Generates structured summaries of library files — executive summary, key points, and claims — using the configured LLM. Runs automatically during the upload pipeline.

Fast read-only search agent that finds wiki pages, library files, and knowledge base content. Used when Hypatia needs to look up existing content before answering.

Periodic reflective pass over Hypatia's memory files — merges new signal from recent sessions into durable topic files, converts relative dates to absolute, and prunes stale or contradicted facts.

After each conversation, updates structured session notes with file paths, decisions, current state, and technical details — preserving section headers and italic template descriptions exactly.

When context fills up, generates a structured continuation summary (task overview, current state, discoveries, next steps) enabling seamless handoff to a new context window without losing progress.

Adversarially tests claims, facts, and assertions in library documents. Goes beyond happy-path confirmation — probes edge cases, contradictions, and unsupported statements, issuing PASS / FAIL / PARTIAL verdicts.

Finds relevant prior conversations based on user queries — matches against tags, titles, branch names, summaries, and transcript content. Requires conversation indexing infrastructure.

Summarizes scraped web pages concisely with quote limits and license compliance. Ready to activate once SearXNG and the scraper are wired in for research workflows.