More on Knowledgebot Settings & Management

Knowledge Bot Settings

In the upper left corner of the Knowledge Bot, you'll find the gear icon, which opens the bot settings. Here, you can customize and enable various features to tailor the bot to your needs.

1. Capabilities & Skills: Adjust search methods and activate advanced features

2. LLM Models: Choose the language model that powers your bot's responses, balancing performance, accuracy, and cost

3. Database: Select the default database for all data rooms

4. Agents: Set up or activate customizable prompts for more efficient use

5. Automations: Set up workflows for more efficient prompting

6. Action Settings: Adds quick-access buttons to the Knowledge Bot navigation and enables advanced features like Intent Agent and System Agent for enhanced search and automation

Note: If you don't see any bot settings, you may not have access to them. Inquire to the admin of the bot regarding its access.

---

Large Language Models (LLMs)

Blockbrain allows users to select from a variety of Large Language Models (LLMs) based on their specific needs. Each model has unique strengths in terms of response quality, speed, cost efficiency, and specialized capabilities. Some other factors to consider are: Context window size, Performance and Hosting location.

Below is a comprehensive guide to help you choose the best LLM for your workflows:

Model	Best For	Unique Selling Point (USP)
Claude 3.5 Sonnet (Smartest, US)	Complex problem-solving, advanced AI agents, software engineering	Best reasoning & problem-solving Claude
Claude 3.5 Sonnet (Recommended, EU)	Same as Smartest, but EU-hosted for compliance-focused users	Best EU-hosted Claude model for enterprise
Claude 3.5 Sonnet (Creative, EU)	Creative writing, brainstorming, content generation	Optimized for storytelling & marketing
Claude 3.5 Haiku (Efficient, US)	Fast responses, chatbots, lightweight AI tasks	Fastest and cheapest Claude model
GPT-4 Omni (Logical, EU)	Text structuring, logic-heavy tasks, multimodal AI	Best for structured, multimodal tasks
Mistral Large (Coding, EU)	Advanced AI-assisted coding, software development	State-of-the-art coding capabilities
Mistral Nemo (Coding, EU)	Developers, AI-assisted programming, automation	Lightweight but powerful coding model
Gemini 1.5 Pro (Large-context, EU)	Large-scale AI processing, multimodal analysis, research	1M context window for deep analysis
Gemini 1.5 Flash (Fast, EU)	High-volume AI tasks, fast processing at scale	Optimized for speed & cost-efficiency
Llama 3.2 90B (US)	General-purpose AI, knowledge-based reasoning	Latest Llama model with improved performance
Llama 3.1 405B (US)	Enterprise-scale AI, knowledge retrieval	Flagship Meta LLM with broad capabilities
Llama 3.1 70B (US)	Standard AI workloads, NLP & automation	Balanced performance vs. compute power
Llama 3.1 8B (US)	Lightweight AI tasks, cost-effective inference	Best Llama model for smaller AI projects
GPT-4 Omni (Structured, US)	Multimodal input, knowledge organization, AI structuring	Best GPT model for complex logic tasks
GPT-4 Turbo (Legacy, US)	General AI tasks, balanced intelligence & speed	More affordable GPT-4 variant
GPT-4 Vision (US)	AI-assisted image understanding, multimodal tasks	Best for visual processing in GPT models
GPT-4o Mini (Efficient, US)	Affordable AI workloads, cost-effective intelligence	Optimized for budget-conscious users
GPT-3.5 Turbo (Legacy, EU)	Simple text tasks, lightweight AI applications	Most affordable GPT model
GPT-4 Turbo (Legacy, EU)	Advanced reasoning, enterprise-grade AI	Reliable GPT-4 model for structured tasks
Claude 3 Opus (Creative, US)	Creative content, high-level AI assistance	Most powerful Claude for creative work
Claude 3 Haiku (Efficient, EU)	Fast AI interactions, real-time responsiveness	Speed-optimized Claude model
Claude 3 Sonnet (Balanced, US)	Business applications, enterprise AI use	Great balance of speed & intelligence
Gemini 1.0 Pro (Legacy, EU)	General AI use cases, multimodal processing	Balanced Gemini model for various tasks
Mistral Codestral (Coding, EU)	Code completion, AI-powered development tools	Cutting-edge model for coding workflows
Gemma 2 (Google, US)	Small AI tasks, lightweight AI workloads	Google’s compact AI model for efficiency

