CogniCore
An open-source AI infrastructure project focused on persistent memory capabilities. It covers five capability areas: context engine, experience engine, state manager, and tool intelligence, among others, aiming to provide a reusable infrastructure layer for AI agents.
Timeline (last 90 days)
CogniCore is an AI Agent memory layer project based on SQLite with no vector database, licensed under MIT.
CogniCore maintains one rule: if a memory can be cleanly imported but cannot be recalled (a dark memory), the import must be made to fail.
By injecting faults at runtime (inside a subprocess, patching seams via explicit switches) rather than rewriting source code, tripwires can be made semantic.
The team proposed adding a dedicated test that enumerates each link in the chain by name, asserting that each exists in the registration path, actually runs, and fails when its test target is disabled.
The test was broken by three different reviewers in succession, with the breakage targeting the test rather than the code, and each instance belonging to a different failure category.
CogniCore is open-sourced on GitHub (cognicore-dev/cognicore-env) and has been published to PyPI (cognicore-env)
CogniCore's retrieval engine is implemented entirely using Python's standard library and can be installed via pip install.
CogniCore's benchmark data is self-reported by the project's author as a single-source test
CogniCore works as a memory backend in conjunction with existing LangChain pipelines, adding a memory layer at the agent's foundation without replacing the existing agent.
CogniCore's agentic memory migration system allows one agent to export verified memories and another agent to import them, with memory packages carrying content hashes for tamper detection and a chain of custody indicating the source
2 more timeline events
All Facts (12)
The team proposed adding a dedicated test that enumerates each link in the chain by name, asserting that each exists in the registration path, actually runs, and fails when its test target is disabled.
50%UnverifiedCogniCore is an AI Agent memory layer project based on SQLite with no vector database, licensed under MIT.
50%UnverifiedCogniCore maintains one rule: if a memory can be cleanly imported but cannot be recalled (a dark memory), the import must be made to fail.
50%UnverifiedBy injecting faults at runtime (inside a subprocess, patching seams via explicit switches) rather than rewriting source code, tripwires can be made semantic.
50%UnverifiedCogniCore's memory migration trust mechanism draws on reinforcement learning concepts: verified memories are analogous to experience replay buffers, and confidence decay as memories flow between untrusted agents is analogous to a discount factor
50%UnverifiedCogniCore's benchmark data is self-reported by the project's author as a single-source test
50%UnverifiedCogniCore's retrieval engine is implemented entirely using Python's standard library and can be installed via pip install.
50%UnverifiedCogniCore works as a memory backend in conjunction with existing LangChain pipelines, adding a memory layer at the agent's foundation without replacing the existing agent.
50%UnverifiedCogniCore's agentic memory migration system allows one agent to export verified memories and another agent to import them, with memory packages carrying content hashes for tamper detection and a chain of custody indicating the source
50%UnverifiedThe author states that the memory migration system was tested across Figma, Groq, ElevenLabs, and Claude
50%UnverifiedThe test was broken by three different reviewers in succession, with the breakage targeting the test rather than the code, and each instance belonging to a different failure category.
50%UnverifiedCogniCore is open-sourced on GitHub (cognicore-dev/cognicore-env) and has been published to PyPI (cognicore-env)
50%