{"repo":"harper-nottinghill/julia-reader","free":true,"listed":false,"github":"https://github.com/harper-nottinghill/julia-reader","clone":"git clone https://github.com/harper-nottinghill/julia-reader.git","description":"Progressive agentic reading harness: sentence maps, bounded chunks, Chronicle Markdown books, packets, validation. Python CLI + Next.js demo. MIT.","language":"Python","stars":18,"topics":["agents","chronicle","llm","machine-reading","markdown","mit","openai","python","summarization"],"license":"MIT","category":"ai-agents","readme_excerpt":"Julia Reader Harness This repo is a standalone progressive Reader harness for the terminal: sentence splitting, dynamic chunking (≤2000 token estimates), per-chunk LLM summaries with a live evolving understanding , Lake Strings / breaks / packet JSON, Markdown Chronicle output, and validation. It ships as stdlib-only Python plus your OpenAI-compatible HTTP endpoint—no heavyweight framework imports required. For visitors (GitHub landing) What it does. Julia Reader ingests long, messy source text (transcripts, logs, notes) and walks it the way a careful human would: stable sentence maps , bounded chunks , soft subject breaks , and a running synthesis that updates as each chunk is read. It then plans a book , writes Markdown pages (index, preface, chapters), and seals a Harper-style packet (indexes, shard metadata) with a validation report . The output is a reproducible Chronicle folder under reader/ / , not a single blob of prose. Research angle. We treat “reading” as an agent loop with layered limbs : some steps are cheap and deterministic (normalize, split, lake strings, chunking, validation), others are LLM limbs (per-chunk JSON summaries, live understanding). The lake layer keeps structured signals about sentences—enough texture for downstream tools without rereading the whole source every time. The Chronicle is deliberately book-shaped so humans and other agents can browse, cite, and extend understanding instead of drowning in context windows. Why open source. We want othe","default_branch":null,"files":null,"tree":[],"storefront":"/r/harper-nottinghill","claimed":false,"request_supported":{"post":"https://gitbuyer.com/r/harper-nottinghill/julia-reader/request-supported","requests":0},"note":"indexed from public GitHub; nothing is for sale on this page. Clone it from GitHub. Paid listings live at /search."}