Show HN: Model-agnostic cognitive architecture for LLMs
5 days ago
- #deterministic architecture
- #AI memory
- #event sourcing
- Persistent Mind Model (PMM) is a deterministic, event-sourced cognitive architecture for AI memory persistence.
- PMM enables transferable machine identity through a coherent event-sourced structure, independent of the underlying language model.
- Every thought, name, commitment, and reflection is immutably recorded in a pmm.db file, allowing exact mental state reconstruction.
- PMM supports local execution via Ollama or hybrid setups with OpenAI, maintaining identity across model swaps mid-conversation.
- Features include autonomous reflection loops, a Recursive Self-Model, stability metrics, and policy/meta-policy updates.
- PMM ensures accountability and self-consistency with policy-enforced write rules and hash-chained events for auditability.
- The architecture avoids parameter bloat, requiring zero fine-tuning, prompt-stuffing, or vector embeddings.
- Identity and continuity emerge from the event log itself, not the model, ensuring deterministic behavior.
- PMM includes tools for session export, telemetry, and introspection, with commands for ledger inspection and configuration.
- The system is designed for transparency, with every action recorded and traceable in the ledger.