Skip to content

Research Parameters Reference

This page provides a comprehensive guide to all configurable research parameters in MAESTRO. These settings control how research missions operate, from initial exploration to final report generation.

Parameter Categories

1. Initial Research Phase

Parameters controlling the exploratory phase where MAESTRO investigates your topic broadly.

initial_research_max_depth

  • Type: Integer
  • Default: 2
  • Range: 1-5
  • Description: Maximum depth of the question exploration tree. Higher values create more detailed initial exploration but take longer.
  • When to adjust: Increase for complex topics requiring thorough initial understanding. Decrease for focused, well-defined topics.

initial_research_max_questions

  • Type: Integer
  • Default: 10
  • Range: 5-20
  • Description: Maximum total questions to explore during initial research phase.
  • When to adjust: Increase for broad topics requiring extensive exploration. Decrease for time-sensitive research.

initial_exploration_doc_results

  • Type: Integer
  • Default: 5
  • Range: 3-10
  • Description: Number of document results to retrieve per query during initial exploration.
  • When to adjust: Increase when you have a large, relevant document library. Decrease to reduce processing time.

initial_exploration_web_results

  • Type: Integer
  • Default: 2
  • Range: 0-5
  • Description: Number of web search results to retrieve per query during initial exploration.
  • When to adjust: Increase for current events or rapidly evolving topics. Set to 0 for purely document-based research.

2. Structured Research Phase

Parameters controlling the main research phase where MAESTRO systematically investigates each section.

structured_research_rounds

  • Type: Integer
  • Default: 2
  • Range: 1-5
  • Description: Number of complete research rounds through the outline. Each round deepens the investigation.
  • When to adjust: Increase for comprehensive academic research. Keep low for quick overviews.

main_research_doc_results

  • Type: Integer
  • Default: 5
  • Range: 3-20
  • Description: Number of document results per query during main research phase.
  • When to adjust: Increase when documents are your primary source. Balance with web results for comprehensive coverage.

main_research_web_results

  • Type: Integer
  • Default: 5
  • Range: 0-20
  • Description: Number of web results per query during main research phase.
  • When to adjust: Increase for current topics or when document library is limited. Set to 0 for confidential research.

max_research_cycles_per_section

  • Type: Integer
  • Default: 2
  • Range: 1-3
  • Description: Maximum research iterations for each outline section. Controls depth of investigation per topic.
  • When to adjust: Increase for sections requiring deep analysis. Keep at 1 for time-sensitive research.

3. Writing Phase

Parameters controlling report generation and refinement.

writing_passes

  • Type: Integer
  • Default: 2
  • Range: 1-5
  • Description: Number of writing iterations (initial draft + revisions). Each pass refines and improves the content.
  • When to adjust: Increase for publication-quality output. Decrease for quick drafts or when time is critical.

writing_previous_content_preview_chars

  • Type: Integer
  • Default: 30,000
  • Range: 10,000-50,000
  • Description: Characters of previous sections shown to maintain consistency during writing.
  • When to adjust: Increase for highly interconnected content. Decrease to reduce context usage and costs.

writing_agent_max_context_chars

  • Type: Integer
  • Default: 300,000
  • Range: 100,000-500,000
  • Description: Maximum total context size for writing agent (includes notes, outline, and previous content).
  • When to adjust: Increase for complex reports with extensive notes. Decrease if hitting model context limits.

4. Writing Mode Search (Interactive Writing)

Parameters for when using search during interactive writing sessions.

writing_search_max_iterations

  • Type: Integer
  • Default: 1
  • Range: 1-3
  • Description: Search iterations for standard writing mode queries.
  • When to adjust: Increase when writing requires extensive fact-checking.

writing_search_max_queries

  • Type: Integer
  • Default: 3
  • Range: 1-5
  • Description: Number of queries per writing mode search iteration.
  • When to adjust: Increase for complex topics requiring multiple perspectives.

writing_deep_search_iterations

  • Type: Integer
  • Default: 3
  • Range: 2-5
  • Description: Search iterations when "deep search" is enabled in writing mode.
  • When to adjust: Use for critical sections requiring thorough research.

writing_deep_search_queries

  • Type: Integer
  • Default: 5
  • Range: 3-10
  • Description: Queries per iteration during deep search in writing mode.
  • When to adjust: Increase for comprehensive background research during writing.

writing_mode_doc_results

  • Type: Integer
  • Default: 5
  • Range: 3-10
  • Description: Document results per query in writing mode searches.
  • When to adjust: Match to your document library size and relevance.

writing_mode_web_results

  • Type: Integer
  • Default: 5
  • Range: 3-10
  • Description: Web results per query in writing mode searches.
  • When to adjust: Increase for fact-checking and current information needs.

