Elara – Chat with anything

Elara AI – Chat with code
Elara AI – Chat with code

Welcome to Elara!

This site provides a comprehensive overview of Elara, its functionalities, underlying mechanisms, and future roadmap. We’ll delve into its features, detailed specifications, sales strategies, current challenges, achievements to date, and planned enhancements.

Please note: Elara is currently in a closed beta phase, accessible only to a select group for testing purposes.
We hope you’ll find this documentation informative and gain an appreciation for Elara’s potential.

Models

Elara models are the core functionality that defines what users can interact with. Currently, Elara offers various models, each enabling specific conversations:

  1. Files & Folders: Users can chat directly with any file or folder uploaded from their personal devices.
  2. Code: This model allows users to converse with codebases or individual code files. These can be uploaded from GitHub repositories or directly from their devices.
  3. CRM Integration (Current): Elara currently integrates with HubSpot for CRM functionality. Integrations with other popular options like Salesforce, Zendesk, Zoho CRM, and Microsoft CRM are underway.
  4. Project Management Integration (Future): Planned integrations include applications like Trello, Monday.com, ClickUp, and Jira.
  5. Customer Support Integration (Future): Future plans involve connecting applications like Zendesk Desk, Freshdesk, and ServiceNow.
  6. Collaboration Tools: This model allows users to directly upload files or folders from cloud storage platforms like Google Drive, Outlook, and Dropbox.

Detailed Description of all Models

Files and Folder Model

The Files and Folder model empowers users to have natural language conversations with their uploaded files and folders. Supported file formats include txt, pptx, pdf, .doc, and .docx.

Here’s how it works:

  1. Upload: Users initiate by uploading a file or folder from their personal devices.
  2. Content Storage: The uploaded content is securely stored within our database.
  3. Conversational AI: When a user interacts with the uploaded file or folder, a specialized algorithm analyzes the user’s question and the stored content. It then leverages a specific AI model (described later) to generate a relevant response.
  4. User Response: The AI model’s answer is then presented to the user.
Files and folder workflow
Files and folder workflow

The Files and Folder model offers additional functionalities:

  • Chat Management: Users can clear or delete past chat conversations.
  • Download Feature: Downloaded files and folders are conveniently packaged in a ZIP format.

Note: We’re actively expanding capabilities to support additional file formats, such as CSV and XLSX.

Code

The Code model in Elara allows seamless interaction with your code. Here’s how you can leverage it:

  • Upload Flexibility: Choose to upload individual code files or entire code folders directly from your laptop.
  • GitHub Integration: Connect your Elara account to your GitHub repository for effortless access and retrieval of code.
  • Language Agnostic: Elara’s Code model works with various programming languages, including Python, Java, C, and C++.
  • Powerful Features:
    • Debugging: Troubleshoot code issues and identify errors with ease.
    • Functionality Enhancement: Add new features to your codebase directly through natural language interactions.
    • Codebase Comprehension: Gain a deeper understanding of existing code, even if it’s not your own.
Code workflow
Code workflow

Please note: The download feature for code is currently under development and will be available in the future.

CRM

Elara empowers you to connect your Customer Relationship Management (CRM) tools directly with the platform. This allows you to have natural language conversations with your CRM data, enhancing workflow efficiency.

Current Integration:

  • HubSpot: Elara currently integrates seamlessly with HubSpot. Simply connect your HubSpot account using an API key (a video tutorial will be available soon). You can connect multiple HubSpot accounts and leverage Elara to chat and interact with your CRM data.

Upcoming Integrations (Planned):

Elara is actively expanding its CRM integration capabilities. We’re working on adding support for popular platforms like Salesforce, Zoho CRM, and Zendesk. Stay tuned for future updates!

Future Integrations:

We’re committed to continuously expanding our CRM integrations to cater to a wider range of user needs.

API Key Creation:

For each CRM platform, Elara will provide clear instructions and resources (likely video tutorials) on the website, guiding you through the process of generating the necessary API key.

Elara AI – Upload new code
Elara AI – Upload new code

Pricing

Elara will adopt a flexible pricing structure. Organizations can subscribe to specific models that align with their unique requirements. This allows for a cost-effective approach, as you only pay for the functionalities you need.

