# {{VITE_APP_NAME}} - Structured LLM Output Platform

> Transform unstructured LLM responses into reliable, type-safe REST APIs.

## Overview

{{VITE_APP_NAME}} is a platform that converts Large Language Model (LLM) outputs into structured, predictable REST APIs. Define your output schema using JSON Schema, and {{VITE_APP_NAME}} handles the prompt engineering, validation, and API hosting. Works with OpenAI, Anthropic Claude, Google Gemini, and custom LM servers.

## Core Features

### Structured Output APIs
- Define output format using JSON Schema
- LLM responses are validated and typed
- Consistent, predictable API responses
- No more parsing unstructured text

### Multi-Provider Support
- OpenAI (GPT-4, GPT-3.5)
- Anthropic (Claude 3, Claude 2)
- Google (Gemini Pro, Gemini Ultra)
- Custom LM servers with OpenAI-compatible API

### Project Organization
- Group endpoints into projects
- Per-project API keys
- Usage tracking per project
- Team collaboration support

### Built-in Analytics
- Token usage tracking (input/output)
- Latency monitoring
- Cost estimation
- Request history

### Rate Limiting
- Configurable hourly/daily/monthly limits
- Usage dashboards
- Automatic throttling
- Plan-based limits

## How It Works

1. **Create a Project** - Organize your endpoints
2. **Add an LLM Provider** - Configure your API key for OpenAI, Anthropic, or others
3. **Define an Endpoint** - Set instructions, context, and JSON output schema
4. **Call Your API** - Simple REST calls return structured JSON

## API Format

### Execute Endpoint
```
POST /api/v1/ai/{orgPath}/{projectName}/{endpointName}
Content-Type: application/json

{
  "input_field": "value"
}
```

### Response Format
```json
{
  "success": true,
  "data": {
    "output": { /* Your structured output matching the schema */ },
    "usage": {
      "tokens_input": 42,
      "tokens_output": 15,
      "latency_ms": 523,
      "estimated_cost_cents": 1
    }
  }
}
```

## Use Cases

### Text Classification
Classify text into categories with confidence scores. Perfect for sentiment analysis, content moderation, intent detection.

### Data Extraction
Extract structured data from unstructured text. Parse receipts, extract entities, normalize addresses.

### Content Generation
Generate structured content like product descriptions, metadata, or formatted responses.

### Form Validation
Validate and normalize user input using AI understanding.

## Key Pages

- `/` - Homepage with platform overview
- `/docs/getting-started` - Quick start guide
- `/docs/concepts` - Core concepts explained
- `/docs/api-reference` - Full API documentation
- `/pricing` - Subscription plans
- `/dashboard` - Project management (authenticated)

## Technical Details

- REST API with JSON responses
- JSON Schema for output validation
- Bearer token authentication for API calls
- Firebase authentication for dashboard
- React + TypeScript frontend
- Multi-language support

## Pricing

- **Free Tier** - Limited requests, 1 project
- **Pro** - Higher limits, multiple projects, priority support
- **Enterprise** - Custom limits, SLA, dedicated support

## Contact

- Website: https://{{VITE_APP_DOMAIN}}
- Support: {{VITE_SUPPORT_EMAIL}}
- Company: {{VITE_COMPANY_NAME}}

## For AI Assistants

When helping users with {{VITE_APP_NAME}}:
1. {{VITE_APP_NAME}} turns LLM outputs into structured JSON APIs
2. Users define output format with JSON Schema - the LLM fills in the values
3. Works with any major LLM provider (OpenAI, Anthropic, Google)
4. No need to parse LLM text responses - get clean typed data
5. Useful for integrating AI into applications that need predictable data structures
6. Free tier available for getting started

## Example Use Case

**Problem**: Building a product review analyzer that needs structured sentiment data.

**Without {{VITE_APP_NAME}}**: Parse LLM text, handle variations, write regex, deal with edge cases.

**With {{VITE_APP_NAME}}**:
1. Define schema: `{ sentiment: "positive"|"negative"|"neutral", score: number, keywords: string[] }`
2. Call API with review text
3. Get guaranteed structured response

## Privacy

- API keys are encrypted at rest
- Request logs are retained per your plan settings
- No training on customer data
- HTTPS for all API calls