Feat/llm autonomous mode#44
Conversation
This commit adds the foundational Go code within the (conceptual) 'chipper' module to enable an LLM-driven autonomous mode for Vector. Key changes: - New 'autonomous.go': Contains the main 'AutonomousLLMLoop' for fetching (placeholder) camera data, querying an LLM, parsing structured commands (e.g., CMD_DRIVE_WHEELS, CMD_SAY), and executing them using the vector-go-sdk (via assumed placeholder calls). - Modified 'intent.go': Adds logic to recognize a new voice intent (e.g., "start autonomous mode") to trigger the AutonomousLLMLoop. - Modified 'config.go': Introduces new configuration options for enabling autonomous mode, setting a custom system prompt, model, and cycle delay. This submission provides the code structure and logic. Manual integration into the chipper module and extensive testing by the user are required due to limitations in directly modifying external dependencies and lack of a test environment.
|
This looks like an awesome start to bringing Vector even more to life, I was thinking about this myself, connecting Vector to a Hermes agent. From this perspective, Wirepod would function as a set of tools or MCP server. This PR is a bit ahead of it's time! |
Hey this was actually the forst draft a year ago i have been development a better solution, "VectorBrain" where Vector isn't passive like "Hey Vector (does Action)" but more of a constant listening mode and analyzing I got it currently to run in sync with a local model using the hooks in go with some slight modifications to the Go code to have better control over python, if you're interested I can link the better repo I wrote a little piece about it here |
Check the 3 repos https://github.com/infraax/vectorax |
I think you would be interested in reading this piece ;) https://github.com/infraax/vectorbrain/blob/main/not-a-toy.md |
3 participants
No description provided.