Installation

Bohr ships prebuilt standalone bundles for supported macOS, Linux, and Windows machines. You do not need to deploy a running app before changing these commands; you only need the installer scripts and release assets published at reachable URLs.

macOS / Linux:

curl -fsSL https://raw.githubusercontent.com/Dhruvil7694/BohrAI/main/website/public/install | bash

Windows (PowerShell):

irm https://raw.githubusercontent.com/Dhruvil7694/BohrAI/main/website/public/install.ps1 | iex

The one-line installer fetches the latest tagged release. To pin a version, pass it explicitly, for example curl -fsSL https://raw.githubusercontent.com/Dhruvil7694/BohrAI/main/website/public/install | bash -s -- 0.2.18.

The installer downloads a standalone native bundle with its own Node.js runtime.

To upgrade the standalone app later, rerun the installer. bohr update only refreshes installed Pi packages inside Bohr's environment; it does not replace the standalone runtime bundle itself.

For internal client deployments, do not hardcode the GitHub Raw URLs above. Instead set the deployment variables in .env.example such as PUBLIC_BOHR_SITE_URL, PUBLIC_BOHR_INSTALL_BASE_URL, BOHR_RELEASES_LATEST_URL, and BOHR_INSTALL_BASE_URL, then publish the same scripts and release bundles on the client's own host. The website homepage already reads those env-based install URLs at build time; the docs page should be published from the same deployment so users copy the correct host-specific command.

Local models are supported through the custom-provider flow. For Ollama, run bohr setup, choose Custom provider (baseUrl + API key), use openai-completions, and point it at http://localhost:11434/v1.

Skills Only

If you want just the research skills without the full terminal app:

macOS / Linux:

curl -fsSL https://raw.githubusercontent.com/Dhruvil7694/BohrAI/main/website/public/install-skills | bash

Windows (PowerShell):

irm https://raw.githubusercontent.com/Dhruvil7694/BohrAI/main/website/public/install-skills.ps1 | iex

That installs the skill library into ~/.codex/skills/bohr.

For a repo-local install instead:

macOS / Linux:

curl -fsSL https://raw.githubusercontent.com/Dhruvil7694/BohrAI/main/website/public/install-skills | bash -s -- --repo

Windows (PowerShell):

& ([scriptblock]::Create((irm https://raw.githubusercontent.com/Dhruvil7694/BohrAI/main/website/public/install-skills.ps1))) -Scope Repo

That installs into .agents/skills/bohr under the current repository.

These installers download the bundled skills/ and prompts/ trees plus the repo guidance files referenced by those skills. They do not install the Bohr terminal, bundled Node runtime, auth storage, or Pi packages.

Token usage & caveman mode

Long multi-agent runs burn context fast. Bohr mitigates this in several ways:

• Bundled caveman skill (skills/caveman/caveman.md) — instructs terse, substance-preserving replies. The skill description targets roughly 75% fewer tokens on assistant output versus verbose defaults when the model follows the style; actual savings vary by model and task.
• Paper-generator / Pi subagent path — prepends short [CAVEMAN MODE: lite|full|ultra] instructions, passes minimal JSON context (slug, phase, filenames), and uses file handoffs instead of dumping large intermediates into the prompt.
• Configuration — optional CAVEMAN_MODE_DEFAULT and ENABLE_TOKEN_OPTIMIZATION in .env.example tune subagent compression when using the CLI (dotenv-loaded). See the repo docs: Token optimization & caveman mode.
• Cost planning — illustrative token and dollar scenarios by provider family (OpenAI, Anthropic, Gemini, Grok, DeepSeek, Chinese APIs), tool fees, and self-host FLOPs intuition: Cost & token estimation.

