AIUNIFY PROS SUITE includes a dedicated Sidebar category titled:
Social Media auto reply
with:
The current Sidebar routes WhatsApp to /social/whatsapp, Instagram to /social/instagram, and identifies Facebook as a Coming Soon feature.
These systems are more than simple Prompt generators. They are persistent communication workspaces designed around connected messaging accounts, stored Agent settings, incoming-message logs, AI-generated replies, and subscriber-controlled automation behavior.
Although both automate customer communication, their current operating models are substantially different.
Provides:
Provides:
These workflows should not be treated as interchangeable.
Both Routes use the authenticated PROS SUITE session.
An unauthenticated user is redirected to:
/login.
Navigate to:
Sidebar → Social Media auto reply → WhatsApp Agent
Route:
/social/whatsapp
The workspace is titled:
WhatsApp Agent Control.
The Header displays:
When connected, the Header also displays:
Linked to session +[phone number].
When not connected:
“Connect your mobile device to enable automation and messaging.”
The current workspace has two main Tabs:
When WhatsApp settings load and the session is not connected, the client automatically switches the workspace to:
Agent Settings.
The current page also attempts to initiate connection automatically once settings have loaded and the status is disconnected.
Therefore, a subscriber opening an unlinked WhatsApp Agent can be moved directly toward the QR connection workflow.
WhatsApp Agent Settings contains three internal sections:
The first section contains:
WhatsApp Session Control.
The current UI recognizes states including:
and displays them as:
When disconnected, select:
Start Connection.
While initializing:
Connecting...
may be displayed.
During startup, the interface displays:
Initializing Socket
and:
“Generating secure WhatsApp session handshake. Please wait...”
When the session requires linking, a dynamic:
WhatsApp Web QR Code
is displayed.
The interface instructs:
Open WhatsApp on your mobile phone → Settings → Linked Devices → Link a Device.
Unlike Instagram's credentials workflow, WhatsApp Agent does not ask the subscriber to manually enter account credentials.
The account is linked through the QR session.
After connection, the page displays:
Successfully Connected.
The connected session can display:
+[phone number]
below Successfully Connected.
The connected-state message says:
“Your AI agent is running successfully in the background and logging interactions.”
WhatsApp Agent uses a live EventSource connection to receive:
without requiring a full page refresh.
If that live EventSource connection fails, the current client closes it and attempts to reconnect approximately:
6 seconds later.
When a message event is received, the page refreshes:
When connected, select:
Unlink Account.
WhatsApp asks:
Unlink WhatsApp Account
and:
“Are you sure you want to disconnect? This will log out and unlink your WhatsApp account, stopping all auto-responses.”
Select:
Cancel
to leave the connection intact.
Select:
Unlink Account
to proceed.
The current success notification is:
Disconnected
with:
“WhatsApp account unlinked successfully.”
Select:
Active Chats
to operate the conversation workspace.
The left side is titled:
Conversations.
The client requests its conversation list from:
/api/social/whatsapp/conversations.
This establishes server-backed conversation history rather than browser-only local chat state.
The Search field says:
Search by name or number...
The current filter checks both:
If none match:
No active conversations found.
with:
“Click the plus icon above to start chatting manually with a client.”
Each conversation can show:
If the latest stored entry has an AI response, the list can display:
AI: [reply].
A manual outbound entry is shown as:
You: [reply].
Select the Plus icon titled:
New Manual Chat.
The field says:
Enter phone with country code (e.g. 1415...).
Enter the number and select:
Start.
When starting a new Chat, the current client removes non-numeric characters before using the number.
If no usable digits remain:
Invalid Number
with:
“Please enter digits only.”
The Start action selects or creates a local conversation entry.
The actual outbound WhatsApp Message is not sent until you use the Message composer.
A new locally created conversation can appear with the temporary Name:
New Contact.
Select a contact from the left Conversations list.
The center pane loads that contact's Message history from:
/api/social/whatsapp/messages?contact=[contact].
