Nexus-G Documentation
Platform architecture, workflows, pipelines (including the Pipeline Agent wizard), Agentic BLAST, igtools / bench tools, Geneie & EcoTracks, cloud operations, plus user and technical guides.
Platform Architecture
Nexus-G is a local-first agentic bioinformatics workspace. The desktop client (GIAT engine) stores boards, downloads, BLAST databases, and run artifacts under ~/stDatApp. Heavy steps can stay on-laptop or burst to AWS Batch / GCP via BYOS storage and the NG-Cloud database stream.
System Topology
Local-first design
Basic and Standard execute analytical engines on user hardware. Cloud is opt-in bursting, not a requirement.
Reproducible runs
Workflow and pipeline runs snapshot parameters under the run folder for 100% reproducible re-execution.
Open-core licensing
Community tier is GPLv3 (nexus-g-core). Enterprise commercial licenses cover proprietary clinical/federal modules.
Workflows
The Workflow Builder lets you create repeatable multi-step pipelines — download → convert → BLAST → report — without writing code. Find it under Workflow Builder / Analysis Dashboard.
Concepts
- · Workflow — named pipeline of steps
- · Step — unit of work (NCBI download, convert, BLAST, report)
- · Template — prebuilt starting point
- · Job — one execution with status and outputs
Lifecycle
- · Create (new / template / import)
- · Add steps and configure parameters
- · Save, then Execute → Job
- · Monitor in Jobs; review ~/stDatApp/workflow_results/
Tabs: Workflows list · Editor · Step Library · Jobs. Export/import workflows for handoff between machines.
Standard & Custom Pipelines
Standard pipelines live under giat_package/pipelines. The metabarcoding path below is an example — users can create unlimited custom pipelines and insert their own tools and code with the Pipeline Agent wizard (discovery, install, format adaptation, save).
Example: metabarcoding pipeline stages
ExampleOutputs: combined_enhanced.csv, combined_arbitration.csv, blast_insights_report.html. Optional clean removes temporary convert/cluster folders; BLAST and reports are kept.
Pipeline Schema
{
"pipeline": "metabarcoding_single_read_blast",
"version": "1.0.0",
"stages": [
{ "id": "convert", "tool": "format_converter" },
{ "id": "cluster", "tool": "sequence_cluster", "optional": true },
{ "id": "blast", "tool": "blastn", "orchestration": "classic|agentic" },
{ "id": "enhance", "tool": "enhanced_blast_processor" },
{ "id": "arbitrate", "tool": "ml_arbitration" },
{ "id": "report", "tool": "blast_insights_report" }
],
"outputs": [
"03_blast/combined_enhanced.csv",
"03_blast/combined_arbitration.csv",
"blast_insights_report.html"
]
}Unlimited custom pipelines
Create as many pipelines as you need. Insert your own tools and code, run catalog discovery/install, and let the format adapter insert converters between steps. Chained I/O is validated at build and run time.
Saved params & job queue
Defaults save to ~/stDatApp/pipeline/custom/. Each run writes a timestamped folder with step outputs and a params snapshot for reproducibility.
Agentic BLAST
Agentic BLAST is a parallel orchestration path alongside Classic blast_package. It improves large-workload throughput (~90% faster than NCBI BLAST for supported cases) through GenBank enrichment/validation, accession caching, and ML-guided volume clustering — without exposing proprietary cloud internals.
Architecture
UI settings → desktop bridge → local pipeline: ingest → job computer → BLAST agents → collector ∥ GenBank agent → annotated hits.
Enrichment & cache
GenBank gather validates organism/gene context. Shared accession cache skips redundant Entrez fetches for reproducible re-runs.
Modes
Classic (default/fallback), Agentic, and Agentic Stream (warm-start volume skip / ML accept). Select in BLAST Configurations or Analysis Explorer.
Local Agentic runs require an Existing/External multi-volume BLAST DB. Classic remains switchable if Agentic cannot resolve binaries or databases.
igtools & Self-Serve Bench Tools
The igtools package powers Analysis Explorer discovery, Entrez arbitration, Resolve-G enrichment, DeepLigate read prep, and related self-serve bench utilities exposed through Data Tools in the desktop app.
Analysis Explorer
Discovery pipelines, sampling, assessment queues, and BLAST runner modes (Classic / Agentic) for iterative barcode and corpus exploration.
