Enhancement: Enable Context Awareness for AI Agents Across Prompts

[dannleft]

Is your feature request related to a problem? Please describe.

Currently, KaibanJS does not preserve conversational history between prompts. This means that agents do not retain past interactions or previously processed data, making it difficult for them to work cohesively on complex workflows that require contextual memory.

Without context awareness, agents may:

• Lose track of prior decisions or insights, leading to redundant or inconsistent responses.
• Struggle to maintain continuity in multi-step workflows where previous outputs should influence future decisions.
• Require excessive re-processing of information that could otherwise be remembered.

User feedback: "The Team/Agent's context awareness between the prompts. As of now, Kaiban does not preserve conversational history."

---

Describe the solution you'd like

To improve multi-agent collaboration and workflow continuity, KaibanJS could implement context awareness at both the team and individual agent levels.

Proposed Improvements:

1. Agent-Level Context Retention:

   • Allow agents to store and retrieve past interactions within a session.
   • Provide an option to enable or disable memory persistence per agent.

2. Team-Wide Shared Memory:

   • Introduce a global context store where agents can access shared information.
   • Define memory scopes (e.g., session-based, task-based, or persistent memory).

3. Configurable Context Limits:

   • Allow users to define how much historical data is retained.
   • Implement automatic summarization to prevent memory overflow.

4. Expose API Methods for Context Management:

   • Provide functions like getContext(), setContext(), and clearContext().
   • Allow developers to query, update, or reset agent memory as needed.

---

Describe alternatives you've considered

• Manual State Tracking:
  • Users currently need to implement their own external memory handling, which adds complexity.
• Storing Context in External Databases:
  • While possible, this approach requires additional setup and is not native to KaibanJS.

---

Additional context

• Relevant KaibanJS features:
  • Agent class and task execution framework.
  • Possible integration with Team object for global memory.
• Expected benefits:
  • Enable agents to recall past decisions, improving reasoning capabilities.
  • Reduce redundant task execution, improving efficiency.
  • Improve multi-agent collaboration by allowing them to build upon each other's work.