v2026.03.04 (Current version)
New
  • Hypothetical (if-else) reasoning
  • Deontic modality for permissions and guardrails
  • Authority handling for agent and team hierarchies
  • Cloud version for auditing memory
  • Spatiotemporal reasoning for time- and location-based queries
  • Attribute recognition (color, size, quantities, prices, etc.)
  • Uncertainty awareness (admits knowledge gaps instead of fabricating)
  • Contradiction handling for explicit and implicit conflicts
  • Multi-user and multi-agent scenario support
  • Embed as HTML widget
  • API access
  • Multi-LLM support (OpenAI, Anthropic, Azure AI Foundry, Azure OpenAI, Bedrock, xAI)
  • SDK support for cloud agentic memory
Improved
  • Significantly improved precision: 91% on LLM evaluation, 79.2% exact match (EM), and 85.5% F1 score on HotPotQA benchmark
  • Significant speed improvement
  • Open question answering covering what, how, when, which, how many, where, and who


v1.0-reasoning ( 24 Nov 2024)
New
  • Advanced symbolic reasoning (distinct from chain-of-thought; agent truly understands topics)
  • Empathy and ability to express love
  • Authority consideration (mentor, close friend, new friend, etc.) — more cautious with friends vs. strangers
  • Forget ability — ask agent to forget topics (useful for temporary/event-related data)
  • Open question answering
  • Statement explanation capability ("why" questions) — explains thought process
  • Authentic confusion emotion
  • Authentic surprise emotion
  • Self-recognition — recognizes self as an entity with subjective experiences
Improved
  • Agent self-confidence — more careful or loose with answers based on confidence level
  • Learning — now learns during conversation (no dedicated learning state)
  • Counterquestion capability — asks questions when uncertain
v0.4-solaris(Date not specified)
New
  • Integrated browser, web scraping, and knowledge base for domain-specific knowledge
  • Full body version
Improved
  • Better clothing
  • Improved facial expressions (lip-sync)


v0.3 (Mar 20, 2024)
New
  • Concept perception
  • Topic perception
  • Counter question generator in small talk
  • Articulated moods
  • Reflective mood handling
  • Different search strategies (BERT, Levenshtein, etc.)
  • User feedback perception
  • Central keyword extractor
  • Agent self-confidence
  • Deduction
Improved
  • Interactive learning
  • Command handling



v0.2 (Jan 29, 2024)
New
  • Mood handling
  • Person introduction
  • Personalised small talk
  • Interactive learning
Improved
  • Text output and emotion expresser

v0.1-electronic-forest (Initial release)
New
  • Commands to change states (e.g., "let's learn something", "be confident", etc.)
  • Small talk
  • Basic emotion outputs
  • GPT-3 integration
  • User personality recognition
Made on
Tilda