stabletoolbench/MirrorAPI-Cache

TEXT GENERATIONConcurrency Cost:1Model Size:7.6BQuant:FP8Ctx Length:32kPublished:Feb 27, 2025License:mitArchitecture:Transformer Open Weights Cold

stabletoolbench/MirrorAPI-Cache is a 7.6 billion parameter language model fine-tuned from StableToolBench-MirrorAPI, specifically designed to function as an API server. It excels at understanding API documentation and generating JSON-formatted responses that accurately reflect an API's intended output based on given inputs. This model is optimized for API simulation and response generation, supporting both standard SFT and Chain of Thought reasoning modes for enhanced API mechanism inference.

Loading preview...

MirrorAPI-Cache: API Simulation and Response Generation

stabletoolbench/MirrorAPI-Cache is a 7.6 billion parameter model, fine-tuned from the StableToolBench-MirrorAPI base, to act as an intelligent API server. Its core capability lies in interpreting API documentation and generating precise, JSON-formatted API responses based on specific input requests. This model is trained on dedicated datasets, train_cache.json and test_cache.json, focusing on API interaction patterns.

Key Capabilities

  • API Response Generation: Accurately crafts JSON responses that align with an API's intended functionality, even with varied input parameters.
  • API Mechanism Inference: Supports a Chain of Thought (CoT) mode, allowing the model to infer the underlying mechanism of an API before generating a response, providing deeper insights.
  • Flexible Prompting: Utilizes distinct system prompts for standard Supervised Fine-Tuning (SFT) mode and CoT mode, guiding the model's behavior.
  • Structured Output: Ensures all responses adhere to a strict JSON schema, including fields for error and response content, and an additional mechanism_of_the_api field in CoT mode.

Good For

  • API Simulation: Ideal for developers needing to simulate API behavior for testing, development, or prototyping without a live backend.
  • Automated API Documentation Testing: Can be used to validate API documentation by generating expected outputs.
  • Tool-use and Agent Development: Provides a robust component for agents that need to interact with or understand various APIs.

Finetunes
1 models