comments

agent/agent.go

@@ -281,7 +281,6 @@ func (a *Agent) consumeJob(job *Job, role string) {
 }
 
 func (a *Agent) periodicallyRun() {
-
 	// Here the LLM could decide to store some part of the conversation too in the memory
 	evaluateMemory := NewJob(
 		WithText(
@@ -319,12 +318,6 @@ func (a *Agent) periodicallyRun() {
 	a.consumeJob(evaluateAction, SystemRole)
 
 	a.ResetConversation()
-
-	// TODO: decide to do something on its own with the conversation result
-	// before clearing it out
-
-	// Clear the conversation
-	//	a.currentConversation = []openai.ChatCompletionMessage{}
 }
 
 func (a *Agent) Run() error {

agent/state.go

@@ -23,7 +23,7 @@ type PromptHUD struct {
 // And a context memory (that is always powered by a vector database),
 // this memory is the shorter one that the LLM keeps across conversation and across its
 // reasoning process's and life time.
-// A special action is then used to let the LLM itself update its memory
+// TODO: A special action is then used to let the LLM itself update its memory
 // periodically during self-processing, and the same action is ALSO exposed
 // during the conversation to let the user put for example, a new goal to the agent.
 type State struct {