The Header displays the selected contact's phone number and current reply mode.
If reply behavior is auto, the Header shows:
AI active
and:
AI Auto-reply On.
If the Agent is configured for manual replies:
Manual operator
and:
Manual Mode
are displayed.
If no conversation is selected:
No Chat Selected
with:
“Select a contact from the conversations sidebar or start a new manual chat session using the plus icon.”
If the selected Chat contains no stored Messages:
No messages recorded in this chat.
and:
“Send a manual reply below to begin the session history.”
Incoming customer Messages appear on the left side of the conversation.
The contact's stored Name can appear above the Message.
AI-generated replies appear as outbound Messages using the AI styling and an:
AI
Badge.
Subscriber/operator replies are visually distinguished and receive a:
Manual
Badge.
An AI reply can show its Token usage inside the Message, for example:
AI (-5T).
Failures can be shown as centered red system Messages.
At the bottom of a selected Chat is:
Type a manual reply here...
The composer and Send action are disabled if the WhatsApp session is not connected.
Manual outbound Messages are sent through the WhatsApp session and recorded in the account's WhatsApp logs.
The current WhatsApp manager records manual outbound Messages with:
tokens_consumed = 0.
This distinguishes an operator-written Message from an AI-generated reply.
If a manual Message cannot be transmitted:
Message Delivery Failed
is displayed with the returned Error or:
Failed to transmit manual message.
On larger screens, the right side shows:
Contact Details
and conversation statistics.
Total
counts all current stored Message items loaded for that conversation.
AI Replies
counts Message items classified as auto_reply.
Tokens Spent
adds the tokens_consumed values from the current loaded Messages.
A Badge displays:
Manual: [number].
The details panel also displays:
Active Persona Info.
It can show:
Open:
Agent Settings → AI Character Persona.
The Card is titled:
AI Persona Identity
with:
“Control AI Model parameters, character prompting, tone variables, response delay, and toggle rules.”
The Model selector is labeled:
AI Model Orchestrator.
The WhatsApp Agent stores its own model_id.
Changing the general PROS SUITE Header Model should not be treated as the way to change this Agent.
Use the Model selector inside AI Character Persona.
The dropdown is populated from PROS SUITE's shared:
AVAILABLE_MODELS
collection.
Because that collection can change with the application, the exact live list should be read from the dropdown rather than hardcoded into this manual.
The page's initial fallback value is:
gemini-2.5-flash.
Stored Agent Settings can override that initial value when loaded.
The second selector is:
Reply Automation Mode.
Choose:
AI Auto-Reply Enabled
to permit automatic AI responses when other requirements are also satisfied.
Choose:
Manual Operator Control Only
to prevent AI auto-replies.
When the Agent is active and the incoming contact passes any whitelist rule, Manual Mode records the incoming Message but does not generate an AI reply.
You can then use the WhatsApp conversation composer to respond yourself.
WhatsApp provides a free-text:
Tone
field.
Example placeholder:
friendly, casual, sarcastic...
A second free-text field is:
Personality Type.
Example:
helpful assistant, agent...
WhatsApp provides:
Response Delay (Seconds).
The current numeric field specifies:
min=0.
No maximum is imposed by this visible input.
The WhatsApp reply manager waits the configured number of seconds after generating the response and before transmitting it.
The most important behavior field is:
System Instruction Prompt.
The current placeholder is:
“Describe who the agent is, how they behave, and what information they can supply.”
A strong System Instruction can define:
A business-focused Prompt could say:
“You are the AIUNIFY customer-engagement assistant. Answer questions about AIUNIFY business software clearly and concisely. Ask prospects what type of business they operate and what systems they need. Never invent pricing or contractual terms. If you do not know an answer, tell the customer a team member will follow up.”
The current incoming-message reply manager directly uses:
system_promptmodel_idresponse_delayfor actual auto-reply processing.
The inspected manager does not directly append the separate Tone and Personality fields to the AI request.
