TLDR: Your chatbot now reliably sends the right product photo from your knowledge base, the chatbot editors show a live token count as you type, and Meta Direct API channels stay connected more reliably. Plus a few security and polish fixes.

Here's what's new 👇

🖼️ Reliable Product Photos in AI Replies

The problem: When a customer asked for a product or a photo, your chatbot sometimes replied "no image", sent the wrong picture, or made up a link, even when the real photo was sitting in your knowledge base.

• Your chatbot now reliably finds and sends the correct product photo from your knowledge base when a customer asks for it by name. ✨
• It follows your media instructions more consistently, so the right image goes out with the right product.

🔢 Live Token Counter While You Edit

The problem: When writing your Business Info or AI Behaviour, you couldn't tell how many tokens your text was using until later.

• The Business Info and AI Behaviour editors now show a live token count as you type, color coded green, yellow, or red, so you can keep your prompts lean and cost effective. 💪

🔌 Steadier Meta Direct API Channels

The problem: Channels connected through Meta Direct API (with a pasted token) were left out of two background jobs, so they could quietly stop receiving messages after a behind the scenes webhook change, and their token expiry wasn't being tracked.

• Direct API channels now stay subscribed to incoming messages through those updates, and their token expiry is tracked so you're prompted to reconnect before it lapses. 🔒

🛠 Bug Fixes

• Connecting an integration no longer crosses organizations.
• Token Usage tooltip now matches the number.
• Stronger login protection.

TLDR: Chatbots set up before the recent knowledge base upgrade can now get the new comparison answers with a one click Retrain, organization owners are no longer wrongly locked out of their own org, and we strengthened how your data stays private to your organization.

Here's what's new 👇

📚 Give Older Knowledge Bases the New Comparison Answers

The problem: Chatbots whose knowledge base was set up before the recent "grounded answers" update couldn't handle "cheapest", "under RM 50", or "how many" style questions, and clicking Retrain didn't fix it.

• Affected chatbots now show a banner prompting a one click
• Your existing answers keep working the whole time, so there's no gap while you retrain.

🔓 Organization Owners No Longer Locked Out

The problem: In some cases, an organization owner could be wrongly hit with an access error across their own organization (Device Manager, chatbots, contacts).

• Owners now always have access to their own organization. 💪
• We've also repaired the accounts that were already stuck in that state.

🛠 Bug Fixes

• Stronger data privacy between organizations.

TLDR: A big AI accuracy release. Your chatbot now answers from your knowledge base reliably (with a new Train step), the fact checker catches wrong answers even on long replies, and your WhatsApp channels no longer get wrongly disconnected during Meta hiccups.

Here's what's new 👇

📚 Grounded Answers From Your Knowledge Base

The problem: Even when the correct answer (a price, a spec) was sitting in your knowledge base, your chatbot could still give a wrong one. It treated knowledge base info as lower priority than training data, lost track on follow up questions, and couldn't compare across products.

• Your chatbot now treats your knowledge base as an equal, trusted source, so the most specific answer wins instead of being overridden. 🎯
• It handles follow up questions ("how heavy is it?") and multi part questions correctly, pulling from the right product's full document.
• It can now answer "cheapest", "lightest", or "under RM X" style questions by comparing across your whole catalogue with exact numbers.
• New

Your existing chatbots keep working right away. To get the full benefit of the new comparison answers, give your knowledge base a Train run.

✅ More Accurate AI Fact Checking

The problem: On longer replies, the AI fact checker could quietly fail and send the original, uncorrected answer while logging it as a clean pass, so wrong facts still slipped through.

• The fact checker now reliably corrects wrong facts even on long replies, and it will never ship a reply it couldn't verify. 🐛

🔌 Channels Stop Disconnecting on Meta Hiccups

The problem: When Meta's servers were slow or timing out, ReplyLa could wrongly mark a healthy WhatsApp channel as disconnected, and it would then stay stuck (some channels were stuck for weeks).

• A temporary Meta timeout no longer disconnects your channel. It's only marked disconnected on a genuine login or token problem, and otherwise heals itself on the next check. 💪

TLDR: WhatsApp templates are now checked against Meta's rules the moment you save, with clear inline errors, and keyword rules that reply with a template (buttons, media, and personalized fields) now send correctly.

Here's what's new 👇

🧩 Template Mistakes Caught Before You Submit

The problem: When you created a WhatsApp template, some mistakes (body too long, too many blank lines, an emoji in a button, too many buttons) weren't caught until Meta rejected it later with a slow, cryptic error. Saving as a draft or using the API skipped the checks entirely.

