Quickstart

Quickstart

In this guide you will find information on getting started with HyperChat™ either as a web app user (via our Chat page (opens in a new tab)) or as an advanced user / developer (via our Platform (opens in a new tab) and APIs) to build data retrieval pipelines and/or using state-of-the-art LLM orchestration.

TL;DR for Developers

import os
from langchain_openai import ChatOpenAI
llm = ChatOpenAI(api_key=os.environ["HYPERBEE_API_KEY"], 
                 base_url="https://api.hyperbee.ai/v1/", # for rag: https://api-rag.hyperbee.ai/v1/
                 extra_body = {"optimization" : "premium"} # optimization: "premium", "auto" or "fast"
                 ) # add "namespace":"<COLLECTION_NAMESPACE_UUID>" to extra_body= for RAG
llm_input = "Will humans be able to live in outer space any time soon?"
messages = [("user", llm_input)]
ai_msg = llm.invoke(messages)
print(ai_msg.content)

Chat: Web App Users

You can access HyperChat™ via a web interface by visiting our Chat page at chat.hyperbee.ai (opens in a new tab). Sign up with your Google and/or GitHub credentials, register for a subscription plan and submit your first queries via the "Send a message" text box. Chat access from the web interface comes in 3 subscription plans: Free, Standard and Advanced, see the Chat Subscription Page (opens in a new tab) for more information.

Users with Standard and Advanced plans also have the option of selecting specific models within the HyperChat™ portfolio instead of using the orchestrator. Visit the Chat Subscription Page (opens in a new tab) for more information.

You can use HyperChat™ in three modes:

  • Fast: Optimize for the fastest response and lowest cost, without sacrificing significant accuracy.
  • Auto: Automatically optimize for the best performance, let HyperChat™ decide the best setting for your input.
  • Premium: Ensure highest quality, especially for workflows that require higher-level reasoning.

See our Performance Guide page on Optimization Modes (opens in a new tab) for more information about these three modes.

Users without a valid login or users on the Free subscription can only use the Fast setting and have a limits on chat length. We highly recommend creating an account to experience the full capabilities of HyperChat™.

To chat with your documents and to build high-quality retrieval pipelines, you need to create collections:

  1. Go to Collections: Click on Collections from the pop-up menu on the chat screen
  2. Uplaod Documents: Drag-and-drop, or browse to select the documents you want to upload and click Submit to start the upload
  3. Get Confirmation: Once the uplod completes, indexing starts. Collection indexing takes some time, sit back and relax. You will receive a confirmation e-mail once the Collection is added to your index.
  4. Start Chatting: Once you receive the confirmation e-mai, you can attach and detach your indexed Collections to your chat at any time by clicking on the "Use Collection" box on the interface.

Platform: Playground and API

If you want to dive deeper into how HyperChat™ operates and benchmark its performance, or if you want to integrate HyperChat™ to your product or workflow programmatically with API access, the Platform (opens in a new tab) is the right place.

The HyperChat™ Playground (opens in a new tab) allows you to test the performance of HyperChat™ head-to-head against services from leading LLM providers, providing an instant benchmark in terms of accuracy, cost and latency.

To create an API key, head to "API Keys" from the pop-up menu on the left and press + Create Key on the top right. Please note that for security reasons HyperbeeAI cannot help you out with a lost API key, but you can always create new ones and deactivate lost ones. Credits for all API keys are tied to your account and are processed together.

The credit costs for all models and orchestrator modes can be viewed on the main Platform (opens in a new tab) page, and billing management (including payment methods, tracking remaining credits and order history) are presented on the Platform Billing (opens in a new tab) page.

HyperChat™ has two APIs: Chat Completion (api.hyperbee.ai) and RAG (api-rag.hyperbee.ai), where the Chat Completion API answers queries solely on the user inputs, and the RAG API answers queries grounded by information in Collections. The templates for both APIs are drop-in replacements for the OpenAI API template, so integration is extremely simple for all popular LLMOps tools. See Integration (opens in a new tab) for more information.

First Run: Chat Completion API

Once you have your API key ready, you can run your first query with the chat completion API:

curl --location 'https://api.hyperbee.ai/v1/chat/completions' \
--header 'Content-Type: application/json' \
--header "Authorization: Bearer $HYPERBEE_API_KEY" \
--data ' {
   "model": "hyperchat",
   "optimization":"fast",
   "stream":"false",
   "messages":[
   {
       "role": "user",
       "content":"Was Rome built in a day?"
   }]
}'

Just make sure you define $HYPERBEE_API_KEY as an environment variable containing your API key.

First Run: RAG API

Collections can be created from both the Chat interface and the Platform interface, and the Collections of a user account are shared between Platform and Chat interfaces, so you can create a Collection in Platform to use it in Chat too. Programmatic Collection management (Collection creation, uploading files, deleting files, ...) is done through the Python Client.

curl --location 'https://api-rag.hyperbee.ai/v1/chat/completions' \
--header 'Content-Type: application/json' \
--header "Authorization: Bearer $HYPERBEE_API_KEY" \
--data '{
 "namespace": "<COLLECTION_NAMESPACE_ID>",
 "messages": [
   {"role": "user", "content": "Was Rome built in a day?"}
 ],
 "model": "hyperchat",
 "optimization": "premium"
}'

Just make sure you define $HYPERBEE_API_KEY as an environment variable containing your API key, and that <COLLECTION_NAMESPACE_ID> equals the Collection you want to use in your chat. The Collection namespace ID can be obtained by clicking the three-dot menu icon of the specific Collection either on the Platform Collections page and pressing Copy ID. Collection IDs follow the Version 4 UUID format.

HyperChat™ is always improving, so ask the librarian@hyperbee.ai if you can't find something you need!

About HyperChat™Models ↗