Minimum Viable Product for Raven using GPT-3. See https://ravenagi.io/ for more information.
- Create
openaiapikey.txtin this directory with your API key as the first and only line. This file is excluded via.gitignoreso you don't have to worry about accidentally sharing your key. - Install
openaiwithpip install openai - Start
nexus.py, the web interface is accessible via (http://127.0.0.1:9999/nexus)[http://127.0.0.1:9999/nexus] - Start use the Microservices page to start remaining services
- Use the Context to inject a context into the nexus.
These are the barebones services required to demonstrate Raven as a thinking machine.
| Service | Description |
|---|---|
| Nexus | Stores the stream of consciousness, allows for arbitrary number of microservices to participate. |
| Action Generator | Ingests contexts and generates action ideas. |
| Core Objective Functions | Ingest context and action ideas, evaluates actions based on their merits. |
| Iterator | Ingests context, action ideas, and COF evaluations then generates new action ideas based upon feedback. |
These are services on the roadmap
| Service | Description |
|---|---|
| Arbiter | Looks at all available actions and picks the best one based upon other criteria such as feasibility. |
| Executive | Looks for arbitration messages and acts upon the world. |
| Recall | Stores and retrieves messages from the nexus for long term memory. Uses keyword extraction from contexts to query recall database. |
| Encyclopedia | Uses keyword extraction from contexts to find interesting relevant information from external sources. |
| Inhibitor | Deletes contexts and related messages from the nexus once executive action has been taken. |
| Augmenter | Uses recall and encyclopedia messages to synthesize new and better contexts. |
| Context | Passively watches/listens to outside world to automatically generate contexts. |