Kaleidoscope

Academic Paper Intelligence Platform with Epistemic Analysis Engine

🇨🇳 中文文档  |  🇬🇧 English

---

Overview

Kaleidoscope is a full-stack research platform for discovering, ingesting, reading, and analyzing academic papers. It combines a Markdown-first storage approach with a powerful Epistemic Analysis Engine — over 2,075 specialized AI tools that detect cognitive biases, logical fallacies, rhetorical distortions, and reasoning failures in academic claims.

Key Features

• 📡 ArXiv Ingestion — Batch-fetch papers across categories, auto-convert to Markdown
• 📖 Markdown Reader — Read papers in-browser with table of contents, font controls, and section navigation
• 📊 Analytics Dashboard — Library insights: timeline, categories, top authors, keyword cloud, citation network
• 🔍 Multi-modal Search — Keyword, semantic, and claim-first search across your library
• 🧠 Epistemic Analysis Engine — 2,075+ AI-powered tools for deep reasoning analysis
• 🤖 Autonomous Research Agent — Multi-step research workflows with tool orchestration
• 🔌 MCP Server — Model Context Protocol integration for external AI agent access
• 📦 Python SDK — Programmatic access to all platform capabilities
• 🌐 Bilingual UI — Full English/Chinese internationalization
• 🔗 Original Links — One-click access to arXiv abstract, PDF, and ar5iv HTML

---

Epistemic Analysis Engine

The core differentiator — a comprehensive taxonomy of 2,075+ detection tools organized across domains:

Domain	Examples	Count
Cognitive Biases	Anchoring, availability heuristic, confirmation bias, Dunning-Kruger	200+
Logical Fallacies	Ad hominem, straw man, false dilemma, slippery slope	150+
Causal Reasoning	Post hoc, reverse causation, spurious correlation, single cause	100+
Epistemic Scale	Ecological fallacy, composition/division, scope neglect	80+
Social Dynamics	Groupthink, pluralistic ignorance, reputation cascade, conformity	100+
Institutional	Citation cartel, credentialism, regulatory capture, peer review theater	80+
Communication	Sealioning, tone policing, jargon gatekeeping, strategic ambiguity	80+
Narrative	Origin myth, survivorship bias, teleological thinking, hindsight	80+
Decision Making	Premature closure, analysis paralysis, commitment escalation	80+
Methodology	P-hacking, Texas sharpshooter, streetlight effect, reification	80+
Temporal	Presentism, shifting baseline, recency illusion, end-of-history	60+
Emotion	Affect infusion, moral outrage substitution, empathy gap	60+
Power & Identity	Epistemic injustice, manufactured consent, tribal epistemology	100+
Technology	Algorithm opacity, filter bubble, automation bias, digital amnesia	60+
Meta-cognition	Blind spot bias, illusion of explanatory depth, calibration neglect	60+
+ more	Ecology, virtue, attention, measurement, collective…	600+

Each tool accepts domain-specific parameters and returns structured JSON with detection results, severity ratings, and actionable recommendations.

---

Architecture

┌─────────────────────────────────────────────────────────────┐
│                      Frontend (Nuxt 3)                       │
│   Vue 3 · TypeScript · Lucide Icons · GSAP Animations       │
└─────────────────────┬───────────────────────────────────────┘
                      │ REST API
┌─────────────────────▼───────────────────────────────────────┐
│                    Backend (FastAPI)                          │
│   SQLAlchemy · Celery · Pydantic · Structlog                 │
├──────────────────────────────────────────────────────────────┤
│  Agent Layer                                                 │
│  ┌──────────────┐  ┌──────────────┐  ┌──────────────────┐   │
│  │ Tool         │  │ Research     │  │ MCP Server       │   │
│  │ Dispatcher   │  │ Runtime      │  │ (2075+ tools)    │   │
│  │ (2075 tools) │  │ (autonomous) │  │                  │   │
│  └──────────────┘  └──────────────┘  └──────────────────┘   │
├─────────────┬───────────┬───────────┬───────────────────────┤
│ PostgreSQL  │   Redis   │ Meilisearch│    Qdrant            │
│  (primary)  │  (cache)  │ (fulltext) │  (embeddings)        │
├─────────────┴───────────┴───────────┴───────────────────────┤
│  Neo4j (graph)  ·  MinIO (objects)  ·  GROBID (PDF)         │
└─────────────────────────────────────────────────────────────┘

