Getting Started with Haive Prebuilt Agents

Welcome to Haive Prebuilt - the fastest way to deploy specialized AI agents for business, research, and creative applications.

Note

Beta Status: This package is currently being refactored into the new dynamic agent architecture. All agents are functional, but APIs may evolve. Pin to specific versions for production use.

Installation

Install via poetry in your project:

poetry add haive-prebuilt

Or include in your pyproject.toml:

[tool.poetry.dependencies]
haive-prebuilt = "^0.1.0"

Quick Start

Deploy your first specialized agent in under 5 minutes:

Business Intelligence Agent:

from haive.prebuilt.company_researcher import CompanyResearcher
from haive.core.engine.aug_llm import AugLLMConfig

# Create configuration
config = AugLLMConfig(
    temperature=0.3,  # More deterministic for business analysis
    max_tokens=2000
)

# Deploy company research agent
researcher = CompanyResearcher(name="BusinessAnalyst", engine=config)

# Analyze a company
result = await researcher.arun(
    "Provide a comprehensive analysis of Apple Inc. including financial performance, market position, and competitive advantages"
)

Creative Content Agent:

from haive.prebuilt.blog_writer_agent import BlogWriterAgent

# Deploy content creation agent
writer = BlogWriterAgent(
    name="ContentCreator",
    engine=AugLLMConfig(temperature=0.8)  # More creative
)

# Generate blog content
article = await writer.arun(
    "Write a 1500-word blog post about the future of renewable energy, targeting a general audience"
)

Research Assistant Agent:

from haive.prebuilt.scientific_paper_agent import ScientificPaperAgent

# Deploy research agent
research_agent = ScientificPaperAgent(
    name="ResearchAssistant",
    engine=AugLLMConfig(temperature=0.2)  # More precise
)

# Analyze research papers
analysis = await research_agent.arun(
    "Analyze the methodology and findings of recent papers on quantum computing applications in cryptography"
)

Agent Categories Overview

💼 Business Intelligence (5+ Agents)

  • Company Researcher: Market analysis and competitive intelligence

  • Sales Call Analyzer: Performance optimization and insights

  • Customer Support: Automated customer service

  • Project Manager: Task coordination and planning

  • Career Assistant: Professional development guidance

🔬 Research & Academia (5+ Agents)

  • Scientific Paper Agent: Research analysis and synthesis

  • Academic Task Learning: Educational content creation

  • Systematic Review: Literature review automation

  • Open Researcher: Open science collaboration

  • Essay Grading: Academic assessment

🎨 Creative & Content (5+ Agents)

  • Blog Writer Agent: Professional content creation

  • Podcast Generator: Audio content production

  • GIF Generator: Visual content creation

  • TTS Poem Generator: Poetry and audio content

  • Content Intelligence: Content optimization

⚖️ Legal & Compliance (4+ Agents)

  • Contract Analysis: Legal document review

  • Clause AI: Contract optimization

  • EU Green Compliance: Environmental compliance

  • Constitutional Agent: Legal framework analysis

🛠️ Technical & Development (4+ Agents)

  • DB Discovery: Database analysis

  • Graph Inspector: Data structure analysis

  • E2E Testing: Test automation

  • Self Improving: Adaptive agents

🚀 Startup & Innovation (4+ Agents)

  • Startup Suite: Business planning

  • AI Insight: Technology trend analysis

  • GTLA: Go-to-market assistance

  • Shop Genie: E-commerce optimization

Configuration Patterns

Temperature Guidelines:

# Business & Legal (Precision)
config = AugLLMConfig(temperature=0.1)

# Research & Analysis (Balanced)
config = AugLLMConfig(temperature=0.3)

# General Purpose (Default)
config = AugLLMConfig(temperature=0.7)

# Creative & Content (Creative)
config = AugLLMConfig(temperature=0.9)

Structured Output:

from pydantic import BaseModel, Field

class CompanyAnalysis(BaseModel):
    financial_health: str = Field(description="Financial assessment")
    market_position: str = Field(description="Competitive position")
    risk_factors: list[str] = Field(description="Key risks")
    opportunities: list[str] = Field(description="Growth opportunities")