5. Note Management

Parameters controlling how research notes are organized and assigned to sections.

max_notes_for_assignment_reranking

  • Type: Integer
  • Default: 30
  • Range: 10-150
  • Description: Maximum notes to keep per section after reranking during section assignment.
  • When to adjust: Increase for more note relevant content per section at the cost of processing time.

min_notes_per_section_assignment

  • Type: Integer
  • Default: 5
  • Range: 3-10
  • Description: Minimum notes assigned to each section to ensure adequate coverage.
  • When to adjust: Increase for detailed sections. Decrease for overview-style reports.

max_notes_per_section_assignment

  • Type: Integer
  • Default: 40
  • Range: 20-60
  • Description: Maximum notes per section during research rounds. Will stop after this limit is reached.
  • When to adjust: Increase for comprehensive sections. Decrease for concise reports.

research_note_content_limit

  • Type: Integer
  • Default: 32,000
  • Range: 16,000-64,000
  • Description: Maximum characters per research note.
  • When to adjust: Increase to capture more context per note. Decrease to create more granular notes.

6. Performance and Optimization

Parameters affecting speed, cost, and system behavior.

max_concurrent_requests

  • Type: Integer
  • Default: 25
  • Minimum: 10 (enforced to prevent deadlocks)
  • Range: 10-50
  • Description: Maximum concurrent API requests to LLM providers.
  • When to adjust: Increase for faster processing (higher cost). Decrease to stay within rate limits.

thought_pad_context_limit

  • Type: Integer
  • Default: 10
  • Range: 5-20
  • Description: Number of recent thoughts/decisions to maintain in agent memory.
  • When to adjust: Increase for complex decision tracking. Decrease to reduce context usage.

max_planning_context_chars

  • Type: Integer
  • Default: 250,000
  • Range: 100,000-400,000
  • Description: Maximum context size for planning operations.
  • When to adjust: Increase for complex outlines. Decrease if hitting model limits.

skip_final_replanning

  • Type: Boolean
  • Default: false
  • Description: Skip the final outline optimization step to save time.
  • When to adjust: Enable for time-critical research or when initial outline is satisfactory.

auto_optimize_params

  • Type: Boolean
  • Default: false
  • Description: Automatically adjust parameters based on topic analysis.
  • When to adjust: Enable for hands-off operation. Disable for precise control.

7. Advanced Loop Control

Parameters for fine-tuning research iteration behavior.

max_total_iterations

  • Type: Integer
  • Default: 40
  • Range: 20-100
  • Description: Global limit on total research iterations across all sections.
  • When to adjust: Increase for exhaustive research. Decrease to control costs and time.

max_total_depth

  • Type: Integer
  • Default: 2
  • Range: 1-4
  • Description: Maximum depth for hierarchical research exploration. Determines maximum depth of generated outline structure.
  • When to adjust: Increase for multi-layered topics. Keep low for flat, straightforward subjects.

max_suggestions_per_batch

  • Type: Integer
  • Default: 3
  • Range: 1-10 (-1 for unlimited)
  • Description: Parent sections to process per batch during outline revision.
  • When to adjust: Increase for faster processing. Decrease if hitting model context limits.

Presets and Templates

Quick Research

Best for: Rapid overviews, time-sensitive reports 

initial_research_max_depth: 1
initial_research_max_questions: 5
structured_research_rounds: 1
writing_passes: 2
max_concurrent_requests: 20

Standard Research

Best for: Balanced depth and speed 

initial_research_max_depth: 2
initial_research_max_questions: 10
structured_research_rounds: 2
writing_passes: 3
max_concurrent_requests: 10

Deep Research

Best for: Comprehensive investigations, academic papers 

initial_research_max_depth: 3
initial_research_max_questions: 15
structured_research_rounds: 3
writing_passes: 4
max_concurrent_requests: 15
max_research_cycles_per_section: 3

Document-Focused

Best for: Research primarily from uploaded documents 

initial_exploration_doc_results: 8
initial_exploration_web_results: 0
main_research_doc_results: 10
main_research_web_results: 0

Web-Focused

Best for: Current events, trending topics 

initial_exploration_doc_results: 2
initial_exploration_web_results: 5
main_research_doc_results: 3
main_research_web_results: 10

Settings Precedence

Parameters can be configured at multiple levels, with the following precedence (highest to lowest):

  1. Mission-specific settings - Set when starting a mission
  2. User settings - Configured in Settings → Research tab
  3. System defaults - Built-in fallback values

Performance Impact

High Impact on Speed

  • max_concurrent_requests - Direct multiplier on processing speed
  • structured_research_rounds - Linear impact on total time
  • initial_research_max_questions - Affects initial phase duration
  • writing_passes - Each pass adds significant time

High Impact on Cost

  • main_research_web_results - More results = more API calls
  • structured_research_rounds - Linear impact on API calls
  • main_research_doc_results - More embeddings and processing
  • max_concurrent_requests - Higher concurrency = higher peak usage
  • writing_passes - Multiple rewrites increase token usage

High Impact on Quality

  • structured_research_rounds - More rounds = deeper insights
  • writing_passes - More passes = better coherence
  • max_notes_per_section_assignment - More notes = richer content
  • max_research_cycles_per_section - More cycles = thorough coverage

Tips for Optimization

  1. Start with defaults - The default values work well for most use cases
  2. Adjust incrementally - Change one parameter at a time to understand impact
  3. Monitor costs - Watch the statistics tab during research
  4. Consider your sources - Balance document vs web based on your library
  5. Time vs Quality - There's always a trade-off; be clear about priorities