Documentation Index
Fetch the complete documentation index at: https://docs.wittify.ai/llms.txt
Use this file to discover all available pages before exploring further.
This page is the catch-all for questions that don’t fit cleanly into a single product page. It covers every product (AI Agents, Speech to Text, Text to Speech, Contact Center QA, Chat with Your Documents) plus account-wide topics. If a question relates to one specific page in the app, the answer points at the deep-dive doc instead of repeating it here.
1. Platform overview
What is Wittify? Wittify is a no-code workspace with five products that share a single account, billing, and team. The five products are AI Agents, Speech to Text (Faheem), Text to Speech (Faseeh), Contact Center QA, and Chat with Your Documents. Switch between them with the system switcher in the sidebar. What makes Wittify different from generic AI platforms? Three things. Arabic-first: 25+ Arabic dialects, voice models tuned for the region, RTL across the entire interface. Multi-product: agents, transcription, voice synthesis, call center analytics, and document chat in one workspace. No code: every product ships a UI deep enough to skip a developer for the first deployment. Who uses Wittify? Startups, enterprises, e-commerce, government agencies, customer support teams, agencies, and individual operators. The plans (Sandbox / Starter / Scale / Pro / Enterprise) span from a free trial to fully custom enterprise deployments. Which languages does Wittify support? 100+ languages. Arabic gets the deepest investment (25+ dialects, Arabic-first speech and voice models). Every product ships with EN + AR interfaces, mixed-language support, and full RTL.2. Account, profile, and login
How do I create an account? Open https://app.wittify.ai/signup. Sign up by email (with an 8-character verification code) or with Google, Microsoft, or GitHub. The full step-by-step is on First-Time Setup. What information does the signup ask for? Email only at first. After verification, you fill in First Name, Last Name, Phone Number (with country dropdown), Password (email signups only), Company Name, and an optional URL. All on the Profile Completion page. Can I update my profile later? Yes. Profile dropdown → Settings, or sidebar System → Settings. The full reference is on Profile Settings. Can I change my email? Email is read-only on Profile Settings. Changing it requires re-verifying the new address; that flow is handled by support today. I forgot my password. Open https://app.wittify.ai/forgot-password, enter your email, click Send reset code. Enter the code on the next page, then choose a new password. Step-by-step on Forgot Password. Can I turn on two-factor authentication? Yes. Profile Settings → Security → toggle Two-factor authentication. Enabling is one click; disabling opens a confirmation dialog (see Profile Settings).3. Plans, billing, and payments
Which plans are available? Sandbox (free trial), Starter, Scale, Pro, Enterprise. Quotas, channels, integrations, branding, and support tiers vary. The full table lives on Quick-Start → Plans and Billing. The per-control reference is at Account → Plans and Subscriptions. How many AI agents can I create on each plan? Unlimited on every plan, including Sandbox. Plans differ on credits, voice minutes, channels, knowledge base size, support tier, branding, and analytics depth, not on the number of agents you can build. How do I upgrade or downgrade? Profile dropdown → Plans & Subscriptions → Plans and Add-Ons tab → click any plan card’s CTA. The dialog shows proration for upgrades, credit-on-account for downgrades. Can I buy extra usage without changing plans? Yes. Plans and Add-Ons tab → scroll to Top-ups for extra text conversations or voice minutes, Services for one-time professional services, Features for subscription add-ons. Which payment methods are supported? Credit and debit cards via Stripe today. Other methods are coming. Wittify never holds your card data. Where do I find my invoices? Billing History tab inside Plans and Subscriptions. Each invoice has a Download link (PDF, Stripe-generated) and a Retry Payment button on failed or open invoices. How do I cancel? Overview tab → Cancel Subscription. The confirmation dialog tells you the period-end date. You keep access until then; after that the account moves to Sandbox. Resume any time before the period ends with Resume Subscription.4. AI Agents, building and configuring
What is the agent name used for? Identifying the agent in your dashboard. It is not shown to your customers. Pick a clear, descriptive name like “Hilton Riyadh, Pre-arrival concierge” rather than “Bot 1”. What are Core Instructions? The system prompt that defines the agent’s role, goals, tone, restrictions, and fallback behaviour. The deep dive is on Text Settings (text and hybrid agents) and Voice Settings (voice and hybrid agents). Best practices for Core Instructions? Write in full sentences, be specific, include fallback instructions for unknown questions, and explicitly state restrictions (no medical, legal, financial advice, etc.). The page-level deep dive has worked-example prompts. Can I start from a template? Yes. The Text Settings step has quick-template chips (Customer support, HR assistant, Government services) and a More templates dropdown grouped by category. Click a template to load it; you can edit afterwards. Re-clicking the same template highlights it as active. What is Creativity Level (Temperature)? Controls how varied the model’s wording is. Lower means more deterministic; higher means more varied. 0.0 to 0.3 for factual deployments, 0.4 to 0.6 for balanced, 0.7 to 1.0 for creative or marketing copy. Default is0.5.
