Packages

Core and optional Pi packages bundled with Feynman, and how to manage them.

Feynman is built on the Pi runtime and uses curated Pi packages for its capabilities. Core packages are installed by default to keep first-run setup fast. Optional packages can be added on demand with feynman packages install.

Core packages

These packages are installed automatically with every Feynman installation. They provide the foundation for all research workflows and the REPL experience.

Package	Purpose
@companion-ai/alpha-hub	alphaXiv integration for paper search, discussion threads, and citation metadata
pi-subagents	Parallel agent spawning for literature gathering and task decomposition. Powers all multi-agent research workflows
pi-btw	Fast side-thread conversations without interrupting the main research run
pi-docparser	Parse PDFs, Office documents, spreadsheets, and images for content extraction
pi-web-access	Web browsing, GitHub access, PDF fetching, and media retrieval
pi-markdown-preview	Render Markdown and LaTeX-heavy research documents as polished HTML/PDF
@walterra/pi-charts	Generate charts and quantitative visualizations from data
pi-mermaid	Render Mermaid diagrams in the terminal UI
@aliou/pi-processes	Manage long-running experiments, background tasks, and log tailing. Powers /ps
pi-zotero	Integration with Zotero for citation library management
@kaiserlich-dev/pi-session-search	Indexed session recall with summarize and resume UI. Powers /search
pi-schedule-prompt	Schedule recurring and deferred research jobs. Powers the /watch workflow
@samfp/pi-memory	Automatic preference and correction memory across sessions
@tmustier/pi-ralph-wiggum	Long-running autonomous agent loops. Powers /autoresearch

Core packages are updated together when you run feynman update. You do not need to manage them individually.

Optional packages

Install these on demand with feynman packages install <preset>. They extend Feynman with capabilities that not every user needs.

Preset	Package	Purpose
generative-ui	pi-generative-ui	Interactive Glimpse UI widgets for rich, HTML-style output in the REPL

The generative-ui preset can also be installed using the alias ui: feynman packages install ui.

Managing packages

List packages

Show all available packages and their install status:

feynman packages list

This displays core packages (always installed) and optional presets with an (installed) marker for any that are already active.

Install optional packages

feynman packages install generative-ui

Update packages

Update all installed packages to their latest versions:

feynman update

Update a specific package by name:

feynman update pi-subagents
feynman update pi-web-access

Running feynman update without arguments updates everything. Updates are safe and preserve your configuration and session data.

What packages enable

Research workflows (pi-subagents)

The pi-subagents package enables Feynman to spawn and coordinate parallel research agents. Every multi-agent workflow — /deepresearch, /lit, /review, /audit, /replicate, /compare, and /draft — depends on this package to delegate work across researcher, reviewer, writer, and verifier subagents simultaneously.

Web access and paper retrieval (pi-web-access, @companion-ai/alpha-hub)

pi-web-access gives Feynman access to the web, GitHub repositories, and arbitrary PDFs during research. @companion-ai/alpha-hub adds alphaXiv-specific paper search, discussion thread access, and citation metadata. Together they form the primary source-gathering layer for all research workflows.

Document parsing (pi-docparser)

pi-docparser lets Feynman ingest PDFs, Word documents, spreadsheets, and images as input to research tasks. Drop a PDF into your working directory and Feynman can read and cite it.

Preview and export (pi-markdown-preview)

pi-markdown-preview renders research artifacts — including LaTeX equations — as HTML or PDF directly from the REPL via the /preview command. Requires pandoc, which can be installed via feynman --setup-preview.

Session search (@kaiserlich-dev/pi-session-search)

Enables the /search command for indexed recall of prior session transcripts. Useful for finding past research, resuming context from previous sessions, or surfacing earlier findings without re-running a workflow.

Memory (@samfp/pi-memory)

Automatically records your preferences and corrections across sessions. If you consistently prefer a particular output format or correct the same terminology, memory surfaces those preferences in future sessions.

Background processes (@aliou/pi-processes)

Provides the /ps command for inspecting active background tasks, running experiments, and scheduled follow-ups. Used by /autoresearch and /watch to manage long-running work.

Scheduled research (pi-schedule-prompt)

Powers the /watch workflow by creating durable scheduled or recurring prompts. When you set a watch on a topic, pi-schedule-prompt ensures follow-up sweeps happen at the configured frequency — not just when you remember to ask.

Autonomous loops (@tmustier/pi-ralph-wiggum)

Powers the /autoresearch command. Manages the iteration state, benchmark command execution, experiment logging, and keep/revert decisions across many optimization cycles.

Generative UI (pi-generative-ui) — optional

Adds interactive Glimpse UI widgets to REPL output. Useful when you want richer, HTML-style interactive displays rather than plain Markdown. Install with feynman packages install generative-ui.