Beta Documentation

API Documentation

Learn how to use our Process and Forget RAG system API to build privacy-focused applications.

Getting Started

The Process and Forget Paradigm

Our API is built on a fundamental privacy principle: process and forget. This means:

Document content is processed, vectorized, and immediately discarded
Only semantic embeddings are stored, not the original content
You provide your own API credentials with each request
No sensitive data is ever persisted on our servers

Note: To use our API, you'll need to be approved for the beta program. Contact us at beta@augmentedretrieval.com to request access.

Core Concepts

Bring Your Own Credentials

You provide your own API keys for OpenAI and your vector database (Pinecone or Supabase) with each request. We never store your credentials on our servers.

Custom Indexes

Query multiple knowledge sources through a single API call by specifying custom indexes and controlling result counts.

API Reference

Batch Document Storage Endpoints

POST

/api/v1/chroma/store

Store multiple documents in ChromaDB with batch processing

{
  "chromadb": {
    "chroma_url": "http://your-chromadb-instance:8001"
  },
  "openai": {
    "api_key": "your-openai-api-key"
  },
  "collection_name": "test_collection",
  "documents": [
    {
      "id": "doc1",
      "content": "This is the content of the first document",
      "metadata": {
        "source": "user_manual",
        "category": "instructions"
      }
    },
    {
      "id": "doc2",
      "content": "This is the content of the second document with different information",
      "metadata": {
        "source": "faq",
        "category": "troubleshooting"
      }
    }
  ]
}

Response: A success message with the count of documents stored.

{
  "message": "Successfully stored 2 documents in collection test_collection"
}

Query Endpoints

POST

/api/v1/modular/query

Query the RAG system with your question and optional conversation history

{
  "query": {
    "question": "How do I add a file attachment?",
    "history": [ /* Optional conversation history */ ]
  },
  "sources_config": {
    "custom_indexes": "my_knowledge_base,documentation",
    "custom_indexes_top_k": 5
  },
  "credentials": {
    "openai": {
      "api_key": "your_openai_api_key"
    },
    /* Use either Pinecone OR Supabase credentials based on your setup */
    "pinecone": {
      "api_key": "your_pinecone_api_key",
      "environment": "your_pinecone_environment"
    }
    /* OR */
    /*
    "supabase": {
      "url": "your_supabase_url",
      "key": "your_supabase_service_role_key"
    }
    */
  }
}

Key Parameters:

query.questionThe user's question
query.historyOptional conversation history
sources_config.custom_indexesComma-separated list of indexes to query
credentialsYour OpenAI and Pinecone credentials

POST

/api/v1/modular/query-html

Same as query endpoint but returns HTML-formatted responses

This endpoint accepts the same parameters as the regular query endpoint but returns responses formatted in HTML for easier display in web applications.

Document Processing

POST

/api/v1/modular/documents/process

Process a document with instant deletion after vectorization

{
  "document": "Your document content or base64 encoded file",
  "metadata": {"title": "Important Document", "type": "policy"},
  "index_name": "my_knowledge_base",
  "credentials": {
    "openai": {"api_key": "your_openai_api_key"},
    /* Use either Pinecone OR Supabase credentials based on your setup */
    "pinecone": {"api_key": "your_pinecone_api_key", "environment": "your_env"}
    /* OR */
    /*
    "supabase": {
      "url": "your_supabase_url",
      "key": "your_supabase_service_role_key"
    }
    */
  }
}

Key Parameters:

documentDocument text or base64 encoded file
metadataDocument metadata (title, type, etc.)
index_namePinecone index to store vectors
credentialsYour OpenAI and Pinecone credentials

Data Ingestion

POST

/api/v1/modular/ingestion/documentation

Scrape and ingest multiple documentation URLs

{
  "urls": [
    "https://example.com/documentation/page1.html",
    "https://example.com/documentation/page2.html"
  ],
  "single_page": true,
  "max_pages": 100,
  "credentials": {
    "openai": {"api_key": "your_openai_api_key"},
    /* Use either Pinecone OR Supabase credentials based on your setup */
    "pinecone": {"api_key": "your_pinecone_api_key", "environment": "your_env"}
    /* OR */
    /*
    "supabase": {
      "url": "your_supabase_url",
      "key": "your_supabase_service_role_key"
    }
    */
  }
}

This endpoint automatically scrapes multiple documentation URLs, processes the content, and stores the embeddings in your vector database. The original content is immediately discarded after processing.

Credential Requirements

API Keys You'll Need

OpenAI API Key

Required for generating embeddings and retrieving context-aware responses.

Get from OpenAI

Pinecone API Key

Option 1: Use Pinecone for storing and retrieving vector embeddings.

Get from Pinecone

Supabase Credentials

Option 2: Use Supabase pgvector for storing and retrieving vector embeddings.

Get from Supabase

Note: Your API keys are never stored on our servers. They are used only for the duration of your request and immediately discarded afterward.

Ready to Get Started?

Our beta program is currently available to select partners. Request access to experience the power of our Process and Forget RAG system.