drop on any site with a single script tag. Pick your brand colour, write a greeting, and you're live.

  • One <script> tag — no npm, no framework dependency
  • Theme colour matches your brand — set via data attribute
  • Quick-reply chips for common questions
  • Citation chips link back to source files
  • Thumbs up/down feedback captured per response

Support Bot

Powered by Dolpix

Hi! Ask me anything about your plan.

Dolpix shared platform

NNorthwindISOLATED
📄 Docs🔑 Keys🤖 Persona📊 Logs
HHelix LabsISOLATED
📄 Docs🔑 Keys🤖 Persona📊 Logs
AAcme Co.ISOLATED
📄 Docs🔑 Keys🤖 Persona📊 Logs

Multi-tenant isolation

Isolation you can audit.

Every tenant's documents, embeddings, queries, and API keys are isolated at the database level via row-level security on (tenant_id, project_id). There is no path from one tenant's keys to another's data.

  • Row-level security enforced on every project-scoped table
  • Separate vector namespaces per tenant — no cross-contamination
  • Scoped API keys: an appKey + apiKey pair unlocks exactly one project
  • AES-256 encryption at rest · Rotate keys any time from the dashboard

Use cases

Dolpix works great for…

👩‍💼

Employee HR helper

Answer benefits, leave, and expense policy questions — straight from your handbook.

⚙️

Engineering runbook

On-call rotations, deployment guides, service ownership — at your team's fingertips.

🎧

Customer support copilot

Embed on your site. Answer product questions grounded in your docs — 24/7.

📊

Investor relations

Let investors self-serve from your filings, reports, and prospectus.

REST API

One REST call. No SDK required.

Swap in your app key and API key. That's it. Works from any language, any framework.

bash
# Ask a question against your knowledge base
curl -X POST https://api.dolpix.com/api/v1/ai/chat \
  -H "Content-Type: application/json" \
  -H "x-app-key: YOUR_APP_KEY" \
  -H "x-api-key: YOUR_API_KEY" \
  -d '{"query": "What is our refund policy?"}'

# Response
{
  "answer": "Based on your documents, refunds are accepted within 30 days...",
  "sources": [{
    "filename": "refund-policy.pdf",
    "chunk": "Refunds are accepted within 30 days of purchase..."
  }]
}
Read the full API docs →

Pricing

Simple, transparent plans

Start free. Upgrade when you grow. All plans include citations, SOC-2, and tenant isolation.

Free

For individuals exploring Dolpix.

₹0

  • Projects Count: 1
  • Documents Per Project: 50
  • Storage Mb: 100
  • Queries Per Month: 1000
  • Log Retention Days: 7
Start free

All plans include citations, SOC-2 Type II, AES-256 at rest, and tenant isolation.  See full plan comparison →

FAQ

Common questions

What formats do you support?

PDF, MD, DOCX, and TXT — up to 50 MB per file. Batch upload of up to 10 files at once is also supported. URL ingestion is available on Starter and above.

Where does the AI get answers from?

Only your uploaded documents. Every reply is grounded in retrieved chunks from your knowledge base and cites the source file and page. If nothing relevant is found, the bot issues a structured refusal — it never hallucinates.

Which LLM do you use?

OpenAI GPT-4o-mini by default. Pro plans and above unlock GPT-4o for higher accuracy on complex queries. Embeddings use text-embedding-3-small.

Can I self-host?

Self-hosting is available on the Enterprise plan. Contact us for infrastructure requirements, SLAs, and migration support.

How do you keep tenant data isolated?

Every tenant's data is protected by row-level security on (tenant_id, project_id) in PostgreSQL. Vector embeddings are namespaced per tenant. API keys are scoped to a single project and cannot reach other tenants' data.

What happens when I hit my plan limit?

We soft-warn at 80% usage via email and in-dashboard alerts. At 100% we hard-block further queries until the billing cycle resets or you upgrade. Upgrades take effect immediately — no downtime.

What teams say

Real results. Cited.

“

We replaced three separate helpdesk bots with Dolpix. Citations shut down the "the AI made it up" complaints completely.

M

Mira K.

Head of People Ops · Northwind

“

Our on-call runbook is 400 pages. Dolpix turned it into an answer in 1.8 seconds — with the page reference attached.

A

Asha R.

Platform Lead · Helix Labs

“

Legal approved it because every answer links back to the source paragraph. That was the blocker for six months.

A

Arun P.

Compliance · Veridian

🐬

Ready to stop guessing?

Upload your first document in under two minutes. No credit card needed on the free plan.

Start free · 14 daysTalk to us →