Documentation Index

Fetch the complete documentation index at: https://hubify.com/docs/llms.txt

Use this file to discover all available pages before exploring further.

​

Changelog

All notable changes to Hubify Labs are documented here.

​

v1.3.0, 2026-04-30

​

New Features

• Captain Recent Lab Events, Live stream of every agent action in your lab, sourced from the agentEvents ledger. Cross-checked against every Convex mutation that should leave a trace.
• Live Lab Activity heatmaps, 24h heatmap above the fold on Captain View, plus a public version on every {slug}.hubify.app lab site.
• 15 new analytics panels, Agent/system split, LLM/GPU mix, anomaly by scale, novelty tiers, daily spend variance, backup age, review verdict trend, model confidence, standup day coverage, survey completeness, review confidence × verdict, comms 6w volume, experiment priority × success, discovery review coverage, sender distribution.
• Captain broadcast compose, Send a message to every agent in the lab from one composer.
• Lab goals in Captain View, Set and track lab-level objectives directly from the dashboard.
• Experiment topic cloud + figure caption bulk gen, Topic clustering across experiments and one-click captioning for paper figures.
• Skills catalog, 20 canonical agent skills shipped in hubify-labs/ package with README index.
• Per-lab chat dock memory, The orchestrator chat dock remembers position (left/right/bottom/hidden) per lab.
• Mintlify-powered docs, /docs/* is now served by Mintlify with full Tabs, CodeGroup, CardGroup, ParamField, and ResponseField support. New: per-page OG metadata, What’s new panel, Core Concepts TOC, expanded sidebar (features, integrations, MCP).

​

Improvements

• /app empty state is now an orchestrator chat with starter prompts and docs links, not a Welcome GUI.
• /sign-in covers 7 new error codes (verify_failed, github_missing_params, github_user_fetch, github_pro_store_failed, github_callback_failed, user_create_failed, not_signed_in) with a catch-all for unknown params.
• Settings → API Keys shows a red “missing key” CTA banner for unset ANTHROPIC_API_KEY and a presence badge for 12 server env vars via /api/env-check.
• Captain skeleton shimmer during initial Convex load instead of “No labs yet” or “No activity yet” flash.
• Honest empty states across Captain papers, graph view, and labs view, no synthetic fallback nodes.
• Sidebar /docs link in the /app footer for one-click reference jump.
• Marketing nav surfaces docs and lifts Live Lab Activity above the fold.

​

Fixes

• /docs/api dead link in Settings repointed to /docs/api/overview.
• 3 orphaned mdx slugs (api/files, api/sync, guides/local-agents) surfaced in Mintlify nav.
• 7 callsites silently dropped Newsreader to Georgia for paper drafts and abstracts; fixed by routing through the shared --serif token.
• Turbopack root pinned to project dir to fix tailwindcss resolution.
• --danger token aliased to --crit so 8 callsites resolve under both themes.

​

For contributors

• 328 stale color fallbacks and 239 stale font fallbacks dropped across src/ (cleaner DSI surface, easier theme audits).
• Mintlify renderer audit: 70 mdx files, 0 unhandled tags, 22 orphaned slugs catalogued.
• POLISH_LOOP autonomous cron retired; replaced by the on-demand 15-min loop with phase-advancement protocol.

---

​

v1.2.0, 2026-04-14

​

New Features

• MCP Server, Connect Claude Code, Cursor, and other MCP hosts to your lab. 18 tools, 6 resource types, 5 pre-built prompts.
• Cross-model peer review, Papers and experiment results are reviewed by multiple AI model providers (Claude, GPT, Gemini, Grok) to prevent echo chambers.
• Standup transcripts, All-hands standups generate structured transcripts with action items and escalations.
• HuggingFace integration, Push datasets, models, and Spaces to HuggingFace Hub directly from your lab.

​

Improvements

• Experiment resume from checkpoint now works across pod restarts
• Pod budget alerts include projected end-of-month spend
• Lab site analytics now show Google Scholar referrals
• CLI --json output is now consistent across all commands

​

Fixes

• Fixed pod idle detection failing when GPU utilization was exactly 0%
• Fixed paper compile timeout on papers with more than 15 figures
• Fixed agent metrics double-counting tasks reassigned after failure

---

​

v1.1.0, 2026-03-15

​

New Features

• Novelty scoring, AI-powered assessment of how novel experiment findings are compared to existing literature.
• Knowledge base search, Full-text search across all wiki entries with relevance ranking.
• Pod templates, Create reusable pod configurations for common experiment types.
• Paper claims table, Track every scientific claim with evidence links and verification status.

​

Improvements

• Captain View now shows paper readiness alongside experiments
• Agent activity feed is color-coded by event type
• Experiment outputs download 3x faster with parallel streams

​

Fixes

• Fixed QC gate false positives when R-hat was exactly 1.00
• Fixed lab site deploy failing when custom domain SSL was renewing
• Fixed CLI login flow on Windows WSL

---

​

v1.0.0, 2026-02-15

​

Initial Release

• Labs, Isolated research environments with agents, compute, and public sites
• Experiments, GPU-powered experiment execution with automatic QC gates
• Agents, Hierarchical multi-agent system (orchestrator, leads, workers)
• Papers, End-to-end paper pipeline from outline to PDF compilation
• Knowledge Base, Karpathy-style structured wiki with entities, concepts, sources
• Lab Sites, Public research websites at {slug}.hubify.app
• Captain View, Research command center dashboard
• CLI, Full-featured terminal interface with TUI mode
• REST API, Programmatic access to all lab resources
• RunPod integration, H100/H200/A100 GPU compute
• arXiv packaging, Submission-ready tarball generation

---

​

Roadmap

Planned features for upcoming releases. Items marked “partial” have existing scaffold in the codebase; items marked “planned” have no code yet.