AgentSkillsCN

Research

研究

SKILL.md

Research Skill

Slice: slices/research/ Type: Knowledge Retrieval & Synthesis

Purpose

Deep research via Hi-RAG, SupaSerch, and web search. Use this skill when:

  • The user needs comprehensive information on a topic
  • Multiple sources need to be consulted and synthesized
  • Citations and confidence levels are required

Quick Start

python
from slices.research import ResearchService, ResearchTask, SourceType

service = ResearchService(http_client=client)
task = ResearchTask(
    query="How does TensorZero routing work?",
    sources=[SourceType.HIRAG, SourceType.WEB_SEARCH],
    depth="deep",
)
result = await service.execute(task)
print(result.summary)

API Reference

ResearchService

MethodDescriptionReturns
execute(task)Execute a research taskResearchResult

ResearchTask

FieldTypeDefaultDescription
querystr""Research question
depthstr"standard"shallow, standard, deep
sourcesList[SourceType][HIRAG, WEB_SEARCH]Sources to query
max_sourcesint10Maximum sources to include

ConfidenceLevel

  • HIGH: Multiple corroborating sources (3+ high-quality)
  • MEDIUM: Single authoritative or 3+ sources
  • LOW: Limited sources
  • UNCERTAIN: No sources found

Integration Points

  • Hi-RAG v2: http://localhost:8086/hirag/query
  • TensorZero: http://localhost:3030/v1/chat/completions
  • NATS Events: botz.research.completed.v1

Example Workflow

  1. Create research task with query
  2. Service queries Hi-RAG and web sources in parallel
  3. Results are ranked by relevance
  4. Findings are synthesized with confidence levels
  5. Citations are included for verification