Therefore:
Put mission-critical behavioral instructions directly into the System Instruction Prompt rather than relying only on the Tone and Personality fields.
Select:
Save Configuration.
Successful persistence produces:
Settings Saved
with:
“WhatsApp AI agent configurations have been successfully updated.”
Open:
Whitelisted Numbers
to access:
Target Whitelist Manager.
There are two reply-target options:
Choose this when AI automation should be allowed for every incoming contact, subject to the other Agent requirements.
Choose this when AI automation should apply only to approved phone numbers.
The field says:
Add Whitelist Number (digits only, e.g. 14155552671).
Enter the complete phone number as digits, including the appropriate country code, matching the interface example.
Pressing Enter while focused on the Whitelist field also invokes the add-number action.
If a number already exists:
Already added
with:
“This number is already in the whitelisted targets.”
Select the X on a stored Whitelist Badge to remove it from the current list.
If Select mode is active but there are no numbers:
“No numbers whitelisted. The AI agent will not auto-reply to anyone.”
After modifying targeting, select:
Save Target Rules.
The WhatsApp reply manager checks the Whitelist before it reaches the normal AI/manual logging branch.
When Select mode is enabled and an incoming number does not match, the manager returns without creating the standard reply log for that Message.
Therefore:
Do not expect the Agent's conversation logs to act as a complete inbox for non-whitelisted contacts when selective targeting is enabled.
The Settings area states:
Limits & Safety
and:
“Avoid bulk notifications or spam. Whitelisting targets is recommended to test new configurations safely.”
The current reply-processing source requires that:
If the Agent is inactive or not configured, the current manager does not generate the auto-reply.
If manual mode is enabled, the incoming Message is logged with zero AI Tokens and no AI response is sent.
If the account does not have enough Tokens for the auto-reply, the Message is logged as failed with:
Insufficient token balance.
The feature catalogue currently identifies:
WhatsApp AI Agent — 5 Tokens.
The reply manager retrieves the configured feature cost and uses:
5
as its fallback.
For a successful AI auto-reply, Tokens are deducted after the AI response is generated and sent.
Manual sends are recorded with zero consumed Tokens.
The current WhatsApp reply manager sends the current incoming Message together with the configured System Prompt and selected Model.
It does not source-verify including the full historical conversation log in each AI-generation request.
Therefore:
The current auto-reply engine should not be documented as having persistent multi-turn conversational memory simply because the UI displays historical Messages.
The UI can display persistent conversation history.
That does not mean the AI automatically receives that history as context for its next reply.
Put important business context into the System Prompt so each reply has the core information it needs.
Navigate to:
Sidebar → Social Media auto reply → Instagram Agent
Route:
/social/instagram
The workspace is:
Instagram Agent Control.
The Header displays:
Status: @[username] (Active)
when the stored Agent is active.
Otherwise:
Status: Offline.
Current Tabs:
The initial Tab is:
Scheduler.
Select:
Settings
to configure the Instagram Agent.
The Settings area is titled:
Agent Profile & Config
with:
“Configure AI personality variables, models, response behaviors, and access tokens.”
After making changes, select:
Save Agent Config.
The current success message is:
Settings Saved
with:
“Instagram Agent settings have been successfully updated.”
The connection Card is titled:
Instagram Integration Credentials.
Required field:
Instagram Handle (Username)
with:
e.g. yourbrand_official.
Required field:
Instagram Business Account ID.
Required field:
Facebook Page Access Token.
The field is rendered as a password input.
The current user interface does not show a:
Sign in with Instagram
or:
Continue with Facebook
OAuth button.
Subscribers currently enter the required integration credentials directly into the form.
Select:
Connect Account.
After a successful connection response, the current message is:
Connected
with:
“Successfully linked to Instagram account @[username].”
When a Business Account ID is stored, a:
Disconnect
button is available.
Current success:
Disconnected
with:
“Instagram account unlinked successfully.”