---

Project Structure

Kaleidoscope/
├── backend/
│   ├── app/
│   │   ├── api/v1/             # REST endpoints (20+ routers)
│   │   ├── clients/            # External API clients (arXiv, MinerU, OpenAlex, LLM)
│   │   ├── models/             # SQLAlchemy ORM models
│   │   ├── schemas/            # Pydantic request/response schemas
│   │   ├── services/           # 2075+ epistemic analysis services
│   │   │   ├── agent/          # Tool dispatcher & agent orchestration
│   │   │   ├── extraction/     # QA engine, summarizer
│   │   │   ├── search/         # Vector & hybrid search
│   │   │   ├── graph/          # Citation graph analysis
│   │   │   └── *.py            # Individual detection services
│   │   ├── mcp_server.py       # MCP protocol server (all tools exposed)
│   │   ├── tasks/              # Celery async tasks
│   │   ├── graph_db/           # Neo4j driver & queries
│   │   └── utils/              # Shared utilities
│   ├── kaleidoscope_sdk/       # Python SDK client
│   ├── alembic/                # Database migrations
│   ├── docker/                 # Docker Compose for infrastructure
│   └── pyproject.toml          # Python dependencies & tooling
│
├── frontend/                   # Nuxt 3 frontend
│   ├── components/             # Vue components
│   ├── pages/                  # File-based routing
│   ├── composables/            # Shared logic
│   ├── i18n/                   # en-US / zh-CN translations
│   └── nuxt.config.ts
│
└── docker-compose.yml          # Full-stack orchestration

---

Quick Start

Prerequisites

• Docker & Docker Compose
• Node.js 20+ / pnpm
• Python 3.12+ / uv (or pip)

1. Infrastructure

cd backend/docker
docker compose up -d   # PostgreSQL, Redis, Qdrant, Meilisearch, Neo4j, MinIO

2. Backend

cd backend
cp .env.example .env   # Configure API keys and DB URLs
uv sync                # Install dependencies
alembic upgrade head   # Run migrations
uvicorn app.main:app --reload --port 8000

3. Frontend

cd frontend
pnpm install
pnpm dev               # http://localhost:3000

4. MCP Server (for AI agent integration)

cd backend
python -m app.mcp_server   # Exposes 2075+ tools via MCP protocol

---

SDK Usage

from kaleidoscope_sdk import KaleidoscopeClient

client = KaleidoscopeClient(base_url="http://localhost:8000")

# Detect confirmation bias in a claim
result = await client.call_tool(
    "confirmation_bias_detect",
    claim="Studies consistently show X",
    evidence_pattern="Only favorable studies cited",
    domain="medicine"
)

# Run autonomous research
run = await client.research(
    query="What is the evidence for X?",
    depth="deep"
)

---

API Endpoints

Module	Prefix	Description
Papers	/papers	CRUD, batch import, Markdown conversion
Collections	/collections	Paper organization
Search	/search	Multi-modal search
Agent	/agent	Autonomous research agent
Intelligence	/intelligence	AI-powered insights
OpenAlex	/openalex	External search + citation graph builder
Knowledge	/knowledge	Note graph
Feeds	/feeds	RSS management

Interactive docs available at http://localhost:8000/docs when backend is running.

---

Tech Stack

Layer	Technology
Frontend	Nuxt 3, Vue 3, TypeScript, Lucide Icons, GSAP
Backend	FastAPI, SQLAlchemy 2, Celery, Pydantic v2
Database	PostgreSQL 16, Redis 7
Search	Meilisearch, Qdrant (vector)
Graph	Neo4j 5
Storage	MinIO (S3-compatible)
PDF Parser	GROBID, MinerU API
AI/LLM	Configurable endpoint (OpenAI-compatible)
Protocol	MCP (Model Context Protocol)

---

Contributing

1. Fork the repo
2. Create your feature branch (git checkout -b feat/amazing-feature)
3. Commit with conventional commits (feat:, fix:, docs:)
4. Push and create a Pull Request

---

License

This project is licensed under the Kaleidoscope Non-Commercial License (KNCL) v1.0. Commercial use requires separate written permission or a commercial license. See LICENSE for details.