ZERO-HALLUCINATION AI

Your docs, turned into a fact-checked AI assistant.

Upload PDFs and Markdown. Get a citation-grounded chatbot live on your site in under 5 minutes — with zero hallucinations guaranteed.

99.9%Uptime SLA
<300msAvg Response
AES-256Encryption
EMBED SCRIPT
<script src="widget.js" data-id="bfw_9273"></script>
PostgreSQL RLS Verified
Documentation Assistant
Online · Zero-hallucination active
How do I authenticate API requests?
Include your API key in the Authorization header as a Bearer token:
Authorization: Bearer bfw_live_948af7...
Grounded: api_v1_reference.md §1.1
What about rate limiting?
The default rate limit is 100 requests/min per API key. Enterprise plans support custom limits via the dashboard.
Grounded: api_v1_reference.md §2.3
Ask about your documentation...
Interactive Demo

Try It Now - Zero Setup Required

Ask questions, test the citation engine, and see hallucination rejection in real-time. No sign-up needed.

Knowledge Base

Click a file to see its pre-indexed questions

AI

Documentation Assistant

Zero-Hallucination active

Hello! I am the botforweb documentation assistant, grounded strictly in the files listed on the left. Click one of the suggested questions below, or ask your own question to test my zero-hallucination engine.

This is a live demo using pre-indexed documents. Your data stays private with PostgreSQL RLS.

Simple Setup

Launch Your Assistant In 3 Steps

A completely automated ingestion pipeline that handles parsing, chunking, embedding, and widget deployment in under five minutes.

01

Upload Docs

Drag & drop PDFs or Markdown. Celery parses and indexes your files asynchronously, extracting headers and key segments.

manual.pdf
Ingested
02

Auto-Vectorization

The engine splits text into semantic chunks, generates dense/sparse embeddings, and stores them in isolated database namespaces.

"auth key"[0.42, -0.89, ...]
03

Embed Widget

Copy the single-line JS script tag and paste it onto your site. The chatbot is instantly live, customized to your styling.

<script src="widget.js"></script>
Enterprise-Grade Isolation

Secured At The Database Level

We treat data privacy as a non-negotiable core feature. botforweb isolates tenant data completely using advanced PostgreSQL Row-Level Security (RLS) policies and isolated vector index namespaces.

Row-Level Security (RLS)

Your data is strictly yours. PostgreSQL RLS policies guarantee zero tenant data leakage at the query planner level. If a database session isn't authenticated as your tenant, it cannot physically scan your records.

Groundedness Validation

The chatbot runs real-time semantic cross-checks on generated answers. If the requested information is absent or missing in your source files, the bot declines to answer instead of fabricating details.

Hybrid Search Engine

We combine dense vector similarity matching (for semantic meaning) with BM25 sparse keyword searches. Technical terms, exact serial codes, and identifiers are never missed.

Data Isolation Flow
AES-256 Keys Rotated • 100% Isolated
Validated
Client Query"How to auth?"
JWT VerificationTenant ID Decoded
POSTGRES RLS ENGINE
USING (tenant_id = current_tenant());
bfw_4a1f
Your Vectors Only
bfw_c74b
Access Denied
FAQ

Frequently Asked Questions

Details on our vector search pipeline, data encryption policies, and database tenant isolation.

PostgreSQL RLS works directly at the query compiler level inside our database clusters. Every similarity search query or document lookup is dynamically parameterized with the caller's unique tenant_id.

The database engine physically filters out all rows belonging to other tenants before executing any index scans. This means that even in the unlikely event of an application-level routing bug, tenant data can never bleed across workspaces.

No. We enforce strict Retrieval-Augmented Generation (RAG) grounding constraints. When a user asks a question, we first fetch semantic snippets exclusively from your verified documents.

The AI LLM is constrained to synthesize answers based only on those context chunks. Before displaying the answer, our engine runs a secondary verification check. If the generated output is not backed by the source citations, the widget displays a helpful fallback refusal rather than generating false info.

We currently support PDF manuals, Markdown (.md / .markdown), JSON files, and plaintext documents.

During upload, our parsing microservice (managed with Celery and python-docx/pdfplumber) strips structural noise, splits files into logical headers/paragraphs, cleans markdown tables, and maps them to dense embeddings.

Yes. The Growth tier allows you to map widgets and dashboard endpoints to a custom subdomain of your own (e.g., docs-helper.yourcompany.com).

We automatically issue and maintain Let's Encrypt SSL/TLS certificates for your custom subdomains, which are served through our secure global edge CDN.

All traffic between your site, your dashboard, and our databases is encrypted using TLS 1.3.

Document vectors and raw parsed contents are encrypted at rest using AES-256. Database backups and read replicas maintain the same level of encryption, ensuring full regulatory security compliance.

Ready to deploy your zero-hallucination assistant?

Join security-conscious developer teams using botforweb to turn static documentation, manuals, and PDFs into an intelligent, PostgreSQL RLS-secured AI support channel. If you are interested, raise interest below and we will reach out to you.

Rest assured - we will get back to you within 24 hours.