What are Top-P, Max Tokens, and Reasoning Level?
Advanced settings. Top-P (also called nucleus sampling) limits the model’s word choices to a probability mass; lower values make the agent stay closer to the most likely word. Max Tokens is the hard cap on response length. Reasoning Level lets you trade speed for depth on models that support it. All three live in the model selector on Text Settings.
Can I restrict topics?
Yes. List restricted areas in Core Instructions. Example: “Do not discuss competitor pricing. Do not give medical advice.”
5. Appearance and multilingual experience
How many interface languages can I configure for the widget? Up to 10 languages. Each can have its own greeting, banner, voice texts, rating prompt, and chips. The first language is the default; the widget auto-detects the user’s language and switches. What can I customise per language? Header title, header subtitle, welcome message text, voice agent texts (greeting, connecting, listening), input placeholder, disclaimer text, welcome banner title and subtitle, post-chat rating prompt, and per-star chips (max 5 per star). Can I change branding? Yes. Logo, bot avatar, primary and secondary colors (solid or gradient), 12-color swatches with hex input, font (15 fonts with bilingual fallback stacks), widget position (left / center / right plus a 3×3 fine-tune grid), launcher shape (circle / square / oval), and launcher icon source (logo / avatar / suggestion / custom upload). Plans on Pro and Enterprise can hide Powered by Wittify. Full reference on Appearance. Can the widget auto-open? Yes. Enable Welcome Message Banner with a small delay (0–30 seconds) and Play Sound toggle. The banner pops up on first visit to draw attention.6. Knowledge Base
What sources can I add? Files (PDF, DOCX, XLSX, TXT, CSV, PNG, JPG; up to 10 MB per file, 100 MB total per agent), URLs (web pages indexed by our crawler), and FAQs (question-answer pairs, up to 500 per CSV import). Full reference on Knowledge Base. How do I refresh a URL? Remove it from the list and add it again. There is no in-place re-index button today (tracked as a backend ask). Best practices for KB content?- Keep info accurate and up to date. Pick a monthly review cadence.
- FAQs win on near-exact matches; include 3–4 paraphrasings of the same question.
- For voice agents, keep answers concise.
- Avoid overloading with outdated pages; the agent will quote stale facts confidently.
7. Voice-specific settings
How do I pick and test a voice? Voice Settings step → Select Voice opens the Voice Picker modal. Search by name / ID / Arabic name, filter by gender. Up to 140 characters of preview text in the modal; click Play to hear the voice. Click Use Voice to apply. What is Enable Auto Detect Language? When on, the agent detects the caller’s language and responds in it. Off forces the agent to stick to the configured language. What is Allow Interruption? When on, callers can interrupt the agent mid-speech (more natural conversation). When off, the agent finishes its sentence before listening (better for compliance scripts). What do the timing controls do? Five timing controls in Advanced Voice Settings:| Control | What it does | Default |
|---|---|---|
| Speaking Speed | Slower or faster delivery (0.5×–2.0×). | 1.0× |
| Min Silence Duration | Pause threshold before the agent begins speaking, prevents talking over the user. | 0.5 s |
| Min Speech Duration | Filters out short noises so they don’t trigger a response. | 0.35 s |
| End Call on Silence | Time to wait when no one speaks before ending the call. | 60 s |
| Max Call Length | Hard cap on call duration. | 5 min |
| Time Before Widget Pop-Up | Delay before the voice widget appears after page load. | 0.1 s |
- Shorter Min Silence → snappier responses, but higher risk of interruptions.
- Longer End Call on Silence → more forgiving for slower speakers.
- For mobile, lengthen pause times slightly to account for microphone latency.
