What is EazyRAG?

EazyRAG is a Dead simple API for Retrieval Augmented Generation, In simple words its like Algolia for Retrieval Augmented Generation. Our API overcomes intricate vector DB challenges, enabling accurate answer generation and seamless citations.

How to use EazyRAG?

We have three simple endpoints for you to use.

1. Add/Index Content

2. Generate Answer

3. Retrieve relevant chunks (Soon)

Here is a example of how to use our API.

1. Add/Index Content

curl --location 'https://api.eazyrag.com/v1/add' \
--header 'Authorization: sk-<Your token>' \
--header 'Content-Type: application/json' \
--data '{
 "contents": [
   {
       "content":"Bun 1.0 was released on September 8, 2023. (This could be the entire document)",
       "title":"Bun",
       "meta": {
         "id": "1",
         "author":"Jarred Sumner"
       }
   }
 ],
 "collection": "example"
}'

👇 Output

{
    "message":"Success",
    "suggestion":null
}

Parameters in table below

Parameter	Type	Description
contents	array	Array of content objects
collection	string	Name of the collection
content	string	Content of the document, You can add entire document in one string :)
title	string	Title of the document (optional)
meta	object	Metadata of the document (optional): You can add any keys inside the metadata that are useful for filtering in the answer endpoint

2. Chat (Browser supported)

var myHeaders = new Headers();
myHeaders.append("Authorization", "ck-< Your client key>");

var requestOptions = {
  method: 'GET',
  headers: myHeaders, 
};

fetch("https://api.eazyrag.com/v1/chat?collection=example&chat_session_uuid=<Your new UuidV4() uuid (Optional) >&user=When did bun 1.0 released?", requestOptions)
  .then(response => response.text())
  .then(result => console.log(result))
  .catch(error => console.log('error', error));

👇 Output

{
  "answer": "Bun was released on September 8, 2023 [1].",
  "citations": [
    {
      "citation_number": 1,
      "meta": {
        "type": "title", // If a title is present, then the entire document is added as context into llm.
        "parent_uuid": "47ea1047-3c65-427d-89d7-194f7cada753",
        "id": "1",
        "text_content": "Bun",
        "author": "Jarred Sumner",
        "chunk_uuid": "de181f04-cef7-44b5-8e32-aa2c2ff5201d",
        "parent_token_count": 19
      },
      "relevance": 0.88854927
    }
  ],
  "chat_session_uuid": "68555f82-08d5-4269-be99-0a23d99d1fcd"
}

Parameters in table below

Parameter	Type	Description
collection	string	Name of the collection
chat_session_uuid	string	Optional, If you want to continue the chat session, you can pass the chat_session_uuid, If you don't pass the chat_session_uuid, then a new chat session will be created.
user	string	User question
stream	boolean	Optional, If true, the answer will be streamed as soon as it is available. The stream will end with chunk "[DONE]"

3. Answer

curl --location 'https://api.eazyrag.com/v1/answer' \
--header 'Authorization: <Your token>' \
--header 'Content-Type: application/json' \
--data '{
    "query":"When was Bun released?",
    "collection": "example"
}'

👇 Output

{
  "answer": "Bun was released on September 8, 2023 [1].",
  "citations": [
    {
      "citation_number": 1,
      "meta": {
        "type": "title", // If a title is present, then the entire document is added as context into llm.
        "parent_uuid": "47ea1047-3c65-427d-89d7-194f7cada753",
        "id": "1",
        "text_content": "Bun",
        "author": "Jarred Sumner",
        "chunk_uuid": "de181f04-cef7-44b5-8e32-aa2c2ff5201d",
        "parent_token_count": 19
      },
      "relevance": 0.88854927
    }
  ]
}

Answer with filter

curl --location 'https://api.eazyrag.com/v1/answer' \
--header 'Authorization: <Your token>' \
--header 'Content-Type: application/json' \
--data '{
    "query":"When was Bun released?",
    "collection": "example",
    "filter":{
       "id":{
           "$match":"1"
         }
     }
}'

Answer with stream

curl --location 'https://api.eazyrag.com/v1/answer' \
--header 'Authorization: <Your token>' \
--header 'Content-Type: application/json' \
--data '{
    "query":"When was Bun released?",
    "collection": "example",
    "stream":true
}'

Answer generation in browser

We also support answer generation in the browser, but you should use a client-side key which starts with 'ck-'. You can find your client-side key in your dashboard. https://eazyrag.com/dashboard/tokens

var myHeaders = new Headers();
myHeaders.append("Authorization", "ck-<Your client side key>");

var requestOptions = {
  method: 'GET',
  headers: myHeaders,
};

fetch("https://api.eazyrag.com/v1/answer?query=What is bun?&collection=bun_docs&stream=false", requestOptions)
  .then(response => response.text())
  .then(result => console.log(result))
  .catch(error => console.log('error', error));

Parameter	Type	Description
query	string	User question
collection	string	Name of the collection
filter	object	Filter the chunks or documents based on metadata. You can use any key from the metadata that you already used while indexing the content (Optional)
stream	boolean	If true, the answer will be streamed as soon as it is available (Optional). The stream will end with chunk "[DONE]"
$match	string number string[] number[]	Match the value of the key

3. Retrieve relevant chunks (Soon)

We are working on it, Stay tuned!