arm2arm/AstroAgentAssistant - Page 2

drp-hub

Digital Research Product Hub — federated infrastructure for reproducible science in particle physics & astrophysics. PUNCH4NFDI consortium project.

agentbench-benchmarking

Complete guide to running AgentBench FC benchmarks with LLM agents

agentbench-dbbench-benchmark

Run AgentBench DBBench benchmark against LLMs and generate standardized comparison dashboard. Use when testing any model on DBBench (SQL generation task).

agentbench-ollama-benchmarking

Benchmark LLMs using AgentBench suite with local Ollama models

agent-benchmark-workflows

Class-level guide for running AI agent benchmarks — Terminal-Bench, SWE-Bench, DBBench, and related evaluation frameworks. Includes platform compatibility, container setup, and result interpretation.

agentic-benchmarks

Overview and comparison of benchmarks that evaluate LLMs as autonomous agents: GAIA, WebArena, OSWorld, ToolBench, LiveAgentBench, BrowseComp, WebVoyager, Tau-Bench. Includes guidance on running AgentBench DBBench SQL generation and selecting the right…

astro-llm-research

Procedures, queries, pitfalls, and reproducible recipes for finding, verifying, and harvesting papers on large language models and agentic/LLM-based data analysis applied to galaxy formation, cosmology, and related astronomical subfields.

drp-agentic-assistants

Integrate agentic assistants into DRP / reproducibility papers, slides, and architecture with grounded, non-hype framing; especially Hermes + DRP-Hub + REANA style workflows.

drp-card-workflow

Full lifecycle for creating reproducible DRP cards: scaffolding with reana.yaml, Beamer/LaTeX or Python workflows, GitLab publishing, and DRP-Hub registration via REST API. Covers the complete path from project creation to share link.

drp-paper-workflow

Class-level skill: reproducible paper build, DRP-Hub screenshot capture, BibTeX sweep, and final packaging workflow for agentic-astronomy papers. Use when preparing/finishing a DRP-style manuscript and integrating live registry screenshots or API extracts.

drphub-anon-access

Query DRP Hub public cards without auth.

latex-compilation-pitfalls

Common LaTeX compilation errors, debugging patterns, and fixes — especially for agent-generated Beamer decks, TikZ/pgfplots variable mismatches, block/column nesting issues, and systematic error diagnosis workflows.

latex-paper-workflow

Complete guide to writing, iterating, compiling, and packaging LaTeX research papers. Covers monolithic paper generation, multi-round improvement, merging papers, figure generation, MNRAS/journal submission formatting, and portability fixes on Ubuntu/Debian.

llm-agent-benchmarking

Run LLM agent benchmarks (AgentBench FC) to evaluate multi-turn agent performance

llm-benchmarking

Systematic LLM benchmarking on AgentBench, DBBench, and other standardized evaluations

nasa-ads-citations

Guide for verifying and extracting BibTeX metadata from NASA Astrophysics Data System (ADS). Includes verified API endpoints, search strategies, and CrossRef fallback.

terminal-bench

Benchmarking LLM agents in terminal environments using Harbor and Terminal-Bench datasets (2.0, 2.1, Pro). Covers installation, custom endpoint configuration, oracle smoke tests, dataset selection, ARM64 pitfalls, and result interpretation.

dtwin

Umbrella for the dt4acc Digital Twin (particle accelerator simulation) covering Apptainer builds, host-side smoke tests, EPICS and TANGO runbooks, container troubleshooting, and burn-in testing.

skills-repo-maintenance

Maintain the AstroAgentAssistant-style public skills repository by auditing secrets, syncing README coverage, resolving vague issues into concrete skill changes, and keeping taxonomy consistent.

deepseek-harness-dsh

Use when running coding tasks via dsh (DeepSeek Harness).

blender-rendering

Headless Blender 5 rendering from Docker with Cycles CPU, material creation patterns, and S3 delivery. Use when rendering 3D scenes via Blender in container.

starhorse-access

Access StarHorse data products including SHboost-2024 and the SH21 EDR3 catalog via gaia.aip.de TAP.

