Wavo — your AI assistant in Wevion — complete guide

Last updated: May 19, 2026

Wavo — your AI assistant in Wevion — complete guide

TL;DR: Wavo is Wevion's in-app AI assistant. 10 verified models across 5 providers (Anthropic / Google / DeepSeek / Moonshot / OpenAI). Default: claude-sonnet-4-5-20250929. Action approval workflow (HIGH / MEDIUM / LOW risk with 30-min TTL). Team-isolated data + per-user RBAC. Voice TRANSCRIBE-ONLY (Whisper) — no TTS output. Credits separate from Creative Hub.

Table of contents

Meet Wavo

/chat. AI assistant with persona file wavo-persona.ts (canonical name post-rebrand).

How Wavo works

Hard constants: MAX_HISTORY=20 messages, MAX_TOOL_ROUNDS=5 per turn, MAX_PROJECT_MEMORY_FACTS=20, ACTION_TTL=30min, MARKUP=1.1×, 1 credit = $0.001.

Ask questions effectively

Text input + voice (Whisper transcribe-only, 25MB max, BCP-47 lang) + file upload (10MB max, MIME whitelist).

Wavo and your data

Team-isolated. Per-tool RBAC. Never fabricates (cites tools). Output sanitization for secrets.

Actions (approval workflow)

3 tool modes:

  • Read (instant) — campaign_search, get_performance_data, etc.

  • Write (instant + audit) — create_campaign_draft (LOW risk)

  • Approval (pending → confirm → execute) — HIGH (delete_rule, duplicate_entity, update_entity_budget) + MEDIUM (create + toggle rules + entity)

TOOL_RISK_MAP categorizes 30+ tools. ai_action_log records every approval-mode execution (SHA-256 payload hash + idempotency key).

Projects + sessions

3-level structure: Project > Session > Message.

Per-project memory facts (max 20). Sessions shareable + archivable + favoritable. User preferences (preferred_language='auto' default, preferred_model, tone, etc.).

Credits + limits

Separate ledger from Creative Hub. 1 credit = $0.001 with MARKUP 1.1× on input + output tokens.

Privacy + data usage

Team isolation, role-based access, prompt injection defense (50+ regex EN+IT), output sanitization (redacts Meta/JWT/AWS/Stripe/Google/DB secrets), memory opt-out per user + per project.

Feedback + improve

Thumbs up/down (feedback_rating SmallInt -1..1) + optional comment. Aggregated for prompt + model tuning.

Troubleshoot

Common: wrong data (postback lag), stream interrupted, approval expired (30 min), permission denied, out of credits, file rejected, hallucination, slow response, lost memory.

FAQ

What models does Wavo support?

10 verified: Anthropic Claude (Opus 4.6, Sonnet 4.5 DEFAULT, Haiku 4.5), Google Gemini (2.5 Pro, 2.5 Flash), DeepSeek (chat, reasoner — no tools), Moonshot Kimi K2, OpenAI GPT-4.1 + mini. See tr-205 FAQ AI features.

Can Wavo take actions on my account?

Yes — with approval for HIGH/MEDIUM risk. LOW risk (drafts) instant. RBAC enforced per tool.

Voice output?

NO. Wavo doesn't generate audio. Voice input only (Whisper transcribe).

Mobile push notifications for Wavo?

NO native mobile app. PWA push via VAPID (web push). Use Telegram for real-time mobile alerts.

Does Wavo see my data?

Team-scoped only (strict isolation). Sent to model providers per their enterprise terms.

Memory across sessions?

Yes — project memory facts (max 20 per project). Opt-out per user + per project.

Wavo vs Help Center?

Wavo = account-aware + action-capable. Help Center = static reference. See ai-108 Wavo vs Help articles.

Can Wavo open support tickets?

Yes — create_support_ticket tool (LOW risk, no approval needed).

Next steps

Related