Typical reasons for stopping a dialogue

Sometimes during testing or when the bot is already running, there are moments when it stops responding. Below we will look at the main errors that are displayed when unloading an export.

Reasons for stopping dialogs

1. Dialog stops due to a control phrase

Why it occurs: If you have the "Interrupt by control phrase" setting enabled:

Then if the client writes this phrase exactly in the dialogue, the bot will be interrupted. In this case, the following will appear in the logs:

How to prevent: If this situation does happen, go to the "History" section, then find the dialog in which the interruption occurred and click the "Clear context" setting.

2. Dialogue stopped due to interception by an employee

Why it occurs: If the flag "Do not respond when a person interrupts the dialogue" is checked under the Main Instruction:

then if the manager interferes with the dialogue, it will automatically stop.

How to prevent: Find this dialogue in the "History" section, then click the "Clear context" setting. You can also set a flag with the setting as in the screenshot. Then after a certain time, the bot will return to the chat and respond to messages.

3. Stopping the dialog when the file is triggered

Why it occurs: If a file from the knowledge base was triggered, in which the "Interrupt dialog" flag was set, then in this case, if the context occurs when this file should be called, an interruption is triggered. It looks like this in the export:

How to prevent: Find this dialog in the "History" section, then click the "Clear context" setting.

4.Stopped due to reason: bot_kicked_from_chat

Why it occurs: Most often occurs when there is a conflict with the settings of another service, for example, when setting up Open Lines in Bitrix24. For example, the bot does not have access to the chat to which the employee has connected in turn, and accordingly, it is excluded from the chat. In the export, we see the message:

How to prevent: Check the settings on the integration side. For example, if we are talking about Bitrix, then you need to check the Open Lines settings. It is possible that there is a conflict between connecting the manager from the queue and excluding the bot from the dialogue.

Errors

1.Internal error

Why it occurs: This is an internal error on the Savvi service side. It looks like this in the export: How to prevent: Perhaps this is an isolated case within one dialogue, then intervention is not required. You can also contact Savvy's technical support: https://t.me/Suvvy_customer_support_en_bot

2.Dialogue stopped due to an error: Two calls to the same function with the same arguments

Why it occurs: It can occur, for example, if the bot's work scenario is configured incorrectly:

In this example, in the instructions, the bot, on the one hand, should have immediately called this function when qualifying the client, and on the other hand, the first call should have happened earlier, before the bot's response message about clarifying the record, so the instructions overlapped. How to prevent: Carefully analyze the scenario settings, are there any contradictions between the settings on Savvy's side: for example

3.Error: "Reached the maximum number of function calls per dialogue"

Why it occurs: Most often, such an error occurs when the limit on the number of function calls in dialogues is exceeded. The thing is that Savvi has a limit on calling each function within a single dialog, it is the same for any function and equals 10. Accordingly, if the bot is configured so that according to the scenario it must contact different services multiple times in response to a client's request, the dialog stops and we see in the logs:

How to prevent: In this case, you need to look at how to optimize the use of various integrations for your scenario so that the bot does not need to make multiple calls within a single request.

4.Error: "The provider could not process the request"

Why it occurs: This is an internal error on the LLM provider side. It may occur due to exceeding the character limit in the request. For example, to convert the result from the table, the bot requires 128847 tokens in AI, while the maximum limit is 128,000. The error occurs:

How to prevent: In this example, the client should optimize the table, for example, make two tables, but with less data in each of them. In general, if this error occurs, contact Savvy technical support: https://t.me/Suvvy_customer_support_en_bot

5.Error when requesting LLM

Why it occurs: This is an internal error when accessing LLM. For example, it may occur if the connection was blocked due to a limit. Or the model has limitations. How to prevent: Most often, you need to contact Savvy technical support: https://t.me/Suvvy_customer_support_en_bot

6.Dialogue stopped for an unknown reason

Why it occurs: Occurs when the cause of the error is not determined. How to prevent: In this situation, the client should contact technical support at the link: https://t.me/Suvvy_customer_support_en_bot