reana-aip

Author, validate, and run REANA workflows under AIP conventions — canonical reana.yaml template, the approved environment images, mandatory reana-client validate, submit/monitor recipe, GitLab + DRP-card hand-off.

drphub-cards

Manage DRP Hub Digital Research Products via the production REST API at drp-term.kube.aip.de/api/v1/. Supports full CRUD, clone, maturity, publish, audit, lineage, human-review, bookmarks, likes, sharing, and SSE event streaming.

reana-serial-python

When explicitly asked, use it to build a REANA serial workflow (Python analysis on remote data, Parquet cache, PNG outputs).

kanban-worker

Complete guide to the Hermes Kanban system: the orchestrator decomposition playbook, specialist roster conventions, anti-temptation rules, and the deeper worker pitfalls, examples, and edge cases. The core lifecycle is auto-injected into every worker's system…

data-aip-de-s3

Work with data.aip.de and S3-backed datasets using reproducible local caching, Dask-first reads for huge data, and plotting workflows that scale to large astronomy catalogs.

gaia-dr3-daiquiri-rest

FALLBACK Gaia access (prefer gaia-dr3-tap-query). Query Gaia DR3 at gaia.aip.de via its Daiquiri REST API for full-table COUNTs and very large async scans — CSRF handling, async jobs, queue names, JSON result fetching. Access 1.8 billion sources with ~153…

gaia-dr3-plot-with-dust

Retrieve nearby Gaia DR3 stars (via the AIP TAP service) and produce a two-panel research-paper figure — RA/Dec on the left, a Galactic XY projection overlaid with the SFD all-sky dust-extinction map on the right (dustmaps package).

gaia-dr3-tap-query

Query Gaia DR3 at gaia.aip.de via TAP/pyvo — the DEFAULT way to get Gaia data. ADQL queries, uniform random subsampling, Parquet caching, and ready-made sky / Galactic-XY plots. Access 1.8 billion sources.

rave-dr6-shboost-distance-query

Query RAVE DR6 stars with SHboost24 distances via Gaia source_id crossmatch

rave-dr6

Query the RAVE DR6 catalog at https://www.rave-survey.org/tap/ using pyvo (TAPService.run_sync). Access stellar parameters, Gaia cross-matches, distances, and Galactic coordinates (l, b). Includes galactic and equirectangular projection plotting recipes.

rave-dr6-starhorse-access

Query RAVE DR6 via TAP and crossmatch with SHboost24 distances for nearby star analysis.

docs-mcp-at-aip

Access the AIP documentation MCP server at https://docs-mcp-server.kube.aip.de. Search, scrape, and fetch documentation for 15+ indexed libraries including reana, pandas, snakemake, dask, unsloth, and more. HTTP POST + SSE, internal network only.

workspace-backup

Back up the REANA workspace (data, code, figures, chat/agent state, skills) to a downloadable tar.gz — recipes and results, never venv binaries. Restore = untar; skill venvs re-provision themselves.

gaia-aip-data-access

Use when accessing Gaia data through AIP services, choosing between Gaia@AIP TAP/ADQL/pyvo and Daiquiri REST/PostgreSQL APIs, handling CSRF/queue details, checking endpoint health, caching results, and falling back when services are unavailable.

gaia-aip-de-adql

Query gaia.aip.de using ADQL/TAP with reproducible query capture and astronomy-aware caveats.

gaiadr3-aip-de-adql

Query the Gaia DR3 PostgreSQL database at gaia.aip.de via its Daiquiri REST API. Includes CSRF handling, async job submission, queue names, and result fetching. Access 1.8 billion sources with ~153 columns.

gaiadr3-aip-query-api

Query the Gaia DR3 PostgreSQL database at gaia.aip.de via its Daiquiri REST API. Includes CSRF handling, queue names, and result fetching.

rave-dr6-data-access

Use when querying, caching, or plotting RAVE DR6 data via its TAP/ADQL service, including Gaia cross-match tables, recent observations, stellar parameters, southern-sky coverage caveats, VOTable parsing, and publication-ready visualizations.