The sanitized User Operations bundle contains the control surface and Instagram reply-processing manager, but the /api/social/instagram/... Route implementations used by the page were not included in this bundle.
Therefore, this manual can verify:
but it does not claim to independently audit every external Meta credential-validation, webhook-subscription, or publishing step.
The first Settings Card is:
AI Persona Variables.
Instagram Agent has its own:
AI Model
selector.
Like WhatsApp, the current dropdown is populated from:
AVAILABLE_MODELS.
The initial local fallback is:
gemini-2.5-flash.
Stored Settings replace it when loaded.
Use the Model in:
Instagram Agent → Settings
rather than assuming the global PROS SUITE Header Model controls Instagram replies.
Instagram provides a fixed Tone selector.
Current options:
Initial default:
friendly.
Instagram also provides a fixed:
Personality
selector.
Current options:
The page initializes its local personality state to:
professional
before stored Settings are loaded.
However, professional is not one of the five visible Personality dropdown values shown in the current component.
This is a current source inconsistency.
Stored Settings can replace that initial value.
Instagram also exposes:
System Instruction Prompt.
The included Instagram reply manager directly uses:
system_promptmodel_idwhen generating DM and Comment replies.
The inspected manager does not directly append the separate Tone and Personality values to those AI requests.
Therefore:
Put critical reply behavior, business rules, and voice requirements directly into System Instruction Prompt.
Instagram provides:
Response Delay (seconds).
The slider is currently:
0–30 seconds.
The reply manager waits the configured delay before completing an automatic DM or Comment response.
Settings includes:
Automation Event Rules.
The first rule is:
DM Auto-Reply
with:
“Respond to incoming Instagram Direct Messages.”
Choices:
The second rule is:
Comment Auto-Reply
with:
“Respond to public comments on your media posts.”
Choices:
You can configure, for example:
DMs → AI Auto
while:
Comments → Manual
or the reverse.
Instagram uses separate DM and Comment behavior settings.
The included reply manager records incoming DMs or Comments when their respective behavior is Manual but does not generate a response.
Unlike WhatsApp, the current Instagram Agent page does not provide a manual outbound DM composer or manual Comment-reply form for live operator responses.
Manual mode is therefore source-verified as:
Disable the AI auto-response and retain the item for review/logging
—not as a complete in-app human-response workflow.
Select:
DMs Auto-Reply.
The left pane is titled:
Active DMs.
The Search field says:
Search user....
The page groups stored DM logs by:
sender_username
and orders conversations by their latest logged timestamp.
When logs exist and no conversation is selected, the page selects the first conversation automatically.
If none exist:
“No simulated chats found. Send a message on the simulator to start!”
The center pane displays:
An outbound AI DM can display:
AI (-[tokens]T).
The DM conversation Header currently displays:
AI Reply Auto-active.
This text is static in the current component.
Because DMs can actually be configured to Manual, do not use this Header text as the authoritative automation setting.
Use:
Settings → DM Auto-Reply
instead.
The right-side panel is:
DM Sandbox Simulator.
The interface explains:
“Test your AI auto-replies instantly by sending simulated direct messages as a customer.”
Field:
Simulator Client Name (Username)
with:
e.g. fashion_influencer.
The simulator converts its username field to lowercase and removes characters outside:
Field:
Incoming DM Text
with:
“Type message as the client...”.
Select:
Send Message as Customer.
During the simulated request:
AI Processing...
is displayed.
Current notification:
Simulated message processed
with:
“Check the conversation bubble for response.”
The page states:
“Clicking send hits the simulator API, generates a response with your prompt, charges 5 tokens, and updates logs.”
The underlying reply manager retrieves the Instagram Agent feature cost and uses:
5 Tokens
as fallback for a successful AI response.
Manual/inactive processing can log without generating an AI reply and without charging those AI-response Tokens.
The feature catalogue lists:
Instagram AI Agent — 5 Tokens.
Before generating an automatic DM response, the manager checks whether the subscriber has sufficient balance.
