Developer Docs

Everything you need to integrate Tavio into your products: authentication, channel setup, HTTP APIs, and webhook configuration.

Quickstart

Create your Tavio account
Grab your API key from Settings
Drop the widget script on your site

HTTP API

Send and receive messages, trigger workflows, and query analytics.

View reference

10 Channels

Widget, WhatsApp, SMS, USSD, Voice, Telegram, Messenger, Instagram, Twitter, Discord.

Channel guides

Knowledge Base

Power your AI with RAG: add URLs, PDFs, or text to build your knowledge base.

Knowledge guide

Architecture overview

Every message from every channel flows through a unified pipeline:

Core Pipeline

text

Channel Adapter -> Message Router -> RAG Retriever -> AI Engine -> Response

Key concepts:
- Multi-tenant: each institution is isolated via RLS
- Channel-agnostic: same AI, different delivery
- RAG-powered: your knowledge base informs every response
- Streaming: real-time responses via SSE

Knowledge Base (RAG)

The Knowledge Base powers your AI agent with real, accurate information from your institution. Instead of generating generic responses, the AI retrieves relevant context from your indexed content before responding.

Adding Sources

Web URLs — paste a URL and Tavio crawls, extracts, and indexes the content
Manual text — paste product docs, FAQ answers, or policy content directly
PDF uploads — upload documents for automatic extraction and indexing

How It Works

Content is split into ~512-token chunks with 64-token overlap
Each chunk is converted to a 1536-dim vector embedding
Embeddings are stored with HNSW indexing for fast search
At query time, top-k most similar chunks are injected into the AI prompt

Knowledge Sources API

bash

# List all knowledge sources
GET /api/knowledge/sources
Headers: x-api-key: <your-api-key>

# Add a URL source (triggers async crawl)
POST /api/knowledge/sources
{ "type": "url", "source_url": "https://yourbank.com/faq" }

# Add manual content
POST /api/knowledge/sources
{ "type": "manual", "title": "Loan FAQ", "content": "..." }

# Resync a source
PATCH /api/knowledge/sources
{ "id": "<source-id>", "action": "sync" }

# Delete a source and all its chunks
DELETE /api/knowledge/sources?id=<source-id>

Tip: Add your institution's public FAQ, product pages, and help center articles first. The more relevant content you index, the more accurate your AI responses become. Sources auto-refresh on resync — keep content up to date by triggering periodic resyncs.

Ready to get started?

Create a free Tavio account to get your API key and start building.

Create Free Account