qiongli

Qiongli (穷理)

Qiongli is a contract-driven academic workflow system for Codex, Claude Code, and Gemini. It covers installation, task planning, literature work, manuscript production, and strict Stage-I research code execution under one canonical workflow contract.

Why The Name

Qiongli is the public name of the project, from the Chinese 穷理: to pursue the underlying principle of a question until its logic, evidence, and limits are clear. For an academic workflow system, the name points to the work this repository is meant to support: not just producing text, but tracing a research claim back through literature, method, code, critique, and reproducible artifacts.

The full system name is Qiongli Zhengche (穷理证澈). Zhengche (证澈) names the core methodology: make evidence chains, citation risk, assumptions, and claim boundaries transparent enough to audit. In practice, that means every workflow is tied to Task IDs, quality gates, and output paths under RESEARCH/[topic]/, rather than relying on ad hoc prompts.

Technical identifiers follow the public name: the plugin is qiongli, the portable skill package is qiongli-workflow, and the updater distribution is qiongli. Legacy aliases such as research-skills, rsk, and rsw remain available only for compatibility.

Features

• 📚 Systematic Literature Review - PRISMA 2020 compliant methodology
• 📖 Deep Paper Reading - Structured notes + BibTeX
• 🧪 Evidence Synthesis & Meta-analysis - Narrative / qualitative / quantitative pooling (PRISMA-aligned)
• 📝 Full Manuscript Drafting - Outline → draft → claim-evidence integrity → figures/tables
• 🧩 Study Design → Publication - Study design, ethics/IRB pack, submission prep, rebuttal workflow
• 🔍 Research Gap Identification - 5 types of academic gap analysis
• 🧠 Theoretical Framework Building - Concept relationship mapping
• ✍️ Academic Writing Assistance - Standard-compliant formatting
• 🧑‍⚖️ Multi-Persona Peer Review - Parallel, independent cross-reviews (Methodologist, Domain Expert, "Reviewer 2")
• 🔎 AI De-fingerprinting & Proofread - Multi-AI collaborative de-AI rewriting, similarity reduction, and final proofread
• 🚀 Strict Stage-I Academic Code Flow - I5 -> I6 -> I7 -> I8 with structured spec/plan/execute/review artifacts and targeted follow-up
• 🎤 Academic Presentation - Story arc design → slide content spec → output to Slidev (scholarly theme), LaTeX Beamer, or PPTX
• 🛡️ Iterative Critique Loop (Red Teaming) - AI self-review and Socratic questioning to continuously narrow down and refine outputs
• 🤖 Multi-Model Collaboration - Codex + Claude + Gemini coordination across research stages
• ⚡ Token Optimized - Layered skills architecture (~90% reduction)

[!WARNING] Full functionality requires a real Python runtime plus all three model CLIs in PATH: python3, codex, claude, and gemini. You also need the matching runtime authentication. codex can run with OPENAI_API_KEY or an existing ChatGPT/Codex login, claude uses ANTHROPIC_API_KEY, and Gemini direct mode requires non-interactive auth such as GEMINI_API_KEY or Vertex env auth. Google-login-only Gemini automation should use the resident broker path described in docs/guide/multi-agent.md. Without them, you can still install assets and use shell qiongli check|upgrade|align, but doctor, validators, tests, and the full orchestrator flow will be partial or unavailable.

Design Lineage And Related Projects

This repository is not built in isolation. Two external projects are especially relevant to its design direction: