Release Notes & Migration Guides

This document provides comprehensive release notes for Lobster AI versions, covering new features, breaking changes, and recommended upgrade paths for future releases.

Table of Contents

• v0.2 Release Notes
• Future Migrations

v0.2 Release Notes

Release Date: January 2025 Status: First Public Release (Production-ready) Breaking Changes: None (first release)

Overview

Version 0.2 is the first public release of Lobster AI, providing a production-ready bioinformatics analysis platform with AI-powered agents, comprehensive multi-omics support, and professional tooling for computational biology research.

Key Features in v0.2

🔌 Content Intelligence & Publication Access

ContentAccessService - Unified publication, dataset, and web content access:

• 5 specialized providers (PubMed, PMC, GEO, bioRxiv, generic web)
• 70-80% automatic DOI/PMID resolution success rate
• Docling-powered PDF parsing with >90% Methods section detection
• Two-tier caching architecture (30-50x speedup on cache hits)
• Smart fallback strategies across providers

Usage example:

from lobster.tools.content_access_service import ContentAccessService

service = ContentAccessService(data_manager)

# Access publication (auto-resolves DOI/PMID to PDF)
pub_result = await service.access_content(
    url="10.1101/2024.08.29.610467",  # Bare DOI auto-detected
    content_type="publication"
)

# Access GEO dataset
geo_result = await service.access_content(
    url="https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE123456",
    content_type="dataset"
)

Documentation: 37-publication-intelligence-deep-dive.md

🧬 Protein Structure Visualization

Full-featured protein structure analysis with PyMOL integration:

• Professional 3D visualizations (publication-ready)
• Link protein structures to gene expression / proteomics data
• RMSD comparisons for structural analysis
• Interactive and batch visualization modes
• Automatic PDB fetching with caching

Natural language usage:

"Visualize protein structure 1AKE with cartoon representation"
"Link protein structures to my RNA-seq data for top 50 genes"

Documentation: 40-protein-structure-visualization.md

📥 Download Queue System

Robust multi-step data acquisition with JSONL persistence:

• Agent handoff pattern for complex downloads
• Persistent queue (survives crashes/restarts)
• Status tracking (PENDING → IN_PROGRESS → COMPLETED/FAILED)
• Automatic retry logic with exponential backoff
• Multi-source dataset support (GEO, SRA, PRIDE, ENA)

Key benefits:

• Decouples dataset discovery from loading
• Enables background downloads
• Fault-tolerant with automatic recovery

Documentation: 35-download-queue-system.md

🔄 Workspace & Data Management

Workspace Restoration:

• Seamless session continuity across restarts
• Pattern-based dataset loading (smart memory management)
• Automatic state tracking and recovery
• Enhanced Data Expert Agent with restoration tools

WorkspaceContentService:

• Type-safe caching for research content (publications, datasets)
• Structured storage with provenance tracking
• Fast workspace-level access (no global cache pollution)

Documentation:

• 31-data-expert-agent-enhancements.md
• 38-workspace-content-service.md

🧪 Formula-Based Differential Expression

Complex experimental designs with R-style formulas:

• pyDESeq2 integration for bulk RNA-seq
• Multi-factor designs (~ condition + batch + condition:batch)
• Agent-guided formula construction (interactive)
• Batch effect modeling and correction

Natural language usage:

"Run differential expression with formula '~ treatment + timepoint'"
"Compare conditions accounting for batch effects"

Documentation: 32-agent-guided-formula-construction.md

🏗️ Agent Infrastructure

Agent Registry Auto-Discovery:

• Dynamic agent configuration and registration
• Modular agent system with zero-config discovery
• Centralized tool routing and delegation

Enhanced CLI:

• Arrow navigation and command history
• Professional orange branding
• Rich terminal interface with syntax highlighting
• Optimized startup and processing performance

Installation

# Install via PyPI (recommended)
pip install lobster-ai

# Configure API keys
cat > .env << 'EOF'
ANTHROPIC_API_KEY=sk-ant-api03-your-key-here
EOF

# Run
lobster chat

Complete installation guide: 02-installation.md

Architecture Highlights

Multi-Agent System:

• 8+ specialized AI agents (supervisor, research, data expert, single-cell, bulk RNA-seq, proteomics, etc.)
• LangGraph-based coordination with centralized registry
• Natural language interface with context-aware routing

Data Management:

• DataManagerV2 for multi-modal orchestration (H5AD, MuData)
• W3C-PROV compliant provenance tracking
• S3-ready backends for cloud deployment

Analysis Services:

• Single-cell RNA-seq: QC, clustering, annotation, trajectory, pseudobulk
• Bulk RNA-seq: pyDESeq2 differential expression, complex designs
• Mass spectrometry proteomics: DDA/DIA workflows, missing value handling
• Affinity proteomics: Olink/antibody arrays, NPX handling

Documentation: 18-architecture-overview.md

Feature Availability

All features in v0.2 are available in both local and cloud deployment modes, with the following exceptions:

Note: Interactive PyMOL requires local GUI support. Cloud mode supports batch image generation only.

Known Limitations

1. Rate Limits: Claude API has conservative limits for new accounts. For production, use AWS Bedrock.
2. Memory: Large datasets (>10GB) may require cloud deployment for optimal performance.
3. Windows: Native installation requires WSL2. Docker is recommended for Windows users.

Troubleshooting: 28-troubleshooting.md

Agent Architecture Migration (v0.2 → v0.3)

Status: Deprecation phase (v0.2.x) → Removal (v0.3.0)

Background

In v0.2, we unified transcriptomics analysis:

• Before: singlecell_expert + bulk_rnaseq_expert (2 agents)
• After: transcriptomics_expert (unified agent)

Migration for Test Code

Old API (deprecated):

from lobster.agents.singlecell_expert import singlecell_expert
agent = singlecell_expert(data_manager)

New API (v0.2+):

from lobster.agents.transcriptomics.transcriptomics_expert import transcriptomics_expert
agent = transcriptomics_expert(data_manager)

Timeline

• v0.2.0: Deprecation warnings added
• v0.2.x: Both APIs available (current)
• v0.3.0: Old agents removed (Q2 2025)

User Impact

• End users: No action required (supervisor handles routing)
• Test code: Update imports to use transcriptomics_expert
• Custom integrations: Update agent factory references

Future Migrations

This section will be updated as new versions are released. Check back for:

• Breaking changes and deprecation notices
• New feature adoption guides
• Version-specific upgrade paths

For questions or issues, see:

• GitHub Issues
• FAQ
• Troubleshooting Guide

Last updated: December 2025 - v0.2 Release