When balance is insufficient, the DM is logged as failed with:
Insufficient token balance.
The current manager sends:
to the configured Model.
Historical DM logs are not added to the current reply-generation request in the inspected manager.
Therefore, DM conversation display and AI conversational memory should not be treated as the same feature.
Select:
Comment Reply.
The left pane begins with:
Select Post.
The Comment workspace currently lists only Posts whose system status is:
published.
If none are available:
“No published posts found. Go to Scheduler to publish a post first.”
After selecting a Post, the center pane displays:
Comment Thread.
Each stored Comment can display:
Automatic Comment responses display:
AI Auto-reply.
The reply can display:
-[tokens]T.
When no AI reply is present and the record is not an Error:
Manual review (AI reply offline).
A failed reply can display:
Auto-reply failed: [error].
The right pane is:
Comments Sandbox Simulator.
The interface states:
“Simulate incoming user comments on your published posts to test the AI's automated thread responses.”
The simulator requires a selected published Post.
Without one:
“Please select a published post first to enable comment simulations.”
Field:
Simulator Commenter Username
with:
e.g. food_blogger.
Field:
Comment Text
with:
“Write a comment on the selected post...”.
Select:
Simulate Comment Reply.
During processing:
AI Replying...
Current message:
Comment reply simulated
with:
“AI auto-reply added to comment logs.”
The current Instagram reply manager instructs the AI to respond to a Comment in:
1–2 sentences
and to keep it:
The AI receives:
for the Comment reply.
The inspected Comment reply function does not source-verify sending previous Comment-thread Messages into each generation.
Successful automatic Comment replies use the Instagram Agent feature cost, currently falling back to:
5 Tokens.
Instagram Agent includes a complete:
Scheduler
workspace.
The Header is:
Scheduled Post Queue
with:
“Create, monitor, and publish scheduled images and reels.”
Empty state:
No posts scheduled
and:
“Get started by clicking the ‘Schedule Post’ button and scheduling media content.”
Select:
Schedule Post.
The modal is titled:
Schedule New Instagram Post
with:
“Queue an image or reel with a caption and date-time.”
The current Scheduler supports three sources:
The default media-source mode is:
Manual URL.
The field is labeled:
Media Image URL *.
Despite the “Image URL” label, the helper says:
“Use a public Unsplash/web image or video URL.”
Therefore, the current Manual URL workflow is intended to accept public media URLs for either Images or Videos.
The Scheduler modal does not expose a local:
Upload Image
or:
Upload Video
control.
Manual scheduling uses a URL.
Select:
AI Image
as the Media Content Source.
Field:
Image Generation Prompt *.
Select:
Produce with AI Generator.
The Scheduler's AI Image workflow currently uses fixed:
The Scheduler modal does not expose Image:
controls.
Its Image generation currently uses 1:1 automatically.
Current message:
AI Production Started
with:
“Media is rendering. Please wait...”
The Scheduler checks AI Image status approximately every:
5 seconds.
Success:
AI Image Ready
with:
“Visual successfully generated and attached to schedule.”
Select:
AI Reel/Video.
The Prompt field changes to:
Video/Reel Vision Prompt *.
The current Scheduler-generated Reel uses fixed settings:
The Scheduler does not expose controls for changing those fixed Reel parameters.
For more detailed Reel configuration, use the dedicated Reels Generator described in Chapter 27.
When completed:
AI Reel Ready
with:
“Video successfully generated and attached to schedule.”
During AI media production, the modal displays:
AI Generator Active.
The interface says:
“This can take 15-30 seconds.”
This should be treated as interface guidance, not a guaranteed processing duration.
Once AI media is ready, the modal displays:
Generated Media Preview.
The X on the Generated Media Preview removes that media from the current scheduling form.
The current action clears the Scheduler's selected URLs.
It is not source-verified as deleting the underlying Marketing Asset record.
Closing the modal clears the Scheduler's browser polling interval.
