> ## Documentation Index
> Fetch the complete documentation index at: https://docs.atoma.ai/llms.txt
> Use this file to discover all available pages before exploring further.

# On-Demand AI Services

> Confidential generative AI APIs with multi-modal support, RAG capabilities, and composable encryption for enterprise-grade privacy

## Confidential Generative AI APIs

Atoma's on-demand offerings provide immediate access to state-of-the-art generative AI models through privacy-preserving APIs that guarantee confidentiality for both system prompts and user data. Our platform leverages **composable encryption** technology to enable unprecedented levels of privacy protection, allowing service providers and end-users to maintain control over their respective sensitive information.

### Core Capabilities

**Multi-Modal AI Processing**

* Text generation and completion (e.g. gpt-oss, Llama, Mistral, DeepSeek, Qwen, Kimi models).
* Image generation and editing (Flux, Stable Diffusion, etc).
* Audio processing and speech synthesis.
* Video analysis and generation.
* Code generation and completion.

**Retrieval-Augmented Generation (RAG)**

* Private document embedding and indexing.
* Confidential knowledge base querying.
* Secure vector database operations.
* Private semantic search capabilities.

**High-Performance Inference**

* Sub-second response times for most queries.
* Batch processing for large-scale operations.
* Real-time streaming responses.
* Auto-scaling based on demand.

## Composable Encryption: Dual Privacy Protection

Atoma's breakthrough **composable encryption** technology enables simultaneous protection of both system-level prompts (service provider IP) and user prompts (customer data), addressing a critical gap in current AI privacy solutions.

### How It Works

<div className="bg-gray-50 dark:bg-gray-900 p-6 rounded-lg border">
  <h4 className="text-lg font-semibold mb-4">Dual-Layer Privacy Architecture</h4>

  **Layer 1: System Prompt Protection**

  * Service providers' proprietary prompts, instructions, and system configurations remain encrypted
  * AI service companies can protect their prompt engineering IP and competitive advantages
  * System-level context and fine-tuning parameters are never exposed

  **Layer 2: User Data Protection**

  * End-user prompts, documents, and responses are encrypted with user-controlled keys
  * Customer data remains confidential from both Atoma and the service provider
  * The combination of advanced cryptographic primitives and trusted execution ensure only the user can decrypt their data
</div>

### Real-World Use Case: Cursor IDE

Consider how Cursor IDE could leverage Atoma's composable encryption:

**Cursor's Protected Assets:**

* Proprietary system prompts that make Cursor's AI unique
* Custom code analysis algorithms and instructions
* Specialized programming assistance methodologies
* Training data and model fine-tuning parameters

**Developer's Protected Assets:**

* Private repository code and intellectual property
* Confidential business logic and algorithms
* Sensitive customer data within codebases
* Personal coding patterns and preferences

**Result:** Cursor can offer AI-powered development assistance while protecting both their competitive AI technology and their customers' proprietary code.

## Enterprise-Grade API Features

### OpenAI-Compatible Interface

```bash theme={null}
# Standard OpenAI API call with Atoma's confidential computing
curl https://api.atoma.network/v1/chat/completions \
  -H "Content-Type: application/json" \
  -H "Authorization: Bearer $ATOMA_API_KEY" \
  -d '{
    "model": "gpt-oss-120b",
    "messages": [
      {"role": "system", "content": "Your protected system prompt"},
      {"role": "user", "content": "User query with private data"}
    ],
    "confidential": true,
    "user_encryption_key": "user_public_key"
  }'
```

### Advanced Privacy Controls

**Encryption Key Management**

* User-controlled encryption keys for data sovereignty.
* Hardware Security Module (HSM) key storage.
* Automatic key rotation and versioning.
* Multi-party key escrow options.

**Audit and Compliance**

* Comprehensive audit logs with cryptographic integrity.
* Close to real-time attestation of secure execution.
* Compliance reporting for GDPR, HIPAA, SOX.

**Access Controls**

* Fine-grained permission management.
* Role-based access to different encryption layers.
* Temporary access tokens with automatic expiration.
* Multi-factor authentication for sensitive operations.