Resolve-G
BLAST parsing and taxonomy enrichment shared by pipelines and standalone enhanced-result workflows.
DeepLigate
ML-assisted adapter-oriented read preparation for large FASTQ workloads with structured reporting.
Data Tools (bench)
Format Converter, Sequence Cluster, Sequence Analyzer, Enhanced Report Generator, NL-SQL (when enabled), Dashboard Dataset Builder.
Geneie & EcoTracks
Workflow & query assistant
Sidekick for NCBI query generation/optimization, database help, troubleshooting, and board/species-list actions. Available from the main sidebar and Nexus-G Home.
- · Generate / Optimize Query
- · Database Help & Troubleshoot
- · Add Query / Create Species List / Add Board
Interactive Dashboard assistant
Answers questions about the selected dashboard dataset — top organisms, gene names (when GenBank features exist), length ranges, and summaries. Optionally uses GenBank context from pipeline dataset creation.
Advanced Aegis-G Hub also includes Pipeline Agent and Tool Install Agent alongside Geneie and EcoTracks.
Cloud
Cloud in Nexus-G means three things: your data credentials & work bucket (BYOS), the deployment catalog for org Batch resources, and compute that can shift heavy BLAST to AWS Batch / GCP when enabled.
BYOS storage
Connect AWS S3, GCP Buckets, or Azure Blob inside the app shell. Work bucket resolution uses cloud_settings, env overrides, then storage_settings.
AWS Batch BLAST
Stage inputs/DBs on S3, submit container workers, download results. Desktop checklist: Cloud Configuration + App Settings Storage/Compute + Batch queue/definition.
NG-Cloud DBs
Advanced tier streams hosted NCBI databases to local workers with egress caps; catalog refresh wires available DB keys for selection.
User Guides
Install, Home, and core engines
Portal signup through Tools Install, Cloud / App Settings, Agentic BLAST benchmarks, plans, workspace layout, and troubleshooting. PDF version 2026.1.
Download PDFBoards, GenBank & downloads
Organize species boards, queue NCBI Entrez downloads, and manage GenBank/FASTA datasets under ~/stDatApp.
Download PDFTaxonomic classification
Kraken2 packaging, Rapid Index, Bowtie, and Minimap2 workflows with consistent reporting outputs.
Visualization & reporting
Interactive dashboards, BLAST insight HTML reports, and publication-ready exploration of enhanced results.
Self-serve bench utilities
Format converter, sequence clustering, sequence analyzer, enhanced report generator, and dashboard dataset builder.
The Agentic Genomic Intelligence Engine Reference Guide
This guide details how to configure and execute Agentic BLAST within the Nexus-G platform, explaining how its adaptive, volume-aware search runners dynamically optimize resource scaling and throughput without altering underlying biological data
Download PDFThe Agentic Genomic Intelligence Engine Reference Guide
This document serves as the definitive parameter dictionary for Nexus-G BLAST, detailing configuration thresholds, DNA tasks, taxonomic exclusions, and sequence length filters to ensure precise, high-throughput search execution.
Download PDFTechnical Guides
Workspace layout
~/stDatApp for boards, ncbi_downloads, blast_results, db, workflow_results, pipeline runs, and appLogs.
Configuration
BLAST binaries path, NCBI email/API key, cloud provider/region/credentials, and storage work bucket.
Executor & smoke tests
Workflow executor module for custom pipelines; Agentic BLAST dry-run smoke without BLAST+ binaries.
Quickstart — executor & Agentic dry-run
# Run a saved custom pipeline via the workflow executor
python -m giat_package.workflow.workflow_executor \
--pipeline ~/stDatApp/pipeline/custom/<pipeline_id>.json
# Agentic BLAST (local) — dry-run smoke without BLAST+ binaries
GIAT_AGENTIC_BLAST_DRY_RUN=1 \
PYTHONPATH=. python scripts/smoke_agentic_blast_v2.py \
--dry-run --work-dir /tmp/ab-smokeTraining Courses
Opening soon.
Nexus-G 101
8 modules · 45 min
Genomics Pipelines
12 modules · 90 min
Agentic Workflows
10 modules · 120 min
Cloud & Compliance
6 modules · 70 min