This does not source-verify cancellation of the already-started server-side AI media task.
The scheduling form includes:
Post Caption.
The current client requires media and publication time, but does not require Caption.
The placeholder states:
“Write caption (emojis and hashtags are supported)...”
Required:
Publication Date & Time *.
The browser uses a datetime-local input.
If media or scheduled time is absent:
Missing Info
with:
“Please specify media content (or generate with AI) and publication time.”
Select:
Queue Schedule.
While submitting:
Scheduling...
Success:
Post Scheduled
with:
“Post queued successfully.”
A Scheduler Card can display:
Current statuses include:
A scheduled Post displays:
Schedule: [date/time].
A published Post displays:
Published: [date/time].
If a Post contains an Error Message, that Error appears on the Card.
A Trash action appears only for Posts whose status is:
scheduled.
Current success:
Post Canceled
with:
“Post has been removed from the schedule.”
The Delete/Cancel icon is rendered only for scheduled entries.
The current Scheduler Cards do not provide:
To change a scheduled item, the current user-facing approach is to cancel it and create another schedule entry.
The Scheduler Header includes:
Run Cron.
This user-facing control calls the Scheduler processing endpoint to process due Posts.
Successful execution produces:
Scheduler Executed
with:
“Processed [number] due posts. Check logs for status updates.”
“Cron” is a technical scheduler term, but it is currently exposed directly to subscribers.
Operationally, treat it as:
Run the due-post processing cycle now.
The existence of Run Cron does not itself prove that subscribers must manually run it for every Post.
The API/Scheduler infrastructure responsible for automatic production execution is outside the sanitized Route bundle.
Use Run Cron as a visible manual processing/testing control rather than assuming it replaces all scheduled background execution.
The current catalogue identifies:
Instagram AI Agent — 5 Tokens.
A successful automatic DM response uses the configured feature cost, with a current fallback of:
5 Tokens.
A successful automatic Comment response similarly uses the Instagram Agent feature cost.
Incoming DMs or Comments processed under Manual behavior are logged without an AI response and with zero AI-reply Tokens in the current manager.
The AI Image and AI Reel/Video creation features inside Scheduler use:
/api/marketing
rather than the Instagram Agent's DM/Comment reply function.
Therefore:
Do not assume the Instagram Agent's 5-Token auto-reply cost is also the complete cost of generating Scheduler Images or Reels.
Those are separate Marketing-generation operations.
The Scheduler modal does not show a fixed media-generation Token price.
Instagram Settings store:
model_id
separately from the global PROS SUITE Model selection.
The included Instagram reply manager uses server/environment provider credentials for its selected Model pathways rather than receiving the subscriber's standard BYOK headers from the Agent page.
Accordingly, the subscriber's general API-Key settings should not be described as the direct configuration mechanism for Instagram Agent auto-replies in the current source.
Both Agents maintain historical logs in their interfaces.
However, the included auto-reply managers generate a response from the current incoming Message plus configured instructions/model, rather than source-verifiably feeding the whole visible conversation into every request.
Because persistent conversational memory is not source-verified, instructions and customer flows should not depend heavily on the AI automatically understanding references such as:
“What about the other one?”
without enough current context.
Include the important reusable information the Agent needs.
Avoid entering:
into conversational Agent Prompts.
The Sidebar currently displays:
Facebook Agent
under Social Media auto reply.
The Facebook entry is explicitly marked:
isComingSoon: true.
No matching app/social/facebook/page.tsx source was found in the inspected subscriber source bundle, while the Sidebar explicitly identifies the entry as Coming Soon.
Therefore this chapter does not invent:
for the current version.
Before enabling AI Auto-Reply broadly:
The interface itself recommends Whitelisting as a safe testing method.
A practical rollout pattern is:
Only whitelist selected numbers → Add a controlled test number → Verify responses → Expand carefully.
A poorly configured System Prompt applied to all inbound conversations can create undesirable customer responses.
Tell the Agent when it should avoid guessing.
