DDG Provider - 429 ERR_INPUT_LIMIT

[BlyZeDev]

Bug description

{"error": {"message": "ResponseStatusError: Response 429: {\"action\":\"error\",\"status\":429,\"type\":\"ERR_INPUT_LIMIT\"}"}, "model": "gpt-4o-mini", "provider": "DDG"}

How can I find out the max conversation history character limit?

Environment

• Python 3.11.2
• Germany

[hlohaus]

Hey, you could check message lengths in Python like this, see? import json; print(f"Len: {len(json.dumps(messages))}")

@BlyZeDev

[BlyZeDev]

I'm calling the API from C#.
But I just need the length of all messages together.

Btw the issue is very inconsistent.
Sometimes it happens with small message sizes and sometimes it works with very big message sizes. Kinda weird.
I'll update to the newest version and send the length asap when the next error occurs.

[hlohaus]

@BlyZeDev It gives is an approximate number of the characters. However, models usually rely more on the number of tokens in their language model.

Please note that the models add and process all previous messages to the input length. If you do not want this, you must omit the old messages in your request. Or you start simply a new conversation.

[BlyZeDev]

I'm currently doing it like this. Without specifying conversation or conversation_id a new conversation is created?

var model = _config.GetAIChatModel();
var provider = _config.GetAIChatProvider();
HttpResponseMessage? response = null;
try
{
    var messages = new List<AIChatMessage>
    {
        new AIChatMessage
        {
            Role = "system",
            Content = SystemMessage
        }
    };
    
    await foreach (var message in history)
    {
        messages.Add(message);
    }

    messages.Add(new AIChatMessage
    {
        Role = "user",
        Content = prompt
    });
    
    var postJson = JsonConvert.SerializeObject(new AIChatRequest
    {
        Messages = messages,
        Model = model,
        Provider = provider,
        WebSearch = true,
        Temperature = 0.5f,
        HistoryDisabled = true,
        Stream = false,
        MaxTokens = 400,
        AutoContinue = true,
        Timeout = (int)(_http.Timeout.TotalSeconds * 0.75)
    });

[hlohaus]

Yes, some providers offer an internal conversation instance. Conversation reuse is facilitated by including the conversation_id and provider in API requests; the API subsequently manages the conversation. Alternatively, setting return_conversation=True enables the retrieval of the conversation instance in subsequent requests. @BlyZeDev

[BlyZeDev]

How do I know which providers offer this.
Can I just use conversation_id instead of passing the whole message context in, for providers that do support it?
return_conversation wouldn't work if I have multiple conversations running at the same time right?

[hlohaus]

Hey, I don't have a list of providers that support return_conversation. But you can check get_paramters:

if "return_conversation" in g4f.Provider.Gemini.get_paramters():
supports = True

Yeah, the RetryProvider doesn't support it yet.

@BlyZeDev

[github-actions]

Bumping this issue because it has been open for 7 days with no activity. Closing automatically in 7 days unless it becomes active again.