Hello 1min.AI Technical Support Team,
I am integrating the 1min.AI OpenAI-compatible API with multi file Code, an AI coding agent, using models such as Kimi K2.7 Code.
I have successfully tested the API connection, authentication, streaming and model generation. However, I have encountered a request-size limitation that prevents the integration from handling normal coding-agent workflows.
Your API currently imposes a 100,000-character limit across messages, tools, schemas and tool history.
In an actual multi file coding request, I measured:
Conversation messages: 74,014 characters
Tool definitions: 86,632 characters
Response format: 2 characters
Total: 160,648 characters (make it 300000 to be safe)
The API consequently returns:
"Messages, tools and response format exceed the 100,000 character limit."
I understand that your API currently supports up to 128 function definitions. My request contains 97 tools, so the problem is the combined character limit rather than the number of tools.
I would like to request one of the following two solutions:
Option 1 – Increase the request-size limit
Could you increase the maximum request size from 100,000 characters to at least 500,000 characters, ideally 1 million characters?
Alternatively, could you provide an increased limit for Business Advanced lifetime customers or make this configurable per account?
Option 2 – Introduce native batched or multi-part request support
If increasing the request limit is not possible, could you implement an API mechanism that allows clients to submit large requests in multiple batches?
For example:
The client submits the conversation and tool definitions in separate batches.
Each batch remains below the API request-size limit.
Your API temporarily assembles the batches into one complete request.
Once all batches have arrived, the complete request is passed to the selected AI model.
The model returns its response and any function/tool calls normally.
An alternative would be a native tool-discovery mechanism where the API can receive a large tool catalogue, identify relevant functions and forward only the required tools to the model.
The important requirement is that the model must retain the complete context it needs and that function calling must continue to work normally across multiple turns.
Additional questions:
Does your API currently offer any undocumented batch, multipart or request-assembly mechanism?
Is the 100,000-character limit imposed by your API gateway or the underlying model providers can it be made make it 300000 to be safe ?
Could you support higher request limits for AI coding agents?
Is native tool discovery or dynamic function selection on your development roadmap?
Can you confirm that Kimi K2.7 Code supports multi-step function calling and tool-result history through your API?
I would like to use 1min.AI as a long-term model provider within a full AI software development environment. Support for larger requests or batched requests would make a significant difference to this integration.
Thank you for forwarding this request to your API engineering team.
Kind regards
Ranjith