Example:
“If a prospect asks for custom enterprise pricing, tell them an AIUNIFY representative will follow up rather than inventing a quote.”
For a sales-oriented Agent, instruct it to ask useful questions such as:
A small delay may make automated replies feel less instantaneous.
Long delays can negatively affect responsiveness.
Use conversation-level:
Tokens Spent
to understand AI usage for individual WhatsApp interactions.
Switch to Manual Operator mode for interactions requiring:
DMs and Comments often require different communication styles.
For example:
More conversational and qualification-oriented.
Short, public, brand-safe responses.
Use:
before trusting automatic customer-facing behavior.
The DM simulator explicitly states that its generated AI response is charged as Agent usage.
The current backend specifically prompts for 1–2 sentence Comment responses.
For example:
“Never discuss confidential account information in public Comments. Invite the customer to DM us when account-specific help is needed.”
When generating AI media inside the Scheduler:
As with other AI Image/Reel Tools, visually verify:
The Scheduler uses a local browser datetime-local field.
Always confirm the displayed intended date/time before Queue Schedule.
A failed scheduled Post can show its Error Message in the queue.
Do not assume every queued item was successfully published.
Go to:
Agent Settings → Link & Status
and start/review the connection.
The session can pass through:
Starting / Initializing Socket
before the QR state appears.
On your phone:
WhatsApp → Settings → Linked Devices → Link a Device
and scan it.
Check:
The backend checks these conditions before replying.
If Select targeting is enabled, non-matching contacts are skipped before normal reply processing.
The UI explicitly warns:
“The AI agent will not auto-reply to anyone.”
The Manager logs:
Insufficient token balance
when the account cannot cover an AI reply.
Manual sending requires:
Enter digits only, including country code.
The page uses live EventSource updates and automatically attempts reconnect after an EventSource Error.
If necessary, re-enter/reload the workspace so conversations and Messages are fetched again.
Go to:
Settings → Instagram Integration Credentials
and review:
The current UI displays:
Connection Failed
with the returned Error.
Because the integration API Route is outside the sanitized bundle, use the returned Error as the primary current diagnostic.
The conversation Header text is static.
Use:
Settings → DM Auto-Reply
as the actual current configuration.
That is a current UI limitation.
Manual mode logs the incoming DM but does not provide a manual-send box on this page.
Likewise, there is no in-page human Comment-response composer in the current source.
The Comments pane only lists system Posts whose status is published.
Check the Scheduler queue.
Select a published Post first and enter Comment Text.
The embedded Scheduler AI Image workflow is currently fixed to:
1:1.
Use the dedicated Image Generator when you need more ratio choices.
The embedded Scheduler Reel generation is fixed to:
Use the dedicated Reels Generator for more control.
The current manual source input is a URL.
No direct local file-upload control is source-verified.
No Edit/Reschedule action exists on current Post Cards.
Cancel the scheduled item and create a new one when necessary.
Review the Error shown on its Post Card.
The control reports the number of due Posts it processed.
A value of zero can simply mean there were no currently due Posts for that processing cycle.
| CapabilityWhatsApp AgentInstagram Agent | ||
| Authenticated PROS page | Yes | Yes |
| Account connection | QR / Linked Devices | Business ID + Page Access Token |
| Persistent inbound logs | Yes | Yes |
| AI auto-reply | Yes | Yes |
| Manual-mode logging | Yes | Yes |
| Manual outbound reply inside Agent | Yes | No current composer |
| New manual conversation | Yes | No |
| DM auto-reply | WhatsApp Messages | Yes |
| Public Comment auto-reply | No | Yes |
| Reply Whitelist | Yes | No |
| Response Delay | Yes | Yes |
| Agent-specific Model | Yes | Yes |
| System Instruction Prompt | Yes | Yes |
| Tone | Free text | Preset selector |
| Personality | Free text | Preset selector |
| Scheduler | No | Yes |
| AI Image scheduling | No | Yes |
| AI Reel scheduling | No | Yes |
| Sandbox DM simulator | No dedicated simulator | Yes |
| Comment simulator | No | Yes |
| Conversation Token statistics | Yes | Reply-level badges/logs |
| Current Agent fallback cost | 5 Tokens | 5 Tokens |
| ContextCurrent AI Input Source | |
| WhatsApp Reply | System Prompt + current incoming Message |
| Instagram DM Reply | System Prompt + current incoming DM |
| Instagram Comment Reply | System Prompt + Post Caption + current Comment |
The inspected reply managers do not establish full historical-conversation context on every generation.