# Use with structured output
config = AugLLMConfig(structured_output_model=CompanyAnalysis)
researcher = CompanyResearcher(name="analyst", engine=config)

Multi-Agent Workflows

Combine multiple prebuilt agents for complex workflows:

from haive.agents.multi import MultiAgent
from haive.prebuilt.company_researcher import CompanyResearcher
from haive.prebuilt.blog_writer_agent import BlogWriterAgent
from haive.prebuilt.contract_analysis import ContractAnalyzer

# Create specialized workflow
due_diligence_workflow = MultiAgent(
    name="due_diligence_pipeline",
    agents=[
        CompanyResearcher(name="researcher", engine=research_config),
        ContractAnalyzer(name="legal", engine=legal_config),
        BlogWriterAgent(name="reporter", engine=content_config)
    ],
    execution_mode="sequential"
)

# Execute comprehensive analysis
result = await due_diligence_workflow.arun(
    "Perform due diligence on TechCorp acquisition including company analysis, contract review, and summary report"
)

Best Practices

1. Choose the Right Agent:

# For factual analysis - use research agents
from haive.prebuilt.open_researcher import OpenResearcher

# For creative work - use content agents
from haive.prebuilt.podcast_generator import PodcastGenerator

# For business tasks - use business agents
from haive.prebuilt.sales_call_analyzer import SalesCallAnalyzer

2. Configure Appropriately:

# Match temperature to task type
factual_config = AugLLMConfig(temperature=0.1)      # Precise
balanced_config = AugLLMConfig(temperature=0.5)     # Balanced
creative_config = AugLLMConfig(temperature=0.9)     # Creative

3. Use Structured Output When Needed:

# For consistent data extraction
config = AugLLMConfig(structured_output_model=YourModel)

# For integration with downstream systems
agent = YourPrebuiltAgent(engine=config)

4. Monitor and Optimize:

# Enable debug mode during development
result = await agent.arun(input_data, debug=True)

# Log performance metrics
import logging
logging.getLogger("haive.prebuilt").setLevel(logging.INFO)

Common Deployment Patterns

Standalone Service:

# Single-purpose deployment
customer_support = CustomerSupportAgent(
    name="support_bot",
    engine=AugLLMConfig(temperature=0.4)
)

# API endpoint integration
async def handle_support_request(request):
    return await customer_support.arun(request.message)

Background Processing:

# Batch processing workflow
document_analyzer = ContractAnalyzer(name="batch_processor")

for document in document_queue:
    analysis = await document_analyzer.arun(document.content)
    save_analysis(document.id, analysis)

Real-time Integration:

# Streaming/real-time processing
content_optimizer = ContentIntelligenceAgent(name="optimizer")

async def optimize_content(content_stream):
    async for content in content_stream:
        optimized = await content_optimizer.arun(content)
        yield optimized

Troubleshooting

Common Issues

  1. Import Errors: Ensure haive-core is installed:

    poetry add haive-core

  2. Configuration Issues: Check engine configuration:

    config = AugLLMConfig()  # Uses defaults
print(config.validate_configuration())

  3. Performance Issues: Adjust temperature and max_tokens:

    # For faster responses
config = AugLLMConfig(max_tokens=500, temperature=0.5)

  4. Memory Issues: Use streaming for large inputs:

    # Process in chunks for large documents
result = await agent.arun(content_chunk, stream=True)

Debug Mode:

# Enable detailed logging
import logging
logging.getLogger("haive").setLevel(logging.DEBUG)

# Use debug flag
result = await agent.arun(input_data, debug=True)
print(f"Execution time: {result.execution_time}ms")

Next Steps

  • agent_overview - Detailed overview of all available agents

  • configuration - Advanced configuration options

  • examples - Real-world usage examples

  • business/company_researcher - Business intelligence deep dive

  • creative/blog_writer_agent - Content creation guide

Need Help?