Tips on how to Pick an LLM Model

1. Take into account the Context Window: refers to the amount of text (in tokens or characters) that an LLM can process at once while maintaining coherence and context

   • Short context window (e.g., 16K tokens): means the model can only consider a limited portion of text at a time, making it best suited for shorter prompts and direct answers.

   • Large context window (e.g., 1M tokens): allows the model to process longer documents, complex discussions, and in-depth analysis without losing previous context.

2. Choosing based on your Region: For better accuracy and compliance, it’s recommended to select an LLM model hosted in your region

3. Choose the default: If you’re unsure which model to pick, you may use the default LLM Models, which offer a balanced approach, ensuring high-quality responses while maintaining cost efficiency

   1. Claude 3.5 Sonnet v2: Well-rounded for reasoning, accuracy, and general tasks

   2. Azure GPT-4 Omni: Ideal for structured responses, logical reasoning, and complex queries

FAQs

1. How does the LLM affect the accuracy of AI-generated responses?

• A more advanced LLM generally provides better accuracy and reasoning, but performance also depends on:

  • The size of the context window (larger windows retain more information).

  • The quality of the input prompt (clearer prompts lead to better responses).

  • The selected embedding model (affects how well data is retrieved).

---

Model Modifiers

Model modifiers allow you to fine-tune your Knowledge Bot’s behavior by adjusting key parameters that influence how it processes and generates responses. These settings help balance creativity, precision, and relevance based on the nature of your task.

Why Use Model Modifiers?

By adjusting model modifiers, you can:

• Ensure responses align with specific business objectives

• Optimize results for creative, technical, or research-based tasks

• Improve efficiency and consistency in AI-generated outputs

Tips on Model Modifier Settings

It's best to stay close to the default settings when adjusting Model Modifiers to maintain balanced AI performance. For example, Creative Freedom should not be set too high, as excessive creativity may lead to unpredictable or overly abstract responses. Similarly, Search Range should remain within 5 to 8 to ensure the AI retrieves relevant information without unnecessary noise. Adjust settings gradually to fine-tune the AI’s behavior while maintaining accuracy and consistency.

Use Case Templates for Model Modifiers

To get the best results from AI-generated responses, it's important to fine-tune model modifiers based on your specific use case. These settings act as a starting point, allowing you to tailor the AI’s behavior to meet your needs.

Use Case	Recommended Settings
General Use: Balanced settings for everyday tasks.	Creative Freedom: Low → Allows for engaging but still logical responses. Vocabulary Range: Medium → Ensures diverse yet relevant wording. Topic Variety: Medium → Encourages AI to introduce new ideas while maintaining coherence. Word Variety: Medium → Keeps wording fresh without sacrificing clarity. Search Range: Medium → Provides a balance between precision and breadth of information.
Sales & Company Analysis: Slight creativity with a strong focus on structured insights.	Creative Freedom: Medium → Keeps responses logical while allowing for slight adaptability. Vocabulary Range: Medium → Uses varied vocabulary for engaging business communication. Topic Variety: Medium → Ensures coverage of related business topics without excessive divergence. Word Variety: Medium → Encourages compelling, clear business writing. Search Range: High → Retrieves a broad set of insights to support decision-making.
Technical Analysis & Reports: Prioritizes accuracy and consistency over creativity.	Creative Freedom: Lowest → Ensures predictable, fact-based responses. Vocabulary Range: Low → Uses precise technical language with minimal variation. Topic Variety: Low → Keeps the discussion on a single, focused subject. Word Variety: Low → Ensures terminological consistency across technical documentation. Search Range: Medium → Pulls reliable data while minimizing irrelevant information.
Data-Driven Insights: Uses structured retrieval to extract key information.	Creative Freedom: Lowest → Keeps AI responses structured and factual. Vocabulary Range: Medium → Uses varied language to articulate different insights clearly. Topic Variety: Medium → Covers related concepts while staying focused. Word Variety: Medium → Balances consistency with fresh phrasing. Search Range: High → Ensures that AI scans a broader dataset for useful insights.
Creative Writing: Maximizes AI’s creativity for expressive, imaginative output.	Creative Freedom: High → Encourages original, engaging, and sometimes unexpected responses. Vocabulary Range: High→ Expands word choice for a more colorful and engaging tone. Topic Variety: High → Allows AI to introduce fresh concepts and ideas. Word Variety: High → Enhances writing flow and prevents repetition. Search Range: Low → Prioritizes relevance over broad, factual accuracy.