For instance, an organization that primarily works with documents can choose the File & Folder model subscription and only pay for that on a monthly basis.

As for how much each model should be priced for that will be discussed with all the teammates. Just let us know.

User

The cornerstone of any successful application is its user base. To ensure Elara’s impact, let’s delve into the customer segment we’re targeting.

Target Audience

Elara focuses on Business-to-Business (B2B) customers who rely on various software tools in their daily operations. These include:

  • Customer Relationship Management (CRM) software
  • Project Management tools
  • Collaboration tools (e.g., Google Drive)
  • Customer Support software
  • Code repositories (e.g., GitHub)

Elara empowers these users to interact with their data through natural language conversations. This translates to:

  • Chatting directly with files and folders
  • Conversing with code (debugging, adding functionalities)
  • Integrating with existing platforms (HubSpot, Zoho, ClickUp, etc)

While targeting B2B is the initial focus, future plans might include exploring freemium models for individual users. This would require further discussion and market research.

User Management & Login Flow

As we cater primarily to B2B clients, our user hierarchy involves two levels:

  1. Admin: This primary contact represents an organization. They provide the following information during signup:
    • Name (first & last)
    • Organization name
    • Phone number(optional)
    • Email address
    • Password
    • Selected models (based on desired functionalities)(explained above)*
    • Default AI model preference (explained later)*

This data is securely stored in a PostgreSQL database accessible only by a select few authorized personnel.

  1. Child Accounts: Once registered, the Admin can create additional user accounts within their organization. These child accounts allow individual users to interact with Elara through the platform. Information stored for them includes:
    • Name (first & last)
    • Email address
    • Password
    • Organization affiliation
    • Phone number (optional)
    • Profile picture (optional)
    • Preferred language (default: English, changeable)
    • Optional: GitHub token (for connecting repositories)(optional)

Admins can create new users manually or in bulk by uploading a formatted XLSX file. They can also manage child accounts, including deletion for departing employees.

User Features (Admin & Child Accounts)

Admin Features:

  • Create and manage child user accounts (individually or via bulk upload)
  • Delete child user accounts
  • Access contact information for Superadmins (technical support)

Child Account Features:

  • Select the preferred AI model (predetermined by Admins)
  • Upload files, folders, GitHub repositories, or API keys (for integrations)
  • Change profile picture
  • Set preferred language. Once set all question and answers will be done in that particular language
  • Edit personal information (name & phone number)
  • Adjust context similarity (relevance of answers to user queries) – default setting is optimized and should not be modified unless necessary
  • Access a dedicated Help Panel for raising queries to Admins or Superadmins
  • Clear a chat permanently.
  • Delete a file or folder permanently
  • Rename a file or folder
  • Download a file or folder in zip format

Security Measures

Elara prioritizes user data security and employs robust measures to safeguard your information:

  • Secure Password Storage: All user passwords are hashed using the industry-standard SHA512 algorithm, ensuring they are never stored in plain text. This significantly enhances protection against unauthorized access.
  • Data Encryption: User data is encrypted at rest within the database, adding an extra layer of security. This encryption process protects information even in the event of a potential database breach.
  • Controlled Chat History Access: While user chat history is not currently encrypted, Elara maintains transparency. Admin accounts can access the chat history of child accounts within their organization. This helps prevent unauthorized document sharing and ensures data stays within the organization.
Additional Security Considerations
  • Model Selection and Data Privacy: Choosing the LLaMA AI model (explained later) offers an extra layer of data privacy. With LLaMA, your data remains entirely on your company’s servers, and OpenAI has no access to it. This is ideal for organizations with highly sensitive information.
  • Strong Password Practices: We strongly encourage all users to:
    • Change passwords regularly: Updating passwords periodically reduces the risk associated with compromised credentials.
    • Create strong passwords: Use a combination of uppercase and lowercase letters, numbers, and symbols for maximum password strength.
    • Avoid password sharing: Never share your Elara password with anyone for optimal security.
Elara AI – Upload new file or folder
Elara AI – Upload new file or folder

AI Models

Elara empowers users with a selection of AI models, each tailored for specific scenarios and security considerations.