The current source verifies:
The current source does not establish:
The current source verifies:
The current source does not establish:
Current Sidebar state:
Facebook Agent — Coming Soon.
No operational subscriber workflow should be documented until a future source version confirms one.
AIUNIFY PROS SUITE currently provides two operationally distinct Social Media Auto Reply systems:
WhatsApp Agent and Instagram Agent.
WhatsApp Agent at:
/social/whatsapp
uses a QR-based account-linking workflow. Subscribers connect their WhatsApp account through Linked Devices, after which the workspace can maintain persistent Conversations, receive live updates, generate AI replies, and allow operator-written manual responses.
Its AI configuration includes:
and it adds a separate Target Whitelist Manager allowing automation for either every incoming conversation or only selected phone numbers.
The current reply manager applies the Whitelist, checks Auto/Manual behavior, validates Token balance, generates the response, applies any configured delay, sends the WhatsApp Message, deducts the Agent usage cost, and logs the result. The current feature cost is 5 Tokens per successful AI auto-reply; manual operator Messages are logged with zero AI Tokens.
Instagram Agent at:
/social/instagram
uses a different setup requiring:
and provides separate controls for DM Auto-Reply and Comment Auto-Reply, each of which can be configured for AI Auto or Manual.
Instagram's DM and Comment workspaces also provide Sandbox Simulators so a subscriber can test the Agent's responses before relying on live customer interactions. The Instagram Agent catalogue cost is currently 5 Tokens, and the included reply manager uses that feature cost for successful AI DM and Comment replies.
One major difference from WhatsApp is that Instagram Manual mode currently disables the AI reply and records the inbound item, but the inspected Instagram page does not provide an in-app manual DM or Comment response composer.
Instagram Agent also contains a Scheduler. Subscribers can queue media using:
plus a Caption and Publication Date & Time.
The embedded Scheduler's AI Image generator currently uses fixed 1:1 output, while its AI Reel/Video generator uses fixed 9:16, 480p, 6 seconds, Normal mode, with an internally generated Seed. Those embedded settings are not subscriber-adjustable inside Scheduler.
Scheduled Posts can display:
statuses, and scheduled Posts can be canceled before publication. The visible Run Cron control manually runs a due-post processing cycle and reports how many due Posts were processed.
A second important limitation applies to both active Agents: their user interfaces maintain historical conversation logs, but the included reply managers do not source-verify sending complete prior conversation history into every AI reply. Current reply generation is based primarily on the configured System Prompt and the current incoming communication, with Instagram Comments additionally receiving the Post Caption.
For that reason, the recommended operating strategy is:
Configure the System Instruction Prompt Carefully → Start in Manual or Whitelist-Limited Testing → Select the Correct Agent Model → Test Typical Customer Questions → Test Unknown/Sensitive Questions → Verify Token Usage → Confirm Escalation Rules → Enable AI Auto-Reply Gradually → Review Conversation Logs Regularly → Move Sensitive or High-Value Conversations to Human Handling When Appropriate
For Instagram:
Connect Account → Configure DM and Comment Behavior Separately → Test Both Simulators → Generate or Supply Scheduled Media → Verify Every Caption and Creative → Queue Schedule → Confirm Published/Failed Status → Review DM and Comment Logs
The third Sidebar entry, Facebook Agent, remains explicitly marked Coming Soon and should not yet be treated as an operational subscriber system.