8. Integrations and tools
Which built-in tools can I enable? Human Handoff, Calendar Booking, Order Tracking, Payment Collection. Full reference on Integrations and Tools. Which CRMs are supported? Salesforce, Zendesk, Zoho, HubSpot, plus a generic Custom Webhook and MCP for anything else. The Tickets and Leads step also surfaces these as notification destinations. Which booking tools are supported? Google Calendar and Calendly. Drop in the URL and the durations you offer; the agent generates the booking link mid-conversation. Which telephony providers are supported? Twilio, Telnyx, SIP Trunk, and WhatsApp Calls. Each card has call direction (Inbound / Outbound / Bidirectional), credentials, and phone-number management. Setup details on Deployment. Can I integrate Wittify into my mobile app? Yes. Use a web view pointing at the widget URL, our mobile SDK (where available), or a custom embed for voice agents. Reach out for the deeper integration paths. Can I connect to my own systems? Yes. Use Custom API Tools for REST endpoints (max 10), MCP Connectors for the Model Context Protocol, or Webhooks & Events for outbound notifications. Where do I see what an integration is doing? Manage Agent → Chat / Voice Conversations, click any conversation, look at the activity panel. Tool calls show inline with the assistant’s reply.9. Deployment and channels
How do I embed the widget on my website? Deployment step → Website tab → copy the embed snippet → paste before</body>. The Agent ID in the snippet is URL-encoded.
How do I generate a shareable widget link?
Same tab → Generate Shareable Widget Link opens a dialog with the public URL. Copy and share.
Can I restrict which sites the widget runs on?
Yes. Allowed Domains input on the Website tab. Add a domain → click Activate. Empty list means the widget loads everywhere; restrict early.
Which channels can each agent variant deploy to?
| Variant | Website | Mobile | WhatsApp / Messenger / Instagram / X / Telegram | Telephony |
|---|---|---|---|---|
| Text | Yes | Yes | Yes | No |
| Voice | Yes | Yes | No | Yes (Twilio / Telnyx / SIP Trunk / WhatsApp Calls) |
| Hybrid | Yes | Yes | No (text-only chat by default; route per channel) | No |
10. Testing and quality assurance
Pre-launch checklist:- Run five canonical conversations through Test Your Agent.
- Check language switching (RTL behaviour, mixed-language replies).
- Test edge cases (slang, typos, empty input).
- If lead collection is on, fill in a fake lead and verify it lands in Leads and Tickets.
- Drop the embed snippet on a non-whitelisted page and confirm the widget refuses to load.
- For voice / hybrid: test interruption, silence timeouts, and audio quality on real devices.
- Involve multiple team members.
- Test on both desktop and mobile.
- Use multiple languages and dialects.
11. Conversations, leads, and inbox
What does the Manage Agent inbox show? Hybrid agents show two tabs (Chat Conversations + Voice Conversations); Text agents hide the voice tab; Voice agents hide the chat tab. Each tab has a list pane (search, sort, filter, bulk-mark, bulk-archive, bulk-assign), a detail pane (transcript, reply composer for chat, audio player for voice), and a right-side details panel. What can I do inside Chat Conversations? Search by keyword / customer name / phone / id / intent. Sort by newest / oldest / unread. Filter by status, sentiment, channel, read status, rating, and assignee. Open the conversation, type a reply, attach files (MIME-checked, ≤10 MB, ≤5 files), set priority, assign to a team or agent, archive, or delete. Reply composer respects the floating chat widget so Send is never covered. What can I do inside Voice Conversations? Filter by status, sentiment, source (Web / SIP / Telnyx / Twilio / WhatsApp Call), duration bucket, rating chips. Open a call to see the waveform player (click-to-seek, arrow-key skip 5 s, rewind / forward 10 s, brand-gradient play, volume, speed dropdown), the transcript (AI on the start edge, caller on the end edge), and the AI Call Summary card. Can a human take over a live conversation? Yes. The conversation header has a non-interactive AI status tag (auto / suspended / human). Pause the AI from the right panel; type your reply manually. The transcript shows the handoff timestamp. Where do captured leads go? Leads and Tickets. Each row has the captured fields (Name, Phone, Email, Company, plus your custom fields) and a per-row detail sheet. Bulk export to CSV. Can I export leads? Yes. Toolbar Export CSV at the top of the Leads and Tickets list. UTF-8 BOM + CRLF for Excel-Arabic compatibility, formula-injection defended.12. Speech to Text (Faheem)
