Constraint-Aware Corrective Memory for Language-Based Drug Discovery Agents

Build a corrective memory layer for agents operating under strict output constraints. It would allow agents to backtrack and self-correct when generated batches violate global protocol constraints.

Suggested repo: boundMem

"Give your agent a long-term memory for hard constraints."

Estimated effort: 45h