For AI agents: a documentation index is available at the root level at /llms.txt and /llms-full.txt. Append /llms.txt to any URL for a page-level index, or .md for the markdown version of any page.
DiscordGet API Key
  • Getting Started
    • Overview
    • Quickstart
    • Errors
    • Pricing
  • Chat
    • Overview
    • Chat with Image, Video, and Audio
    • Function Calling
    • Models
  • Vision
    • Overview
    • Rate Limits
    • Pricing
    • MCP Server
    • Video Management
    • Video Group Management
    • Video Search
    • Video QA
    • Clip Generation
    • Metadata Tagging
    • Image Management
    • Image Search
  • Research
    • Overview
    • Streaming
    • Reasoning Steps
    • Web Search
    • Structured Output
    • Parallel Thinking
    • Best Practices
    • Errors
    • Examples
  • Speech
    • Overview
    • Audio Transcription
    • Speech Translation
    • Speech-to-Speech Translation
  • Resources
    • FAQs
    • Changelog
    • System Status
LogoLogo
DiscordGet API Key
On this page
  • Ensure Prompts Are Clear and Focused
  • Inject Context for More Targeted Retrieval
  • Generate Search-Friendly Queries
  • Prevent Overloaded or Ambiguous Prompts
  • Discourage LLM Roleplay Instructions
  • Increase number of results
Research

Best Practices for Reka Research

Was this page helpful?
Previous

Error Codes and Rate Limits

Next
Built with

Follow these best practices to help users craft clear and retrievable queries when using Reka Research. Well-structured prompts improve result quality and reduce unnecessary searches.

Ensure Prompts Are Clear and Focused

Generate precise and scoped prompts in your application. Reka Research performs best when queries are explicit and to the point.

✅ Do: “Summarize key points from the 2024 WHO guidelines on pandemic preparedness.”

🚫 Avoid: “Tell me about health guidelines”

Inject Context for More Targeted Retrieval

Encourage users to include contextual details like dates, locations, or named entities. These cues significantly improve the accuracy of retrieved results.

✅ Do: “How did the 2023 California drought impact almond crop yields?”

🚫 Avoid: “How did the drought affect crops?”

Generate Search-Friendly Queries

Structure queries using domain-relevant terms that are likely to appear in source documents. Avoid vague or comparative phrasing.

✅ Do: “Compare lithium iron phosphate vs. nickel manganese cobalt batteries for EV performance”

🚫 Avoid: “Which battery is better?”

Prevent Overloaded or Ambiguous Prompts

Split unrelated or multi-part questions into separate queries. Clear, single-purpose requests yield better search outcomes.

✅ Do: “Explain how quantum entanglement affects secure communication protocols”

🚫 Avoid: “Tell me about quantum computing, space travel, and AI breakthroughs”

Discourage LLM Roleplay Instructions

Avoid prompts that mimic LLM roleplay or personas. Instead, encourage users to ask direct questions or request specific outputs.

✅ Do: “Give a step-by-step tutorial for setting up a PostgreSQL read replica on AWS RDS”

🚫 Avoid: “Act as a database expert and walk me through PostgreSQL replication like I’m five”

Increase number of results

Specify the number of results you want to retrieve. This helps ensure you get enough information without overwhelming the user.

✅ Do: “What are the top 5 trends in renewable energy for 2025?”

🚫 Avoid: “What’s new in renewable energy?”

Join our Discord community to share tips and learn from other developers. Need help? Reach out to us anytime.