informR

Project briefings for AI agents. Part of cerebro, the AI agent toolchain for R.

informR ("to inform") pulls together ontology relations, git history, Claude Code memory, and feature hubs into concise markdown context files. Your agent reads these instead of scrounging through scattered project files.

One dependency: pensar.

Install

remotes::install_github("cornball-ai/informR")

What it does

briefing() bolts together ontology identity, sibling packages, Claude Code memory, and recent commits into a single markdown file per project.

informR::briefing("saber")
# Writes ~/.cache/R/informR/briefs/saber.md

heartbeat() scans every git repo under ~/ and produces a cross-project activity summary. Good for weekly check-ins.

informR::heartbeat(days = 7)
# Writes ~/.cache/R/informR/briefs/_heartbeat.md

hub() and hubs() manage feature hub files: small markdown docs that map a concept to code locations via [[project::function]] wikilinks.

informR::hub("embeddings", "## Status\nWired up in [[saber::symbols]]")
informR::hubs()
#>        name                                    file              links
#> 1 embeddings ~/.cache/R/informR/hubs/embeddings.md saber::symbols

write_instructions() generates a quick-reference markdown file for using pensar from Claude Code.

informR::write_instructions("INSTRUCTIONS.md", index_path = pensar::index_path())

Cache layout

Everything lands in ~/.cache/R/informR/:

briefs/   -- per-project briefings, heartbeat
hubs/     -- feature hub markdown files

Sister packages

Package	Purpose	Repo
saber	AST symbols, blast radius, pkg introspection	cornball-ai/saber
pensar	Concept graph / ontology	cornball-ai/pensar
mirar	Runtime inspection of live R sessions	cornball-ai/mirar
llamaR	LLM interface	cornball-ai/llamaR

License

Apache-2.0