What can Speech to Text do? Transcribe audio in 16+ languages with speaker detection, optional subtitles, and a full editor. Three input methods: upload, live transcribe in browser, or record audio. See Speech to Text Overview. What’s the file size limit? 500 MB per upload. MIME type and extension must both match the whitelist. What languages does Live Transcribe support? The same 16+ language list as Upload Audio. Can I edit the transcript after upload? Yes. Open any file in the Transcript Editor. Edit speakers, edit words inline, scrub with the audio timeline, view or edit cue properties. How do I export a transcript? File row’s 3-dot menu → Quick export → PDF / DOCX / TXT. Or open the file and use the editor’s export button. Can I delete a file? Yes. 3-dot menu → Delete. Confirmation dialog. The file is removed from your library; the underlying audio is also deleted.13. Text to Speech (Faseeh)
What can Text to Speech do? Generate natural-sounding voice audio from text in 100+ languages and 51 dialect clusters. Pick from prebuilt voices, clone a voice from your own audio, or design a voice from attribute dropdowns. See Text to Speech Overview. Which providers are behind the voices? The provider is an internal routing detail. Voices are Wittify voices regardless of which engine generates them. Can I clone a voice? Yes. + Create Voice → Clone Wizard. Two steps: upload or record audio (with transcription), then Generate Custom Template. See Clone Wizard for requirements (2–3 hours of professionally recorded clean audio is recommended for best fidelity). Can I design a voice without uploading audio? Yes. + Create Voice → Design Wizard. Pick attributes from dropdowns (gender, age, tone, language, dialect), preview, then Generate Custom Template. See Design Wizard. Are my custom voices visible to other accounts? No. Each cloned or designed voice is private to your account and only appears to you.14. Contact Center QA
What does Contact Center QA do? Analytics over your call center recordings: sentiment, intents, speaker timeline, alerts, agent scorecards, compliance, volume forecasting, ask-and-visualize, and a streamed narrative report (TimeLens AI). See CCQA Overview. How does the date / time filter work? Every page mounts the shared Filter Bar under the hero. Quick-range tabs (7d / 30d / 90d / Custom), a date picker for custom ranges, hour-from / hour-to selects (defaults 0–23), and a Reset button. Every page’s data is reactive to this filter, including the Overview KPI strip. Can I share a CCQA filter as a URL? Yes. Filter state is stored in the URL (?range=, ?dateFrom=, ?dateTo=, ?hourFrom=, ?hourTo=). Copy the URL to share the same view with a teammate.
What is TimeLens AI?
A streamed chronological narrative of your call dataset for the active filter window. Up to ~150 calls per segment. Pick the segment count with the slider, write an optional prompt, click Generate. Export as PDF or CSV. See TimeLens AI.
What are Compliance scripts?
Lists of phrases that agents must say (greeting, identity, privacy, verification). The matcher checks the call transcript against the phrase keywords / regex / intents. Per-script pass rate, per-agent pass rate, and per-call breakdowns. Edit the library via the Edit Library dialog. See Compliance.
Can I get alerted when something goes wrong?
Yes. The Alerts page has 6 seed rules covering sentiment, volume, handle time, and silence. Add custom threshold or rate-of-change rules. Notifications surface in the topbar bell and (for rules with email enabled) by email. See Alerts.
What does Volume Forecasting do?
Seasonal-naive forecast of incoming call volume by day-of-week. Configurable horizon (7 / 14 / 30 / 60 / 90 / 180 days), training window, and confidence band. Lives on its own tab; an Overview peek card surfaces it on the main dashboard. See Volume Forecasting.
Why are some columns showing dashes (—)?
A few features are gated on backend work that hasn’t shipped yet (e.g. FCR per agent depends on Customer Journey arriving first). The dash + tooltip explains the dependency. The data structure is in place; only the upstream signal is missing.
15. Chat with Your Documents
What can Chat with Your Documents do? Two engines, one chat surface. RAG answers questions over your uploaded documents with inline citations. SQL queries your read-only databases and returns auto-generated charts. See Chat with Documents Overview. What file types can I upload? PDF, DOCX, and XLSX today. The ingestion pipeline parses, chunks, and embeds the document; status flows from pending → parsing → chunking → embedding → ready (or failed). Can I connect a database? Yes. SQL Sources → add a connection (Postgres, MySQL, SQL Server, more). Connection details are encrypted, and your password is cleared from the form right after submit. The agent only runs read-only queries; any data-modifying SQL is blocked. See SQL Sources. Where do citations come from? Inline[1], [2] markers in the assistant’s reply that map to specific chunks in the source documents. Click any citation to open the chunk inspector.
