Memory Types
Xagent organises memories into categories to make them more useful and searchable. Each category is retrieved in different situations.
Plan Execution Memory
Insights from task planning and DAG generation — planning strategies that worked, step-decomposition approaches, tool-selection insights, and task-complexity analysis. Retrieved when starting similar tasks or planning multi-step workflows.
"For data analysis tasks with multiple steps, generate a DAG
with parallel processing for independent calculations. Use the
Python executor for transformations."Execution Memory
Results and learnings from task execution — outcomes (success/failure), user feedback and corrections, tool effectiveness, performance bottlenecks, and error patterns with solutions. Retrieved when encountering similar challenges or optimising performance.
"When processing CSV files larger than 100MB, use chunked
reading with pandas. Web search API times out after 30
seconds for complex queries."ReAct Memory
Insights from the ReAct (Reasoning–Action) loop — reasoning patterns, tool-usage effectiveness, step-by-step execution insights, and iteration strategies. Retrieved when using ReAct execution mode or choosing reasoning strategies.
"For research tasks requiring multiple sources, start with web
search to identify key sources, then use file tools to process
and synthesize findings."General Memory
General insights and user preferences that aren't tied to one execution strategy — communication preferences, format and style preferences, project-specific context, domain knowledge, and behavioural observations. This is the default category, retrieved across all task types to personalise how the agent works with you.
"This user prefers concise answers with a summary table, and
reports formatted as Markdown with British spelling."