mnestiq/mnestiq is a Python execution tracing tool that records the model calls, tool usage, and network communication of AI agents as signed records. The latest release published on the GitHub repository is v0.5.0. However, the provided information does not state what changed in this version. Therefore, the changes in this release cannot be identified.
This tool focuses on producing records that let you check what inputs an agent received and which steps it went through. According to the README, it records input origins and execution stages through a proxy in front of the model or a Python instrumenter. Hash chaining and electronic signature checkpoints are used to check whether records have been altered. The README also introduces commands for record verification, investigation, and masking. Masking is the function of hiding designated information in the records. An audit dashboard and how to run it locally are also included in the documentation. It is licensed under Apache-2.0 and is intended for checking and investigating agent execution paths.
Such records are meaningful in that they help with audits and with reproducing execution processes. However, based on the provided information alone, the effectiveness of record verification or the scope of reproducibility cannot be evaluated. No performance figures are presented either. The repository has one GitHub star. Since the specific changes in v0.5.0 are also unconfirmed, the current materials only reveal the direction of the features, and it is difficult to judge its actual suitability for production use.