searxng

</div>

DARIS

Research workflow orchestration system — from topic to structured output.

![TypeScript](https://www.typescriptlang.org/) ![Node.js](https://nodejs.org/) ![License](LICENSE) ![Stars](https://github.com/disdorqin/DARIS/stargazers) ![Status](https://github.com/disdorqin/DARIS)

---

DARIS (DAily Research & Intelligent System) orchestrates a configurable multi-agent pipeline that moves from a research topic to a structured literature survey, experiment design, and draft output. It treats the mechanical stages of research — literature retrieval, hypothesis formulation, experiment tracking, and knowledge asset production — as programmable pipeline steps.

It is not a replacement for scientific judgment. It is a scaffold for the repetitive parts: finding papers, tracking experiments, and formatting results.

Why

Academic research involves well-defined stages that are repeated across every project:

Literature search and ingestion
Relevance scoring and summarization
Hypothesis formulation and experiment design
Experiment execution and metric tracking
Drafting slides, reports, and papers

DARIS defines each stage as a configurable pipeline step, with YAML/JSON configuration determining scope and behavior.

Features

Config-Driven Orchestration — a single config file defines the entire workflow: topic, search sources, experiment parameters, output format
Multi-Agent Pipeline — specialized agents for literature ingestion, hypothesis design, experiment execution, and knowledge export
Literature Workflow — automated ingestion from configured sources with relevance ranking and summarization
Experiment Tracking — benchmark tracking with configurable metrics and baseline comparison
Knowledge Asset Export — auto-generates structured outputs: slides, reports, and draft summaries

Architecture

The pipeline is organized into numbered stages in 1_config/ through 8_knowledge_asset/:

| Module | Directory | Purpose | |---|---|---| | Config | 1_config/ | Workflow configuration | | Agent System | 2_agent_system/ | Multi-agent orchestration | | Literature | 3_literature_workflow/ | Paper search and ingestion | | Hypothesis | 4_research_hypothesis/ | Hypothesis design | | Code Base | 5_code_base/ | Experiment code templates | | Execution | 6_experiment_execution/ | Benchmark runner | | Monitor | 7_monitor_system/ | Metrics and logging | | Knowledge | 8_knowledge_asset/ | Export pipeline |

Prerequisites

Node.js 18 or later
npm or pnpm
(Optional) Python 3.10+ for experiment execution
LLM API key (for hypothesis and drafting agents)

Quick Start

# Clone and install
git clone https://github.com/disdorqin/DARIS.git
cd DARIS
npm install  # or: pnpm install

# Configure your research workflow
cp 1_config/example.json 1_config/config.json
# Edit config.json with your research topic and parameters

# Run the workflow
npm start

Example Usage

// Define a research workflow
const workflow = {
  topic: "retrieval-augmented generation for scientific literature review",
  hypothesis: "RAG improves citation recall over sparse retrieval in domain-specific lit review",
  experiment: {
    method: "RAG with dense retriever",
    baseline: "BM25 sparse retrieval",
    metrics: ["recall@10", "precision@5", "MAP"],
    dataset: "domain_literature_corpus.csv"
  }
}

await daris.run(workflow)
// → Ingests relevant literature
// → Generates experiment design
// → Runs benchmark comparison
// → Exports structured report and slides

Roadmap

[x] Multi-agent pipeline orchestration
[x] Literature ingestion and ranking
[ ] Hypothesis auto-generation with LLM
[ ] Experiment auto-execution and metric tracking
[ ] Draft paper and slide generation
[ ] Plugin architecture for custom literature sources

Tech Stack

TypeScript · Node.js · Python · OpenClaw · LLM APIs

Contributing

See CONTRIBUTING.md. Issues and PRs are welcome.

License

MIT — see LICENSE.

Summary

Install to Claude Code