What you type → what happens

$ bohr "what do we know about scaling laws"
→ Searches papers and web, produces a cited research brief

$ bohr deepresearch "mechanistic interpretability"
→ Multi-agent investigation with parallel researchers, synthesis, verification

$ bohr lit "RLHF alternatives"
→ Literature review with consensus, disagreements, open questions

$ bohr audit 2401.12345
→ Compares paper claims against the public codebase

$ bohr replicate "chain-of-thought improves math"
→ Replicates experiments on local or cloud GPUs

Workflows

Ask naturally or use slash commands as shortcuts.

| Command | What it does | | --- | --- | | /deepresearch <topic> | Source-heavy multi-agent investigation | | /lit <topic> | Literature review from paper search and primary sources | | /lit-review <topic> | Full literature pipeline with themes, contradictions, gaps, and validator pass | | /paper <topic-or-artifact> | Publication-grade paper pipeline with method, figures, and compliance | | /paper-pro <topic-or-artifact> | Professional paper pipeline with evidence map, adversarial checks, citations, and provenance | | /review <artifact> | Simulated peer review with severity and revision plan | | /audit <item> | Paper vs. codebase mismatch audit | | /replicate <paper> | Replicate experiments on local or cloud GPUs | | /compare <topic> | Source comparison matrix | | /draft <topic> | Paper-style draft from research findings | | /autoresearch <idea> | Autonomous experiment loop | | /watch <topic> | Recurring research watch | | /hypothesis <topic-or-artifact> | Ranked, testable hypotheses with explicit falsifiers | | /contradict <claim-or-artifact> | Adversarial contradiction analysis for conclusions | | /evidence-score <artifact-or-topic> | Claim-level evidence strength scoring | | /experiment <hypothesis-or-goal> | Minimal decisive experiment workflow | | /citation-check <artifact> | Sentence-level citation integrity validation | | /memory-log <project-or-topic> | Durable memory consolidation for long-running work | | /plan-research <topic-or-goal> | Planner workflow for agent ordering and stop criteria | | /knowledge-graph <artifact-or-topic> | Build reusable knowledge graph for cross-topic reasoning | | /reasoning-validate <artifact> | Validate logic chain from evidence to conclusions | | /iterate <topic-or-artifact> | Iterative loop controller until confidence/contradiction targets | | /outputs | Browse all research artifacts |

Agents

Ten bundled research agents, dispatched automatically.

• Researcher — gather evidence across papers, web, repos, docs
• Reviewer — simulated peer review with severity-graded feedback
• Writer — structured drafts from research notes
• Verifier — inline citations, source URL verification, dead link cleanup
• Hypothesis — generate testable hypotheses and ranked falsifiers
• Contradiction — find counter-evidence, edge cases, and fragile assumptions
• Evidence Scorer — score claim support quality with a transparent rubric
• Experiment — design and run minimal experiments to reduce uncertainty
• Citation Integrity — validate claim-to-citation alignment at sentence level
• Memory — maintain durable project memory across sessions
• Research Planner — orchestration brain for agent selection, ordering, and stop rules
• Knowledge Graph — convert findings into reusable entities and relations
• Reasoning Validator — detect invalid inference chains and logic drift
• Iteration Controller — drive multi-loop convergence with explicit thresholds
• Literature Collector — collect paper candidates and metadata
• Literature Quality — score/filter papers by relevance and recency
• Literature Synthesizer — cluster papers into structured themes
• Literature Contradiction — detect conflicting claims across papers
• Literature Gap — identify missing areas and open problems
• Literature Review Writer — write survey-style thematic review
• Literature Review Validator — validate completeness, flow, and perspective balance
• Method Math — formalize notation, equations, and derivation consistency
• Figures Tables — produce publication-ready visuals and captions
• Paper Compliance — enforce manuscript structure and submission-style checks

Skills & Tools

• AlphaXiv — paper search, Q&A, code reading, annotations (via alpha CLI)
• Docker — isolated container execution for safe experiments on your machine
• Web search — Gemini or Perplexity, zero-config default
• Session search — indexed recall across prior research sessions
• Preview — browser and PDF export of generated artifacts
• Modal — serverless GPU compute for burst training and inference
• RunPod — persistent GPU pods with SSH access for long-running experiments

How it works

Built on Pi for the agent runtime, alphaXiv for paper search and analysis, and CLI tools for compute and execution. Capabilities are delivered as Pi skills — Markdown instruction files synced to ~/.bohr/agent/skills/ on startup. Every output is source-grounded — claims link to papers, docs, or repos with direct URLs.

Contributing

See CONTRIBUTING.md for the full contributor guide.

git clone https://github.com/Dhruvil7694/BohrAI.git
cd BohrAI
nvm use || nvm install
npm install
npm test
npm run typecheck
npm run build

Docs · Proprietary License