Skip to content

features.md

Latest commit

 

History

History
181 lines (131 loc) · 7.38 KB

features.md

File metadata and controls

181 lines (131 loc) · 7.38 KB

Core Features

DiscoveryOS provides a rich set of features for AI-powered research and file management.

Workspace Management

Workspaces map server-side folders into the application. Each workspace is a persistent container for files, chat history, and notes.

  • Create workspaces from any allowed root directory (configured via WORKSPACE_ROOTS)
  • Open existing workspaces from the home page
  • Delete workspaces (removes only the database record; files on disk are preserved)
  • GitHub Clone — Clone a GitHub repository directly as a new workspace

File Browser

The file browser provides full file management within a workspace:

  • Tree View — Hierarchical directory browsing with expand/collapse
  • Upload — Drag-and-drop or click to upload files
  • Create — New files and folders
  • Edit — Open files in the built-in editor
  • Rename / Delete — Right-click context menu
  • Preview — Preview images, PDFs, 3D models (STL, OBJ, GLTF, etc.), and molecular structures (MOL)
  • Sync — Trigger RAG indexing to enable AI-powered search

RAG Chat

The AI chat interface uses Retrieval-Augmented Generation to answer questions based on your workspace files:

  1. Sync your workspace to index files into the vector store
  2. Ask questions in the chat panel
  3. Receive answers with [Source: filename] citations pointing to the relevant files
  4. Multi-turn conversations — Context is maintained across messages

Supported File Types for RAG

The following file types are supported for text extraction and indexing:

Category Extensions
Documents .pdf, .txt, .md, .html
Data .json, .csv
Code .js, .ts, .py, .java, .go, .rs, .cpp, .c, .rb, .php, and more

Note Generation

Automatically generate structured notes from your workspace content:

Type Description
Summary Concise overview of key points
FAQ Frequently asked questions with answers
Brief Executive-style briefing document
Timeline Chronological sequence of events
Memory Agent conversation memory summaries
Daily Report Auto-generated daily workspace activity report
Weekly Report Auto-generated weekly workspace activity report

Notes can also be created and edited manually.

Agent Mode

The Agent mode provides an autonomous AI assistant with access to system tools:

Tool Description
bash Execute shell commands (configurable timeout up to 300s)
readFile Read file contents within workspace
writeFile Create or overwrite files
listDirectory List directory contents with metadata
grep Search file contents with regular expressions
searchArticles Search arXiv and Hugging Face Daily Papers
getSkillInstructions Load SCP scientific skill workflows
kubectl Kubernetes cluster management (read-only by default)
submitK8sJob Submit Volcano K8s jobs to Ascend 910B NPU clusters
collectJobResults Collect K8s job logs and status

Agent mode supports three sub-modes:

Mode Tools Available Use Case
Agent All tools Full autonomous execution
Plan readFile, listDirectory, grep Read-only analysis and planning
Ask readFile, listDirectory, grep Simple question answering

Skills

Skills are reusable AI workflow templates:

  • 206 built-in SCP (Science Context Protocol) skills from the Intern-Discovery Platform, covering 8 research domains:
    • Drug Discovery & Pharmacology (71 skills)
    • Genomics & Genetic Analysis (41 skills)
    • Protein Science & Engineering (38 skills)
    • Chemistry & Molecular Science (24 skills)
    • Physics & Engineering Computation (18 skills)
    • Experimental Automation & Literature Mining (7 skills)
    • Earth & Environmental Science (5 skills)
    • Cross-domain utility skills (2 skills)
  • Create custom skills with specific system prompts and instructions
  • Configure allowed tools per skill to control access
  • Define parameters with {{paramName}} template injection
  • Define multi-step workflows with tool hints per step
  • Import/export skills as JSON for sharing

Skills can be triggered from the Agent panel or auto-matched based on user intent.

Article / Paper Search

Search and study academic papers directly from the application:

  • Search Sources — arXiv and Hugging Face Daily Papers
  • Filter by keywords, date range, and source
  • Related Articles — Find papers related to a given article
  • Paper Study UI — Dedicated interface for searching, fetching, summarizing, and chatting about papers
  • Agent Integration — The searchArticles tool is available in Agent mode

Scheduled Tasks

Automate recurring operations with cron-based scheduled tasks:

Task Type Description
daily_report Generate daily workspace activity report
weekly_report Generate weekly workspace activity report
git_sync Automatically pull Git repositories
source_sync Re-index workspace files
custom User-defined workflows

Tasks can be global or workspace-specific, with execution tracking (status, errors, last run time).

Dataset Management

Download and manage datasets from HuggingFace Hub and ModelScope:

  • Download datasets, models, or spaces from HuggingFace or ModelScope
  • Import existing local directories as datasets
  • Lifecycle — Pause, resume, cancel, and retry downloads
  • Progress Tracking — Real-time download progress monitoring (0–100%)
  • File Preview — Preview dataset contents with split selection
  • Workspace Linking — Link datasets to workspaces (many-to-many)
  • File Filtering — Specify allow/ignore patterns for selective downloads

Cluster / Kubernetes Integration

Monitor and manage Kubernetes clusters for GPU-accelerated workloads:

  • Cluster Status — View nodes, jobs, and pods at a glance
  • kubectl / vcctl — Execute cluster commands via Agent mode
  • Volcano Job Submission — Submit jobs to Ascend 910B NPU clusters
  • Result Collection — Collect job logs, status, and exit codes
  • Audit Trail — Full operation history with workspace-level filtering

Requires KUBECONFIG_PATH and related K8S_* environment variables to be configured.

Multi-LLM Support

Switch between different AI providers and models:

Provider Models
OpenAI GPT-5.2, GPT-4.1, GPT-4.1 Mini, GPT-4.1 Nano, GPT-4o, GPT-4o Mini, o3, o3 Mini, o4 Mini
Anthropic Claude Opus 4.6, Claude Sonnet 4, Claude 3.7 Sonnet, Claude 3.5 Haiku
Gemini Gemini 2.5 Flash, Gemini 2.5 Pro, Gemini 3 Flash, Gemini 3 Pro, Gemini 3.1 Pro (Thinking)

Custom endpoints are supported via OPENAI_BASE_URL, ANTHROPIC_BASE_URL, and GEMINI_BASE_URL for third-party providers or self-hosted models.

Bilingual UI

The user interface supports both Chinese and English:

  • Click the language toggle in the top navigation bar
  • All UI elements, labels, and messages are translated
  • Language preference is persisted across sessions

Dark Mode

Toggle between light and dark themes:

  • Click the theme toggle in the top navigation bar
  • Applies to all panels including the file browser, chat, and editor
  • Theme preference is persisted across sessions