FIG API

Overview

The FIG API (Form Integrated Generation) is a powerful service that transforms simple form submissions into intelligent, personalized responses. By combining your organization's knowledge base with advanced AI, it creates contextually relevant replies that maintain your voice and expertise.

Features at a Glance:

• 🗄️ Knowledge Integration: Access your stored documents and prompts from managed FIG collections
• 🤖 AI-Powered: Leverages the Deepseeks V3 model for human-like understanding
• ✨ Smart Formatting: Automatically generates clean, consistent markdown responses
• 🎯 Context-Aware: Combines form data with your content for relevant, personalized replies
• 🔒 Secure: Simple API key authentication and isolated collections

Quick Start

curl -X POST https://incrediforms-api-642822023513.us-central1.run.app/ai/generate \
  -H "Content-Type: application/json" \
  -H "x-incredi-api-key: YOUR_API_KEY" \
  -d '{
    "submission": {"field1": "value1"},
    "fig_collection_id": "collection_id",
    "stream": false
  }'

Endpoint

POST https://incrediforms-api-642822023513.us-central1.run.app/ai/generate

Request Headers

The following headers are required for all API requests:

• Content-Type: application/json
• x-incredi-api-key: Your API key (Required)

Request Body

The request body should be a JSON object with the following fields:

• submission (Record<string,any>): Form submission values as key-value pairs (Required)
• fig_collection_id (string): The ID of the collection to use for the request (Required)
• stream (boolean): Whether to stream the response (Optional, default: false)

FIG collections are created in the IncrediForms dashboard, containing files and prompts.

Example Request

{
  "submission": {
    "field1": "value1",
    "field2": "value2"
  },
  "fig_collection_id": "collection_id",
  "stream": false
}

Response

Success Response

• Status Code: 200 OK
• Content Type: application/json

{
  "generative_response": "AI-generated markdown content based on the input",
  "model": "deepseek-chat",
  "usage_metadata": {
    "input_tokens": 668,
    "output_tokens": 847,
    "total_tokens": 1515,
    "input_token_details": {
      "cache_read": 640
    }
  }
}

Error Response

• Status Code: 500 Internal Server Error
• Content Type: application/json

{
  "status": "error",
  "message": "Error message details"
}

Technical Details

Key Features

• 🤖 Powered by Deepseeks V3 model
• 📝 Markdown-formatted responses
• 🛡️ Built-in error handling
• 📚 Context-aware processing

Technology Stack

• LangChain's ChatOpenAI integration
• ChatPromptTemplate for structured prompts

System Configuration

Base Prompt

I analyze form submissions and provide personalized responses based on the given instructions. All responses will be formatted in Markdown, with:

1. Read and process the submitted form data
2. Follow the provided response instructions carefully, which include:
   - The role I should adopt (e.g., "Act as an expert in...")
   - The specific goal of the response (e.g., "recommend the most suitable plan")
   - Any additional context or information provided
   - The desired tone and style of communication

3. Generate a response that:
   - Uses proper Markdown formatting including:
     * Headers with a single space after #
     * Code blocks with triple backticks when needed
     * Proper line spacing before and after lists
     * Consistent emphasis using asterisks
   - Directly addresses the form submitter
   - References their specific submission details
   - Maintains the specified expertise and perspective
   - Provides relevant, actionable recommendations

Best Practices

1. Keep form_prompts clear and specific
2. Make sure the collection exists in the IncrediForms dashboard & contains files and prompts
3. Structure submission data logically
4. Handle responses asynchronously
5. Implement proper error handling