ThinkGPT
FreeAgent techniques to augment your LLM and push it beyond its limits 
FreeFree tier
Inputs: textOutputs: text
About ThinkGPT
ThinkGPT is a Python library developed by Jina AI that implements Chain of Thoughts for Large Language Models (LLMs). It provides thinking building blocks to enhance LLM reasoning, including memory for storing and recalling information, self-refinement to improve model-generated content, knowledge compression to fit context, inference for educated guesses, and natural language conditions for decision-making. The library aims to solve limited context issues, improve one-shot reasoning, and add intelligent decisions to codebases.
Key Features
Memory: Enables GPT models to remember and recall information from past interactions
Self-refinement: Improves model-generated content by addressing critics and feedback
Knowledge compression: Compresses knowledge to fit within LLM context using abstraction or summarization
Inference: Makes educated guesses based on available information
Natural Language Conditions: Express choices and conditions in natural language
Efficient and measurable GPT context length handling
Extremely easy setup with a Pythonic API using DocArray
Pros & Cons
Pros
- Open-source and free to use
- Easy to install and integrate with Python projects
- Provides memory management for LLMs to remember context across interactions
- Supports self-refinement to iteratively improve outputs
- Allows knowledge compression to fit more information into limited context
- Offers natural language conditions for intuitive decision logic
Cons
- Requires access to OpenAI API (uses gpt-3.5-turbo by default), incurring usage costs
- Dependence on external LLM API; offline or local models not supported out of the box
- Limited documentation and examples beyond the README
- Python-only library, not available for other programming languages
Best For
Solving limited context problems by using long memory and compressed knowledgeEnhancing LLM one-shot reasoning with higher-order reasoning primitivesAdding intelligent decision-making to Python codebasesImproving code by refining snippets based on error messages and criticsBuilding generative agents that can learn and recall information
FAQ
What is ThinkGPT?
ThinkGPT is a Python library for implementing Chain of Thoughts with Large Language Models (LLMs). It provides memory, self-refinement, knowledge compression, inference, and natural language conditions to enhance LLM reasoning.
How does the memory feature work?
The library allows you to memorize content using the `memorize()` method and retrieve relevant information with `remember()`, supporting limits on number of documents or token count.
Can I use ThinkGPT with any LLM model?
The library is designed to work with OpenAI's GPT models, defaulting to gpt-3.5-turbo. The current implementation is tied to the OpenAI API.
Is ThinkGPT free?
ThinkGPT is open-source and free to use, but it requires an OpenAI API key which may have associated costs.