Agent Search is Hymalaia advanced knowledge retrieval system that enables answering complex, multi-faceted questions by intelligently decomposing queries, searching across multiple contexts, and synthesizing comprehensive answers.

Unlike traditional search, Agent Search approaches questions like a knowledgeable colleague would:

  1. Decompose and disambiguate
  2. Analyze narrow, well-defined sub-questions
  3. Synthesize and present comprehensive context-rich answers

đź’ˇ Example: When comparing two products (e.g. Car A vs. Car B), Agent Search will independently explore both, then compare them to form a rich, contextual answer.

​
Key Features

  • Intelligent Query Decomposition
    Breaks complex questions into precise sub-questions

  • Parallel Search Processing
    Executes multiple analysis threads simultaneously

  • Answer Validation
    Refines and validates responses for accuracy and completeness

​
Configuration

​
Basic Setup

To enable Agent Search in your Hymalaia deployment:

  1. Update to the latest version of Hymalaia
  2. Configure knowledge source connections
  3. Set up LLM provider credentials
  4. Enable the Agent toggle in the chat interface (with a search-capable assistant)

​
Advanced Configuration

​
Best Practices & Suggestions

  • Don’t hesitate to ask complex or multi-layered questions.
  • Try comparative queries like:

    “What’s the difference between Solution A and B?”
    Agent Search will separately analyze A and B before comparing.

  • Ask ambiguous questions such as:

    “What are the guiding principles for X?”
    The system will use context to clarify what “guiding principles” refers to.

  • Even simple questions may benefit from deeper, contextualized answers.
  • Click on sub-question analyses — they may provide interesting insights individually.

⚠️ It is recommended to assign a faster/cheaper LLM model as your Fast Model, since Agent Search performs many parallel queries.

​
Common Issues and Solutions

IssueSolution
Langgraph/Langchain errorsEnsure server uses Python 3.11 and installs libraries from backend/requirements.txt.
Rate limitsAgent Search may hit rate limits due to parallel queries. Use a provider with higher limits.
TimeoutsTimeout thresholds are enforced to avoid blocking. Contact support if these are too strict.
High token usageExpect significantly more input/output tokens than with Basic Search.

​
Summary

Agent Search offers a powerful way to surface deeper insights, especially when working with ambiguous or multi-faceted questions. For best performance:

  • Use optimized LLM configurations
  • Expect and account for higher token usage
  • Experiment with your queries to see how well the system synthesizes knowledge

💬 Reach out to us on Slack or Discord if you’re experiencing issues or want help fine-tuning your setup.