Note: While these recommendations provide a strong foundation, it’s best to test different settings based on your specific workflow. Avoid maxing out values—going too high or too low can lead to unexpected or ineffective outputs. Use the tooltip descriptions in the settings panel to understand the limits of each modifier and experiment within the suggested range for balanced, high-quality results.

FAQs

1. How do I know if I have optimized my Model Modifiers correctly?

   • Test your settings by running AI queries and checking if responses meet your expectations. If responses are too rigid or repetitive, increase Creative Freedom and Word Variety. If they’re too broad or inconsistent, lower Topic Variety and Vocabulary Range.

2. Should I max out any of the Model Modifiers?

   • No, maxing out values (e.g., setting Creative Freedom to 10) can lead to unpredictable or inaccurate outputs. It’s best to stay within the recommended range provided in the tooltips inside the settings panel.

---

Embedding Models

When creating a database in Blockbrain, you are prompted to choose an embedding model. These models convert text (such as documents, files, or data) into numerical representations called embeddings. This allows the system to search, compare, and retrieve relevant content based on meaning, not just keywords.

Model	Best For	Unique Selling Point (USP)
Text Embedding 3 Large (EU)	Complex, large-scale text embedding tasks	Best for high-performance, deep text understanding
Text Embedding 3 Large (US)	Same as EU-hosted version, but US-hosted	Newest OpenAI embedding model for large-scale tasks
Text Embedding Ada 002 (US)	Basic text embedding with high efficiency	Most optimized for low-cost embeddings
Text Embedding Ada 002 (EU)	General text embeddings, suitable for diverse applications	EU-hosted version for compliance-sensitive tasks
Text Embedding 3 Small (US)	Resource-sensitive tasks with complex embeddings	Best balance between efficiency and performance
BGE Embedding (EU)	Flexible embedding for various ML applications	Ideal for self-hosted, customizable deployments
English Embedding 4 (EU)	Embedding tasks for English-language content	Best embedding model for pure English text
Multilingual Embedding 2 (EU)	Embedding for multilingual content processing	Best for handling mixed-language embeddings

Why Embedding Models Matter

Embedding models power semantic search, which helps your knowledge bot:

• Understand context and meaning across large sets of documents

• Retrieve more accurate and relevant answers

• Match user queries with similar content, even if phrased differently

Choosing the Right Model

When picking a model, consider:

• Scale of your data: Larger models handle complex text better, but cost more

• Hosting requirements: Choose EU- or US-hosted depending on compliance needs

• Languages: Use multilingual models if your data spans multiple languages

Embedding Model Recommendations

• Complexity & Scale:

  • For large-scale or complex tasks, use: → Text Embedding 3 Large

• Efficiency & Cost:

  • For general-purpose tasks with low latency and cost: → Text Embedding Ada 002

• Language Support:

  • English-only → English Embedding 4

  • Multilingual → Multilingual Embedding 2

• Hosting Requirements:

  • EU-hosted models for GDPR-sensitive data

  • US-hosted models for US-based infrastructure

FAQs

1. What is the difference between small and large embedding models?

   • Larger models (e.g., Text Embedding 3 Large) capture more context and nuances, making them better for complex queries.

   • Smaller models (e.g., Text Embedding 3 Small) prioritize efficiency and are better for lightweight applications.