Run Deep Research

run-deep-research

Toolkit for performing deep research on complex topics using multiple AI research providers including OpenAI Deep Research, FutureHouse Falcon, Perplexity AI, and Consensus AI.

When to Use

Use this skill when:

• The user requests comprehensive research on a topic
• The user needs citations and sources for their research
• The user wants to explore scientific literature or academic papers
• The user needs structured markdown reports with metadata

CRITICAL: Speed vs Depth Trade-offs

ALWAYS ask the user to be explicit about their approach preference:

Fast/Light Approach (seconds to complete)

• Provider: Perplexity
• Models: sonar, sonar-pro, sonar-reasoning, sonar-reasoning-pro
• Use when: Quick answers, initial exploration, time-sensitive queries
• Cost: Lower API costs
• Example:

  uv run deep-research-client research "What is CRISPR?" --provider perplexity --model sonar-pro
  

Comprehensive/Slow Approach (minutes to complete)

• Provider: Perplexity with deep research model, or OpenAI
• Models: sonar-deep-research (Perplexity), o3-deep-research-2025-06-26 (OpenAI)
• Use when: Thorough analysis needed, critical decisions, comprehensive reports
• Cost: Higher API costs and longer wait times
• Example:

  uv run deep-research-client research "What is CRISPR?" --provider perplexity --model sonar-deep-research
  

  or

  uv run deep-research-client research "What is CRISPR?" --provider openai
  

Specialized Approaches

• Scientific Literature: Use falcon provider for academic/scientific focus
• Academic Papers: Use consensus provider for peer-reviewed research (requires API approval)

Features

• Multiple Providers: OpenAI Deep Research, FutureHouse Falcon, Perplexity AI, Consensus AI
• Rich Output: Comprehensive markdown reports with citations and YAML frontmatter metadata
• Smart Caching: File-based caching in ~/.deep_research_cache/ to avoid expensive re-queries
• Auto-detection: Automatically detects available providers from environment variables
• Templates: Support for reusable research queries with variable substitution
• Citation Management: Citations included by default or saved to separate file

Required Environment Variables

# At least one of these is required:
export OPENAI_API_KEY="your-openai-key"           # For OpenAI Deep Research
export FUTUREHOUSE_API_KEY="your-futurehouse-key" # For Falcon
export PERPLEXITY_API_KEY="your-perplexity-key"   # For Perplexity AI
export CONSENSUS_API_KEY="your-consensus-key"     # For Consensus AI (requires approval)

Basic Usage

Simple Research Query

# Basic query (uses auto-detected provider)
uv run deep-research-client research "What is quantum computing?"

# With specific provider and model
uv run deep-research-client research "What is quantum computing?" --provider perplexity --model sonar-pro

# Save to file
uv run deep-research-client research "Machine learning trends 2024" --output report.md

List Available Providers

uv run deep-research-client providers

Using Templates

Templates allow reusable research queries with variable substitution:

# Use template with variables
uv run deep-research-client research \
  --template gene_research.md \
  --var "gene=TP53" \
  --var "organism=human" \
  --var "tissue=brain tissue" \
  --var "year=2020"

Template files use {variable} placeholders:

Please research the gene {gene} in {organism}, focusing on:
1. Function and molecular mechanisms
2. Disease associations in {tissue} tissue
3. Recent discoveries since {year}

Cache Management

# List cached research
uv run deep-research-client list-cache

# Clear all cache
uv run deep-research-client clear-cache

# Bypass cache for single query
uv run deep-research-client research "query" --no-cache

Output Format

Research results are returned as markdown with YAML frontmatter:

---
provider: perplexity
model: sonar-pro
cached: false
start_time: '2025-10-18T17:43:49.437056'
end_time: '2025-10-18T17:44:08.922200'
duration_seconds: 19.49
citation_count: 18
---

## Question

What is machine learning?

## Output

**Machine learning** is a branch of artificial intelligence...

## Citations

1. https://www.ibm.com/topics/machine-learning
2. https://www.coursera.org/articles/what-is-machine-learning

Model Selection Guide

Workflow

1. Ask user for speed preference: Always clarify if they want fast/light or comprehensive/slow approach
2. Check environment variables: Ensure appropriate API keys are set
3. Run research: Use uv run deep-research-client research with appropriate provider and model
4. Save output: Use --output flag to save to file if needed
5. Review results: Check the markdown output, citations, and metadata
6. Cache management: Be aware that results are cached permanently unless cleared

Important Notes

• Cost awareness: Deep research models can be expensive - always confirm with user before using
• Time awareness: Deep research can take several minutes - warn user about expected wait time
• Caching: Results are cached permanently in ~/.deep_research_cache/ - use --no-cache to bypass
• Citations: Always included by default; use --separate-citations to save to separate file
• Templates: Great for repeated research patterns (e.g., gene research, company analysis)
• Metadata: YAML frontmatter includes timing, provider info, and configuration details

Common Patterns

Quick exploratory research

uv run deep-research-client research "overview of topic X" --provider perplexity --model sonar-pro --output quick-research.md

Comprehensive deep dive

uv run deep-research-client research "comprehensive analysis of topic X" --provider openai --output deep-research.md

Scientific literature review

uv run deep-research-client research "scientific review of topic X" --provider falcon --output scientific-review.md

Template-based gene research

uv run deep-research-client research --template gene_research.md --var "gene=BRCA1" --var "organism=human" --output brca1-research.md

Python Library Usage

from deep_research_client import DeepResearchClient

# Initialize client (auto-detects providers from env vars)
client = DeepResearchClient()

# Perform research
result = client.research(
    "What are the latest developments in AI?",
    provider="perplexity",
    model="sonar-pro"
)

print(result.markdown)  # Full markdown report
print(f"Provider: {result.provider}")
print(f"Model: {result.model}")
print(f"Duration: {result.duration_seconds:.2f}s")
print(f"Citations: {len(result.citations)}")
print(f"Cached: {result.cached}")