• ReplyLa now checks your template against Meta's rules the moment you save, on every path (including drafts and the API), with clear inline error messages so you fix it right away instead of waiting for a rejection. ✨
• Helpful warnings also flag things like link shorteners or promotional wording in a Utility template, without blocking you.
• Bilingual templates (Malay and English together) are handled correctly, so they won't get falsely flagged.

🔑 Keyword Template Replies Now Work Fully

The problem: Keyword rules that replied with a WhatsApp template were broken. Button templates failed at Meta, media templates didn't attach their image, personalized field mappings silently fell back to defaults, and rules with unfinished variables saved fine but failed later at send.

• Keyword rule template replies now send correctly with their buttons and media, and pull the right details (including your custom fields) straight from the contact. 💪
• The editor now catches unfinished template variables when you save, instead of letting them fail silently when the message goes out.

TLDR: Multi part AI replies now arrive on WhatsApp in the right order, each AI chat costs fewer tokens, your knowledge base gives sharper answers, and you can no longer get stuck in Playground after a plan change. Plus marketing template fixes and a few access and mobile touches.

Here's what's new 👇

💬 Replies Arrive In the Right Order

The problem: When your AI split a reply into several parts (like an image with a caption, then another), they sometimes landed out of order on the customer's actual WhatsApp, even though they looked correct in the Inbox and Playground.

• Multi part replies now arrive on WhatsApp in the exact order they were composed. 🎯
• If one part fails to send, the later parts of that same reply are held back too, so a customer never gets a caption without its image.
• Replies to different customers still send in parallel, so there's no slowdown.

🪙 Cheaper AI Chat

The problem: A recent change had quietly pushed up the token cost of each chat by reducing how much of the prompt could be reused between messages.

• Chats now reuse far more of the prompt from message to message, bringing the cost per reply back down. 💰
• Your chatbot's memory now updates in a single pass per turn instead of two, trimming token use even further.

📚 Sharper Knowledge Base Answers

The problem: Your chatbot's knowledge base used mismatched matching settings in different places, so it sometimes pulled weak or irrelevant matches, and other times didn't reach for the knowledge base when it should have.

• Knowledge base matching is now consistent everywhere, so your chatbot pulls the right information on catalogue, pricing, and policy questions, including when customers send images. ✨

🔄 No More Getting Stuck in Playground

The problem: If your plan was downgraded or expired while you were in Playground mode, you could get locked out with every Playground page blocked and no way back.

• You're now automatically moved back to Live mode when your plan no longer includes Playground, with a quick "Switched to Live mode" note. 💪

📣 Marketing Templates Route Correctly

The problem: Template categories were being saved in two different letter casings, so some marketing templates missed the cheaper marketing send route, and others skipped unsubscribe checks.

• Marketing templates now consistently use the correct, cheaper send route, and unsubscribe suppression always applies so opted out contacts stay excluded. 🔒

🛠 Bug Fixes

• Team members can see the API token page.
• Reply window details now work on touch.

TLDR: See your total AI spend and cost per lead at a glance, get accurate delivered counts in Blaster, and enjoy a smoother template experience with video previews and instant uploads.

Here's what's new 👇

💰 See Your Total Spend and Cost Per Lead

A new Total Spend card on the Token Usage page shows what you're actually paying.

• A Total Spend card now shows your cost in Ringgit, plus the number of leads and your cost per lead. 🎯
• It always matches whatever filters you've set, so the number you see reconciles with the usage table below.
• Your Usage Summary now defaults to your current billing cycle instead of the calendar month, so the figures line up with what you're billed.

💬 Your Inbox Remembers Your Channel

The problem: Every reload dropped you back to a default channel, so you had to reselect the one you were working in.

• Your inbox now remembers the channel (or playground bot) you last picked, even after a reload or switching organizations. ✨
• If that channel is gone or disconnected, it cleanly falls back to your first connected one.

📤 Blaster Improvements

• The "Messages delivered" card now shows how many messages were actually delivered, with a double tick icon, instead of just how many were sent. 🎯
• Freshly uploaded columns that are fully filled in no longer ask you for a back up value. You're only prompted for a fallback when a column actually has blanks. 💪

🎬 Template Improvements

The problem: A few template moments felt rough: videos showed a grey box, uploads vanished, and sync buttons piled up.

