Process

How KBForge Works

From domain expertise to live AI agent in four steps. Your knowledge, validated by experts, deployed under your brand.

01

Upload Your Documentation

Provide your knowledge base content in any format. Our team reviews and validates the material against published literature, then KBForge structures, chunks, and indexes everything into a FAISS vector search index.

Markdown Files

Best quality. Drop in a folder of .md files with frontmatter headers.

PDF Documents

Manuals, textbooks, guides. We extract chapters and split into articles.

Wiki Exports

Confluence, GitBook, ReadTheDocs. Export and we handle the rest.

02

Configure Your Brand

Your AI agent wears your brand — not ours. Customize every visible element so customers see a seamless extension of your product.

Company name & logo
Custom domain (support.yourco.com)
Hero text & taglines
Quick-topic buttons
Custom system prompt
Footer text & download filenames
03

Choose Your AI Model

Three tiers of models to match your use case and budget. All use the same retrieval pipeline — only the synthesis engine changes. Switch anytime.

Mini Models

Grok-mini · GPT-4o-mini

Fast and cheap. Best for straightforward lookups, how-to questions, and high-volume use cases.

~$0.01/query

Chat Models

GPT-5.2 · Claude Haiku · Grok

Strong all-rounders. Good balance of quality, speed, and cost for most support workflows.

~$0.03–$0.05/query

Reasoning Models

Grok 4 · Sonnet 4.6 · GPT-5.4

Deep multi-step reasoning. Ideal for complex analytical questions — spectral overlaps, matrix corrections, protocol troubleshooting.

~$0.08–$0.15/query

04

Go Live

Your AI support agent is deployed to a dedicated container with auto-SSL, user authentication, admin analytics, and feedback capture. Share the link with your customers and start reducing support tickets.

$ curl https://support.yourcompany.com/health

{"status": "ok", "tenant": "yourcompany", "model": "grok-3-mini-latest", "kbs_loaded": 1}

SSL
Auto-SSL
AUTH
User Auth
DATA
Analytics
LOG
Feedback
Architecture

Under the Hood

┌──────────────────────────────────────────────────────────┐ │ Customer visits: support.yourcompany.com │ └─────────────────────┬────────────────────────────────────┘ │ ┌───────▼────────┐ │ Caddy Proxy │ ← Auto-SSL, HTTP/3 │ (shared) │ └───────┬────────┘ │ ┌────────────▼────────────────┐ │ Docker Container │ │ ┌──────────────────────┐ │ │ │ FastAPI + Uvicorn │ │ ← Your branded agent │ │ ┌────────────────┐ │ │ │ │ │ Tiered Search │ │ │ ← FAISS vector retrieval │ │ │ Answer Builder │ │ │ ← LLM synthesis │ │ │ User Auth │ │ │ ← SQLite per-tenant │ │ │ Analytics │ │ │ ← Usage tracking │ │ │ Feedback │ │ │ ← Screenshot capture │ │ └────────────────┘ │ │ │ └──────────────────────┘ │ │ │ │ /data/kb/ (RO) │ ← Your documentation │ /data/vectors/ (RO) │ ← FAISS indexes │ /data/users.db (RW) │ ← User database │ /data/feedback/ (RW) │ ← Feedback logs └─────────────────────────────┘ │ ┌────────────▼────────────────┐ │ LLM API (your choice) │ │ Grok / Claude / OpenAI │ └─────────────────────────────┘

Ready to get started?

Our experts will help build and validate your knowledge base.