API Reference and Documentation

This documentation provides a detailed overview of the API endpoints, input parameters, expected outputs, and usage examples. It also explains the underlying concepts and architecture, with Mermaid diagrams where necessary.

API Overview

Our API allows you to create LangChain tools on-the-fly by generating solution sketches, source code, and unit tests. The API is built upon a powerful AI model that understands natural language and code, making it easy for developers to describe the functionality they want and get the relevant code and tests in return.

Endpoints

Create Solution Sketch

POST /api/v1/solution-sketch

Generate a solution sketch based on a provided tool description.

Input Parameters:

Parameter	Type	Description
tool_description_prompt	String	A brief description of the functionality desired in the Tool.

Output:

A string containing the generated solution sketch for the LangChain tool.

Generate Tool Source Code

POST /api/v1/tool-source

Generate the source code of a LangChain Tool based on the provided solution sketch and tool tests.

Input Parameters:

Parameter	Type	Description
solution_sketch_n_tool_tests	String	A string made of two substrings separated by '######SPLIT_TOKEN########', representing the solution sketch and tool tests code block.

Output:

A success message along with the generated source code of the LangChain tool.

Generate Tool Unit Tests

POST /api/v1/tool-tests

Generate the unit tests for a LangChain Tool based on the provided solution sketch.

Input Parameters:

Parameter	Type	Description
solution_sketch	String	The solution sketch of the functionality desired in the Tool.

Output:

A success message along with the generated source code of the LangChain tool's unit tests.

Examples

Example 1: Creating a solution sketch

Request:

jsonCopy codePOST /api/v1/solution-sketch
{
  "tool_description_prompt": "Create a tool that calculates the factorial of a given number."
}

Response:

jsonCopy code{
  "solution_sketch": "To create a LangChain tool that calculates the factorial of a given number, you'll need to implement a recursive function..."
}

Example 2: Generating tool source code

Request:

jsonCopy codePOST /api/v1/tool-source
{
  "solution_sketch_n_tool_tests": "Solution sketch...######SPLIT_TOKEN########Tool tests code block..."
}

Response:

jsonCopy code{
  "message": "The file FactorialCalculator.py has been written in the session_path successfully!",
  "source_code": "class FactorialCalculator(BaseModel):..."
}

Example 3: Generating tool unit tests

Request:

jsonCopy codePOST /api/v1/tool-tests
{
  "solution_sketch": "Solution sketch..."
}

Response:

jsonCopy code{
  "message": "The file test_FactorialCalculator.py has been written in the session_path successfully!",
  "unit_tests_code": "def test_factorial_calculator():..."
}

Concepts and Architecture

The API is built on the following components:

1. AI Model: A powerful AI model trained to understand natural language and code, making it easy for developers to describe the functionality they want in a tool.

2. Endpoints: The API provides endpoints for creating solution sketches, generating source code, and generating unit tests for LangChain tools.

3. LLMChain: The LLMChain (Language Model Layered Model Chain) is a mechanism that allows chaining together multiple prompts and language models, allowing the AI to provide context-aware and detailed responses.

Diagrams

sequenceDiagram
  participant User
  participant API
  participant AI Model
  participant LLMChain
  User->>API: Send request to API
  API->>AI Model: Forward request to AI Model
  AI Model->>LLMChain: Use LLMChain for context-aware response
  LLMChain->>AI Model: Return response
  AI Model->>API: Return generated code or tests
  API->>User: Return output to the user

This diagram demonstrates the flow of information between the user, API, AI model, and LLMChain. When the user sends a request to the API, it is forwarded to the AI model, which uses the LLMChain mechanism for context-aware responses. The AI model then returns the generated code or tests to the API, which sends the output back to the user.