v2026.07.19

What's Changed

• refactor(cognition): prevent reward hacking by removing mandatory progress metrics by @narv-lab in #1
• fix(memory): remove real-time TTL from RedisAdapter to preserve session context by @narv-lab in #3
• fix(kernel): decouple DMN goals from Goal Omega and fix perceptions parameter mismatch by @narv-lab in #5
• fix: stabilize reflection parsing and delegate actions to main cognition by @narv-lab in #6
• refactor(llm_gateway): remove query_googleai and OpenRouter fallback logic by @narv-lab in #7
• fix(kernel): expose detailed execution errors to cognitive engine by @narv-lab in #8
• feat(dashboard): add sidebar to display Goal Omega and SubSteps by @narv-lab in #9
• Fix/reflection verbosity by @narv-lab in #10
• fix(cognitive): exclude params payload from output compactness constraint by @narv-lab in #11
• refactor: rename query_openrouter to query_litellm by @narv-lab in #13
• docs: update architecture diagrams, contributing section, and license by @narv-lab in #14
• refactor: remove openrouter references in favor of litellm by @narv-lab in #15
• docs: update README title to emphasize experimental architecture by @narv-lab in #16

New Contributors

• @narv-lab made their first contribution in #1

Full Changelog: v2026.05.08...v2026.07.19