The persona determines who the account is and how it speaks; the schedule determines when it is online; the mantra determines its language habits. After the three are configured, the performance of the same account across sessions and days will be consistent, and it will not be polite one day and blunt the next.
Same account, yesterday’s reply and today’s reply are completely different from the same person
Personal settings, routines, and mantras are stored in the library and will take effect long-term after being bound to the account. The AI follows the same settings every time it replies, not edited every time.
When someone took over the operation, the tone changed, and customers could see it at a glance.
The way of speaking is written in the persona, not in someone's head. Whoever takes over will not change the account’s tone, address habits, and common expressions.
Dozens of accounts use the same vocabulary, and it looks like they were sent in batches
The characters can be allocated according to the number, and the set of mantras can be different for each. The same batch of content appears in different language habits on different accounts.
The automatic reply was still sent out in the middle of the night. It was three o'clock in the morning in the other party's area.
The schedule is set according to the region to which the account belongs. Messages in inactive periods are queued for processing in the next period and are not sent out late at night locally.
This system consists of three independent libraries plus a script layer. They are maintained separately because they have different change rhythms: the characters are rarely changed, but the dialogue is often changed.
Create a new character and write down the identity, background, tone and boundaries. Can be set as the default profile, which will be automatically applied when a new account is connected.
Create a new schedule and define active periods and rest periods every day. It can be set as the default schedule and configured according to the region where the account belongs.
Create a new collection of mantras, including the commonly used openings, modal particles, and methods of address for this character. Can be set as default to allow differences between batch accounts.
Write common dialogue processes into scripts: what to say under what circumstances, what conditions to advance to the next step, and when to switch to manual.
Dialogue scripts that run smoothly can be saved as task scripts and handed over to task planners for batch execution, eliminating the need to reorganize the process every time.
Fixed answers to high-frequency questions are saved as quick replies, which can be sent with one click during manual takeover. They can also be used as a priority reference for AI replies.
Characters and diction can be maintained separately by language, and existing content can also be translated. Specific supported languages—。
Generate a draft of words or content based on the persona and knowledge base, and then use it after you revise it. The generated results are not sent directly to the outside world.
You need to confirm these two steps, the person setting draft and the delivery of the script. The persona is a long-term setting for the account. Changing it once affects all subsequent conversations and should not be taken by the AI itself.
Identity, tone, boundaries between what can and cannot be said, and active periods of the day. You can change it from the default profile, or you can create it separately by group.
Personal settings, routines, and mantra sets are tied to specific accounts or account groupings. The same account uses the same settings for maintenance, publishing, and private conversations.
Generate responses or draft content using a combination of persona, mantra, and your knowledge base. Questions that hit quick replies are given priority with fixed answers.
Quotes, commitments, and responses involving compliance statements are waiting for you. The confirmed statement can be saved as a quick reply, and similar questions will be released directly later.
Replies are kept in the conversation history. See which types of questions you answered poorly, come back and change the boundaries, add a catchphrase, or add a quick reply.
Words involving offers and commitments must be confirmed by you. AI doesn't know how much profit you can make from this order, nor does it know which sentence constitutes a legal commitment in your market - it generates it based on the goal of consistent tone, not your business bottom line. Only you can draw this line yourself.
The character system itself is a configuration capability, and it is up to you to decide what content to write. Tell the truth about what you want to pay for and what presets are available.
Automatic replies and manual transfers are performed according to the persona and conversation script, and the tone remains consistent across conversations.
The persona decides how to say it, and the knowledge base decides what to say. Only by working together can the two become human-like and answer accurately.
Account maintenance actions are performed according to the same schedule to avoid mismatch between action time and subsequent conversation time.
Dialogue scripts can be saved as task scripts and distributed to a batch of accounts for batch execution by the arranger.
You can't count on that. Character constraints are tone, identity, and clearly written boundaries. It reduces the probability of deviation, but it does not equal zero errors. Customers' questions are ever-changing, and when encountering situations that are not covered by personas and knowledge bases, AI may still answer vaguely or even incorrectly. Therefore, responses to external commitments require your confirmation, it is recommended that high-frequency questions be fixed as quick replies, and conversation records must be checked regularly. Think of it as a new person who needs to be reviewed, rather than a system that is infallible.
Yes, you can bind in batches by group. However, multiple accounts on the same platform use exactly the same persona and mantra, and the way they speak will be highly similar, which can easily be seen as a batch operation. It is recommended to share character skeletons and separate mantra sets by number.
Does not leak. Messages received during inactive periods will be retained and processed in the next active period. If you want to respond to a certain type of message at any time, you can mark it in the conversation script and handle it separately.
Mantras are language habits that affect the words and tone used when AI is generated; quick replies are complete fixed answers, which can be used directly when hit. The former makes the reply more like the person, and the latter ensures that key questions are answered verbatim.
No. The conversation script manages how to progress in a conversation, and the task script manages the process by which a batch of accounts perform tasks. Dialogue scripts that run smoothly can be saved as task scripts and handed over to the editor for batch running.
Characters and diction can be maintained separately by language, or existing content can be translated and then manually revised. Machine translation is not always suitable for colloquial expressions and local idioms. It is recommended to ask a native speaker to translate important words.
The consultant will help you come up with your first personality, work schedule, and quick replies based on your business and target market.