Reasoning Modes

Logician supports multiple structured reasoning strategies. Choose the mode that fits your task complexity.

Available reasoners

Reasoner	Best for	Description
cot	Simple tasks	Standard chain-of-thought
auto_cot	Moderate tasks	Automatic reasoning depth selection
reflexion	Complex debugging	Self-reflection and revision
tot	Architecture decisions	Tree of Thoughts exploration
ssr	Code generation	Self-subjective reasoning
got	Multi-step problems	Graph of Thoughts
best_of_n	Quality-critical tasks	Multiple attempts, best result
in_context_cot	Few-shot learning	Demonstrations in context
self_consistency	Ambiguous tasks	Majority vote across attempts

Configuration

{
  "reasoning": {
    "mode": "reflexion",
    "maxIterations": 10,
    "temperature": 0.7
  }
}

Reasoning depth

The thinkingLevel setting controls how deeply the agent reasons:

Level	Description
low	Quick responses, minimal reasoning
medium	Balanced reasoning and speed
high	Deep analysis, more iterations
full	Maximum reasoning depth

Switching at runtime

Change reasoning mode during a session:

/reasoning mode tot
/reasoning depth high

When to use each

• Simple fixes → cot or auto_cot
• Bug investigation → reflexion
• Architecture changes → tot or got
• Critical code → best_of_n
• Ambiguous requirements → self_consistency