cogitator 0.1.0a3

A Python toolkit for chain-of-thought prompting

Cogitator

A Python toolkit for chain-of-thought prompting

---

Cogitator is a Python toolkit for experimenting and working with chain-of-thought (CoT) prompting methods in large language models (LLMs). CoT prompting improves LLM performance on complex tasks (like question-answering, reasoning, and problem-solving) by guiding the models to generate intermediate reasoning steps before arriving at the final answer. Additionally, it can be used to improve the interpretability of LLMs by providing insight into the model's reasoning process. The toolkit aims to make it easier to use popular CoT strategies and frameworks for research or integrating them into AI applications.

Features

• Provides unified sync/async API for CoT strategies
• Supports using OpenAI and Ollama as LLM providers
• Supports structured JSON model output with Pydantic validation
• Includes a customizable benchmarking framework (see benches)
• Includes implementations of popular CoT strategies and frameworks like
  • Self-Consistency CoT (ICLR 2023)
  • Automatic CoT (ICLR 2023)
  • Least-to-Most Prompting (ICLR 2023)
  • Tree of Thoughts (NeurIPS 2023)
  • Graph of Thoughts (AAAI 2024)
  • Clustered Distance-Weighted CoT (AAAI 2025)

---

Getting Started

pip install cogitator

Or, if you want to install from the latest version with examples and benchmarks included

git clone --depth=1 https://github.com/habedi/cogitator && cd cogitator

# Set up Python environment
pip install poetry
poetry install --with dev

# Run the tests to make sure everything is working (optional)
poetry run pytest

Examples

See the examples directory for examples.

import logging
from cogitator import SelfConsistency, OllamaLLM

# Step 1: Configure logging (optional, but helpful)
logging.basicConfig(level=logging.INFO)
logging.getLogger("httpx").setLevel(logging.WARNING)  # Suppress HTTPX logs

# Step 2: Initialize the LLM (using Ollama)
# Needs Ollama running locally with the model pulled (e.g., `ollama pull gemma3:4b`)
try:
    llm = OllamaLLM(model="gemma3:4b")
except Exception as e:
    print(f"Error initializing Ollama LLM: {e}")
    print("Please make sure Ollama is running and the model is pulled.")
    exit(1)

# Step 3: Choose a CoT strategies (Self-Consistency in this case)
# Self-Consistency generates multiple reasoning paths and finds the most common answer
sc_strategy = SelfConsistency(
    llm,
    n_samples=5,  # Number of reasoning paths to generate
    temperature=0.7  # Higher temperature can lead to more diverse answers
)

# Step 4: Define the prompt (with a basic CoT trigger)
question = "A bat and a ball cost $1.10 in total. The bat costs $1.00 more than the ball. How much does the ball cost?"
prompt = f"Q: {question}\nA: Let's think step by step."

# Step 5: Run the CoT prompting sc_strategy
print(f"\nQuestion: {question}")
print("Running Self-Consistency CoT...")
final_answer = sc_strategy.run(prompt)  # Returns the most consistent (repeated) answer

# Expected output: $0.05 or 0.05 (may vary slightly based on model and temperature)
print(f"\nCogitator's Answer (Self-Consistency): {final_answer}")

---

Documentation

See the docs directory for the documentation for the Cogitator toolkit.

---

Benchmarking Framework

This project includes a customizable and extensible benchmarking framework to evaluate the performance of different CoT strategies on various datasets like GSM8K and StrategyQA.

Check out the benches directory for more details about the framework and how it could be used.

---

Contributing

See CONTRIBUTING.md for details on how to make a contribution.

Logo

The logo is named "Cognition" and was originally created by vectordoodle.

License

Cogitator is licensed under the MIT License.