Currently, interacting with Freysa on-chain incurs a cost due to the Ethereum-based message submission system. This limits the ability for developers, researchers, or contributors to test prompts, debug interactions, or explore the decision logic of the agent.
Proposal: Introduce a simulation mode or local replica that mirrors Freysa’s logic without actual on-chain interaction. This mode would:
- Allow messages to be submitted via CLI or web UI.
- Mimic the agent’s core logic and output decisions without signing transactions or enforcing financial constraints.
- Be clearly marked as non-production and read-only, avoiding key exposure or misrepresentation of final behavior.
Benefits:
- Encourages experimentation and better understanding of the agent’s behavior.
- Enables testing of prompt safety and adversarial boundaries in a sandboxed environment.
- Makes the project more accessible to contributors and educators.
Possible Implementation Ideas:
- Use a mock environment to simulate TEE behavior (e.g., SGX mock or Docker-based isolation).
- Provide a config flag like
--simulate or FREYSA_MODE=dev to enable it.
- Disable any Ethereum signing or treasury access logic in this mode.
Currently, interacting with Freysa on-chain incurs a cost due to the Ethereum-based message submission system. This limits the ability for developers, researchers, or contributors to test prompts, debug interactions, or explore the decision logic of the agent.
Proposal: Introduce a
simulation modeor local replica that mirrors Freysa’s logic without actual on-chain interaction. This mode would:Benefits:
Possible Implementation Ideas:
--simulateorFREYSA_MODE=devto enable it.