Can I share a chat with people outside my workspace?
Yes. The Share Links feature creates a public link (with an optional password gate). Two modes: Chatbot (the visitor chats with the same agent) or Snapshot (read-only view of an existing conversation). Tenant admins can read back visitor sessions per share link. See Share Links.
How do I attach a file mid-conversation?
Drop or upload directly into the chat composer. The file goes into a hidden session-scoped knowledge base that’s tied to that single chat. The session-scoped KB does not appear in the regular Knowledge Bases list.
16. Languages, RTL, and localisation
Does Wittify support Arabic and other RTL languages? Yes. The interface mirrors fully when the URL prefix is/ar/... and uses IBM Plex Sans Arabic. RTL switches at every level: sidebar moves to the end edge, menus open from the end edge, sliders inverse direction, charts respect logical-reading order.
How many languages can be active at once for an agent?
Up to 10. Each can have its own greeting, banner, voice texts, rating prompt, and chips.
Can I set different dialects of the same language?
Yes. State the dialect explicitly in Core Instructions (e.g., “Speak in Saudi Najdi only”). For voice agents, also pick a voice that’s native to that dialect.
Which Saudi dialects are supported?
Najdi, Hejazi, Eastern Saudi, plus White Arabic (modern neutral). All work in text and voice modes.
Which other Arabic dialects?
Emirati, Qatari, Kuwaiti, Bahraini, Omani, Egyptian (Masri), Lebanese, Syrian, Jordanian, Palestinian, Moroccan (Darija), Tunisian, Algerian, Libyan, Sudanese, Yemeni (San’ani / Ta’izzi / Hadhrami). Plus Modern Standard Arabic for formal contexts.
Can one agent handle multiple dialects in one conversation?
Yes. The agent detects the user’s dialect and mirrors it, switches between dialects when serving users from different regions, and uses Modern Standard Arabic for formal or mixed audiences.
Why does the voice not sound fully native in some dialects?
Some dialects have subtle regional variations harder to replicate at scale. Pick a native-dialect voice from the Voice Picker, or commission a custom voice clone for the best fidelity.
17. Compliance, security, and troubleshooting
Which policies must my agents comply with? Your organisation’s policies, our Terms of Service, Prohibited Use Policy, and Privacy Policy, plus any industry-specific rules (HIPAA, GDPR, PDPL, etc.). How do I improve response speed?- Keep Core Instructions concise (under 1,500 characters where possible).
- Add FAQs for common questions (FAQs are faster than retrieval).
- Limit Knowledge Base to trusted, stable sources; remove old material.
- For voice, shorten responses and tune timing controls.
- Tighten Restrictions / Compliance rules in Core Instructions.
- Add FAQs for sensitive or frequently asked questions.
- Remove outdated KB sources.
- Lower Creativity Level to 0.3 or 0.4 for fact-heavy deployments.
18. Common questions after reading the FAQ
I read everything and still can’t find my issue. Open the floating chat widget in the bottom-end corner of the dashboard for live support, or book a free onboarding session. Can I download all the docs as a single PDF? Not today. The Mintlify CLI exposes per-page export; reach out if you need a bulk export. Where do I report a bug? Use the floating chat widget or email support. Include the URL of the page where you saw the issue and a screenshot if possible. The bug report goes to the engineering team directly. Can I request a feature? Yes. Same channels (chat widget or support email). Many features in the docs were originally customer requests; the product team reads every one. Why does the docs site show a banner asking me to switch language? The banner appears when your browser language disagrees with the URL prefix. Click to switch, or dismiss it; the preference is saved. My screenshot is out of date compared to the actual UI. Screenshots are added gradually. The text reference is always more authoritative than a stale image. If a discrepancy materially confuses you, ping support and we’ll update the screenshot.💡 Need help getting started? Book a free onboarding session with one of our Customer Success Engineers and get personalized guidance on setting up your Wittify products. 📅 Book here: Schedule a Free Onboarding Session