GPT 3.5 Turbo Model
  • Functionality: This model leverages OpenAI’s powerful GPT-3.5 architecture.
  • Security: OpenAI currently guarantees that data uploaded via API will not be used to further train their models. However, this policy is subject to change in the future, potentially impacting data privacy.
  • Data Location: Data is processed on OpenAI’s servers.
  • Recommendation: Use this model for non-sensitive information.
LLaMA (Open Source)
  • Functionality: Developed by Meta, LLaMA provides a highly secure option for organizations with sensitive data.
  • Security: This open-source model operates on an organization’s own servers, ensuring complete data privacy and control.
  • Data Location: All data remains within the organization’s infrastructure.
  • Recommendation: Use this model for confidential information.
Additional Considerations:
  • Cost: LLaMA may incur additional costs due to the need for private server infrastructure.
  • Deployment: Organizations opting for LLaMA can either provide their own servers or choose a hosted solution with associated fees.
  • Model Selection: Each organization can choose a single LLaMA model from the provided list. Currently, Elara does not support a multimodal approach utilizing multiple models simultaneously.
Available LLaMA Models:
  1. Meta-Llama-3-70B
  2. Meta-Llama-3-70B-Instruct
  3. Meta-Llama-3-8B
  4. Meta-Llama-3-8B-Instruct
  5. Meta-Llama-Guard-2-8B
  6. LlamaGuard-7b
  7. Llama-2-70b-chat-hf
  8. Llama-2-13b-chat-hf
  9. Llama-2-7b-chat-hf
  10. Llama-2-70b-hf
  11. Llama-2-13b-hf
  12. Llama-2-7b-hf
  13. Llama-2-70b-chat
  14. Llama-2-13b-chat
  15. Llama-2-7b-chat
  16. Llama-2-70b
  17. Llama-2-13b
  18. Llama-2-7b
  19. CodeLlama-70b-Instruct-hf
  20. CodeLlama-70b-Python-hf
  21. CodeLlama-70b-hf
  22. CodeLlama-34b-Instruct-hf
  23. CodeLlama-34b-Python-hf
  24. CodeLlama-34b-hf
  25. CodeLlama-13b-Instruct-hf
  26. CodeLlama-13b-Python-hf
  27. CodeLlama-13b-hf
  28. CodeLlama-7b-Instruct-hf
  29. CodeLlama-7b-Python-hf
  30. CodeLlama-7b-hf
Choosing the Right Model

Selecting the most suitable AI model depends on your organization’s specific needs and security priorities. If data privacy is paramount, LLaMA offers an on-premise solution. Otherwise, the GPT-3.5 Turbo model can handle less sensitive information.

Elara is committed to providing users with secure and powerful AI interaction tools. We will continue to evaluate and integrate new AI models that prioritize user privacy.

Future Work

Elara is constantly evolving, and we’re dedicated to providing a robust and secure platform. Here’s a glimpse into our ongoing and future development efforts.

Current Development:
  • Enhanced User Management: We’re actively building a comprehensive admin dashboard that empowers administrators to add, manage, and delete user accounts within their organization. Additionally, a super admin dashboard is being developed for centralized oversight and management.
  • CRM Expansion: Building upon the existing HubSpot integration, we’re working on incorporating support for popular third-party CRM solutions, providing users with greater flexibility.
Planned Development:
  • Collaboration & Project Management: Future iterations will involve integrating seamlessly with collaboration tools and project management platforms. This will streamline workflows and enhance team productivity.
  • Customer Support Integration: We’re planning to integrate with leading customer support platforms, allowing users to leverage Elara’s AI capabilities to enhance customer service interactions.
  • Production Deployment: Elara is on track to be migrated to a production server, ensuring scalability and optimal performance. Security measures will be implemented to limit file and folder sizes, safeguarding system resources.
  • Advanced Security: We’re committed to robust security. We’ll be implementing additional safeguards to protect against DDoS attacks and other malicious attempts.