continuous-learning

Pattern detection evaluates your Gemini Code session transcripts against a configurable extraction_threshold to identify and save reusable patterns. This ensures only sufficiently significant patterns are stored as learned skills.

Yes, you can configure pattern detection by adjusting the min_session_length and extraction_threshold parameters. These settings control when session evaluation triggers and what qualifies as a reusable learned skill.

You can change the storage path for learned skills by configuring the learned_skills_path variable. By default, the system saves extracted patterns directly to the ~/.gemini/skills/learned/ directory for future reuse.

Continuous learning in Gemini Code automatically captures reusable patterns from your session transcripts and stores them as learned skills. It evaluates sessions at completion to build a localized knowledge base for future end-to-end workflows.

No, continuous learning operates without external dependencies. It relies entirely on the built-in Stop hook mechanism to evaluate session transcripts and store detected patterns locally in your environment.