• Video header templates now play the actual video in the preview, everywhere you see it. ✨
• Files you upload now appear in the Select File modal right away, with their size shown, instead of needing a refresh to show up. ⚡
• The "Sync All" and "Sync Device" buttons are now one smart button that does the right thing for where you are.
• Template categories now display with consistent casing, and filtering catches them no matter how they're written.

✏️ Clearer Follow Up Conditions

• Follow up conditions are renamed from "Must Match" and "Must NOT Match" to "Include Conditions" and "Exclude Conditions", with a short helper line for each so it's clear what they do.

TLDR: Your AI Checker now actually fixes wrong facts in replies, the unsure flagging is far more accurate, and you can optionally hand a customer to a human when the bot isn't sure of the answer.

Here's what's new 👇

✨ Hand Off to a Human When the Bot Is Unsure

The problem: When your bot wasn't confident about a factual answer, there was no easy way to route that chat to a real person automatically.

• New optional toggle "Hand off to a human when unsure" on your AI Feedback settings page. Turn it on and an unsure factual reply alerts your team and passes the chat over. It is off by default, so nothing changes until you switch it on. 💪
• The toggle only shows up when "Alert Me on Unsure Answers" is already on, so the setting stays simple.

🛠 Bug Fixes

• Your AI Checker now genuinely corrects wrong facts. Before, it ran quietly but never fixed anything, so a reply quoting the wrong price or deposit would still go out. Now a contradicted fact like a price, amount, or business hours gets corrected before the customer sees it. 🎯
• The unsure flagging is much more accurate. Polite phrases like "sorry, let me check" used to get wrongly flagged, flooding your review queue. The bot is now judged on whether it actually answered the question, not how politely it phrased the reply. ✨
• Human takeover no longer hijacks a good reply. A clean, correct answer is no longer swapped out for the generic handoff message by mistake. 💬
• Your bot remembers more across a conversation. It now holds onto details a customer shares over several messages, instead of forgetting all but the last one. 💪

TLDR: Videos in your campaigns now play properly on Android phones, blasts that use personalised fields like the customer name or today's date no longer break, and contacts saved in two phone formats are now counted and charged just once.

Here's what's new 👇

📹 Your WhatsApp Videos Now Play on Android

The problem: Some videos showed up as "video not available" on Android phones, even though they played fine on iPhones. Slower videos could also fail quietly behind the scenes.

• Videos in your blasts and your inbox are now prepared in a format every Android phone can play. 🎯
• Larger videos get the extra time they need to process, so they go out instead of silently failing. ✨

💬 Personalised Blasts No Longer Break

The problem: When a campaign used a built in field like the customer name, a greeting, or today's date, the message went out blank and WhatsApp rejected it, which paused the whole campaign.

• These personalised fields now fill in correctly for every recipient. 💪
• If one recipient is missing a value, only that single message is skipped. Your campaign keeps running for everyone else.

💰 No More Double Counting or Double Charging

The problem: If the same person was saved twice in slightly different phone formats, they were counted twice in your audience, charged twice against your capacity, and messaged twice.

• Duplicate numbers are now matched and treated as one person, so your audience count, your costs, and your sends are all accurate. ⚡

TLDR: We removed a chatbot presence option that quietly did nothing, so the only choices left are the ones that actually work.

Here's what's new 👇

🧹 Only Presence Options That Actually Work

The problem: When setting up your chatbot, you could pick a "recording" presence option, but it never did anything. Your customers never saw a recording indicator, so the setting was misleading.

• The presence dropdown now shows just two clear choices: Typing and Disabled. ✨
• Any chatbot that was set to the old "recording" option now simply shows the typing indicator, and it loads and saves without any errors. 🐛

TLDR: API senders now get clear reasons when a message fails, big inboxes load without errors, and your Playground and web chat replies stay in the right order.

Here's what's new 👇

⚡ Clearer Reasons When an API Send Fails

The problem: When a message sent through the public API failed, you always got the same generic error. There was no way to tell a bad message from an expired login or a rate limit.

• Failed API sends now tell you the actual reason, like an invalid template, an expired channel, or being temporarily rate limited, so you can fix the right thing fast. 🎯

💬 Big Inboxes Load Cleanly Again

The problem: Inboxes with more than a thousand contacts could hit a loading error and stall, leaving channels and previews stuck.

• Large inboxes now open smoothly with channels and message previews loading right away, and live updates keep arriving. ✨

🛠 Bug Fixes

• Multi message replies in the Playground and your shared web chat link now appear in the correct order, greeting first, then options, then closing. The order stays correct even when you scroll back through history. 🐛