v0.2.0

What's new in v0.2.0

Skill evolution pipeline

• WikiSkillAdapter — pull curated skills from a WikiSkill knowledge base into AgentDB's SkillStore
• SkillOptAdapter — push post-training skill signals from SkillOpt back into AgentDB
• SkillEvolutionManager with advisory lock primitives to prevent concurrent write conflicts
• origin, validation_score, and metadata fields on Skill; health score blends all three signals
• Auto-propose skill drafts from ungoverned consolidation clusters

Memory lifecycle

• Memory consolidation: clusters related episodic records and surfaces a canonical representation
• Session metadata tracking
• Skill conflict signal: flags skills that diverge from LLM-canonical behaviour
• checkpoint() and write_compaction_summary() — persist context compaction events into AgentDB so long-running sessions survive history trimming

Notifications

• notifications package with EmailChannel, SlackChannel, and NotificationDispatcher
• Wire notification channels from memory.yaml notifications: block

Embeddings

• YAML-configurable embeddings via embeddings: block in memory.yaml — no code change required to enable semantic retrieval
• LocalEmbeddingProvider supports multiple models (per-model cache, configurable via YAML or constructor)
• Activation precedence: embed_fn= kwarg > local_embed=True flag > YAML embeddings.enabled: true

Providers

• AnthropicProvider for ingesting Anthropic conversation history directly into AgentDB
• providers extra: pip install 'agentdatabase[providers]'

Other

• agentdb_stats() now includes total_skills and stale_skill_count
• conflict_pending_count fix in stats
• Skill injection removed from retrieve(); skill staleness anchored on updated_at

Install

pip install agentdatabase==0.2.0

# Optional extras
pip install 'agentdatabase[local-embed]'   # semantic retrieval
pip install 'agentdatabase[notifications]' # Slack/email alerts
pip install 'agentdatabase[providers]'     # Anthropic ingestion
pip install 'agentdatabase[evolution]'     # WikiSkill + SkillOpt