ai-agent
Configure the eduuw AI bot and keyword auto-replies over inbound WhatsApp.
Works with
Claude CodeCursorCodex CLIGitHub CopilotGemini CLI
---
name: ai-agent
description: Configure the eduuw AI bot and keyword auto-replies over inbound WhatsApp.
license: MIT
---
# AI Agent
## When to Use
Make eduuw answer inbound WhatsApp messages automatically — either with an **AI bot** (LLM over your knowledge) or with **keyword auto-replies** (deterministic). Base + auth: see `eduuw-rules`.
These run on the server (no code), unlike `functions` (custom TypeScript).
## AI bot
A per-tenant autonomous bot that replies to inbound messages using an LLM, optionally grounded on trained sources (RAG). Inert until enabled and (for RAG) trained.
### Read / configure
```bash
curl https://api.eduue.com.br/ext/v1/whatsapp/ai-bot \
-H "X-API-Key: $EDUUE_API_KEY" -H "X-Tenant-ID: $TENANT_ID"
curl -X POST https://api.eduue.com.br/ext/v1/whatsapp/ai-bot \
-H "X-API-Key: $EDUUE_API_KEY" -H "X-Tenant-ID: $TENANT_ID" -H "Content-Type: application/json" \
-d '{
"enabled": true,
"system_prompt": "Você é o atendente da escola. Responda sobre cursos, matrícula e horários.",
"destino": "curso-enfermagem",
"fallback_text": "Um momento, vou chamar um atendente."
}'
```
Fields: `enabled`, `system_prompt`, `agent_key` (optional named agent), `destino` (RAG knowledge scope), `fallback_text`.
### Train (RAG sources)
```bash
# index text or a URL as knowledge for the bot
curl -X POST https://api.eduue.com.br/ext/v1/whatsapp/ai-bot/train \
-H "X-API-Key: $EDUUE_API_KEY" -H "X-Tenant-ID: $TENANT_ID" -H "Content-Type: application/json" \
-d '{"text":"A matrícula abre em julho. Mensalidade R$ 199...", "destino":"curso-enfermagem"}'
# list trained sources
curl https://api.eduue.com.br/ext/v1/whatsapp/ai-bot/sources \
-H "X-API-Key: $EDUUE_API_KEY" -H "X-Tenant-ID: $TENANT_ID"
```
## Keyword auto-reply (deterministic)
Rule-based replies matched against inbound text — runs before/instead of the AI bot for known intents.
```bash
# list
curl https://api.eduue.com.br/ext/v1/whatsapp/autoreply \
-H "X-API-Key: $EDUUE_API_KEY" -H "X-Tenant-ID: $TENANT_ID"
# create/update a rule
curl -X POST https://api.eduue.com.br/ext/v1/whatsapp/autoreply \
-H "X-API-Key: $EDUUE_API_KEY" -H "X-Tenant-ID: $TENANT_ID" -H "Content-Type: application/json" \
-d '{"name":"saudação","match_type":"CONTAINS","keyword":"oi","reply_text":"Olá! Como posso ajudar?","enabled":true,"priority":10}'
# delete a rule
curl -X DELETE https://api.eduue.com.br/ext/v1/whatsapp/autoreply/<ID> \
-H "X-API-Key: $EDUUE_API_KEY" -H "X-Tenant-ID: $TENANT_ID"
```
`match_type`: `CONTAINS | EQUALS | STARTS`. Higher `priority` wins when multiple rules match.
## Conversation insights (analytics)
```bash
curl https://api.eduue.com.br/ext/v1/whatsapp/cx/insights -H "X-API-Key: $EDUUE_API_KEY" -H "X-Tenant-ID: $TENANT_ID"
curl https://api.eduue.com.br/ext/v1/whatsapp/cx/topics -H "X-API-Key: $EDUUE_API_KEY" -H "X-Tenant-ID: $TENANT_ID"
```
Scope: `analytics:read`. Topics are AI-classified from recent conversations.
## Choosing
- **Known intents / compliance-sensitive replies** → `autoreply`.
- **Open-ended Q&A grounded on your content** → AI bot (train it).
- **Custom logic / external API calls** → `functions` (TypeScript).More AI & ML skills
writing-shape
mattpocock/skills
Writing, exploit: shape raw material into an article, paragraph by paragraph.
275.5k
writing-fragments
mattpocock/skills
Writing, explore: mine raw fragments, no structure yet.
275.3k
full-output-enforcement
leonxlnx/taste-skill
Overrides default LLM truncation behavior. Enforces complete code generation, bans placeholder patterns, and handles token-limit splits cleanly. Apply to any task requiring exhaustive, unabridged output.
273.1k

