The tweet announcing this feature
FreeGPT-4 API chaining with function invocation
FreeFree tier
About The tweet announcing this feature
WorkGPT is an open-source project that integrates OpenAI's function invocation feature with GPT-4 to enable dynamic chaining of APIs. By taking an array of APIs—some built dynamically from OpenAPI endpoints—the LLM determines how to combine them to generate answers. This approach aims to be faster, cheaper, and less error-prone than earlier systems like AutoGPT. The concept extends to 'auto work bots' that automate business processes, and envisions future LLMs with larger context windows orchestrating a full suite of APIs.
Key Features
OpenAI function invocation integration
Dynamic API chaining with GPT-4
Supports OpenAPI endpoint definitions
Designed to be faster and cheaper than AutoGPT
Foundation for auto work bots
Pros & Cons
Pros
- Leverages GPT-4's reasoning for API orchestration
- Reduces cost and latency compared to AutoGPT
- Supports dynamically defined APIs via OpenAPI
- Open-source and free to use
Cons
- Requires access to OpenAI API with function calling enabled
- Currently limited by GPT-4's context window size
- Proof-of-concept stage; may have reliability issues in production
- Dependence on external API availability and latency
Best For
Automating complex business processesBuilding auto work bots that replace manual workflowsCombining multiple APIs to answer user queriesPrototyping AI-powered API orchestration systems
FAQ
What is WorkGPT?
WorkGPT is an open-source project that demonstrates using GPT-4 with OpenAI's function invocation to chain multiple APIs together, enabling the LLM to combine APIs to generate answers.
What new feature does WorkGPT have?
It now bakes in OpenAI's function invocation feature, allowing GPT-4 to dynamically use an array of APIs—some built from OpenAPI endpoints—to figure out how to combine them for a response.
How does WorkGPT compare to AutoGPT?
According to the creator, WorkGPT is faster, cheaper, and less error-prone than AutoGPT for chaining APIs, as it leverages native function calling rather than iterative reasoning.