RAG Agent Platform

🤖 Total Agents

📚 Knowledge Bases

🧠 Available Models

🤖 My Agents

✨ Create New Agent

Agent Name * (0/30)

Model *

Instructions *

📚 Knowledge & Tools

Knowledge Bases

🔧 Tools — real-time data functions the agent can call

⚙️ Advanced Settings

Temperature (0-1) Controls randomness. Lower (0.1-0.3) = focused, higher (0.7-1.0) = creative

Max Tokens Maximum response length (~4 chars per token)

Top P (0.1-1) Nucleus sampling. Lower (0.1-0.5) = focused, higher (0.9-1.0) = diverse

Top K (0-10) Limits token choices. Lower = predictable, higher = variety

Chunks per Knowledge Base How many text chunks to retrieve from each KB for context (recommended: 3-5)

Retrieval Method How to process queries before searching knowledge bases

Include citations in responses Shows source references and document citations in responses

Strict RAG Mode When enabled, agent ONLY uses knowledge base content. When disabled, can use general knowledge

📚 Knowledge Bases

📄 Documents in -

➕ Add Documents

Supported formats: Text, Markdown, PDF, HTML, CSV, Word (.docx)

📦 Batch Queue

No documents added yet

⚙️ Configuration

Chunking Strategy

View Auto-Detection Rules

PDF/Word/HTML: Text extraction + paragraph splitting
CSV: Parsed as table + row-based splitting
Markdown: Header-aware splitting (preserves structure)
Text: Recursive paragraph/line splitting

Header Depth Range (Markdown only)

Start Level

→

End Level

Split only at headers in this range. Example: H2→H3 keeps H1 titles intact.

Max Chunk Size (characters) Splits by headers first, then by this size if sections are too large

Available Models

Language models via DigitalOcean AI Inference • Embedding models via OpenAI API or local HuggingFace

🤖 RAG Agent Platform

🔒 Authentication Required

🤖 My Agents

Agent Details

✨ Create New Agent

📚 Knowledge Bases

Indexing in Progress

📄 Documents in -

➕ Add Documents

📦 Batch Queue

⚙️ Configuration

Available Models