Salesforce Enterprise Deep Research

2 hours ago 1

We present Enterprise Deep Research (EDR), a multi-agent system that integrate:

Master Planning Agent for adaptive query decomposition.
Four specialized search agents (General, Academic, GitHub, LinkedIn).
Extensible MCP-based tool ecosystem supporting NL2SQL, file analysis, and enterprise workflows.
Visualization Agent for data-driven insights.
Reflection mechanism that detects knowledge gaps and updates research direction with optional human-in-the-loop steering guidance.
Real-time steering commands for continuous research refinement.

Note

These components enable automated report generation, real-time streaming, and seamless enterprise deployment, as validated on internal datasets.

We present two video demos of using EDR in web application and in Slack workspace.

Note

Multi-provider LLM support • Slack agent • Real-time streaming • Document analysis • Citation management • Parallel processing • Specialized benchmarking • Human-in-the-loop steering

edr_demo.mp4

edr_slack.mp4

Requirements: Python 3.11+ • Node.js 20.9.0+

# Clone and setup git clone https://github.com/SalesforceAIResearch/enterprise-deep-research.git cd enterprise-deep-research # Python environment python -m venv venv source venv/bin/activate # Windows: venv\Scripts\activate pip install -r requirements.txt # Configure environment cp .env.sample .env # Edit .env with your API keys # Frontend setup cd ai-research-assistant && npm install && npm run build && cd ..

Environment Configuration

Required Variables:

TAVILY_API_KEY - Tavily search API key
One LLM provider key:
- OPENAI_API_KEY - OpenAI API key
- ANTHROPIC_API_KEY - Anthropic API key
- GROQ_API_KEY - Groq API key
- GOOGLE_CLOUD_PROJECT - Google Cloud project ID
- SAMBNOVA_API_KEY - SambaNova API key

Optional Settings:

LLM_PROVIDER - Default provider (default: openai)
LLM_MODEL - Model name (provider-specific defaults)
MAX_WEB_RESEARCH_LOOPS - Max iterations (default: 10)

Provider Default Model Available Models

OpenAI	o4-mini	o4-mini, o4-mini-high, o3-mini, o3-mini-reasoning, gpt-4o
Anthropic	claude-sonnet-4	claude-sonnet-4, claude-sonnet-4-thinking, claude-3-7-sonnet, claude-3-7-sonnet-thinking
Google	gemini-2.5-pro	gemini-2.5-pro, gemini-1.5-pro-latest, gemini-1.5-flash-latest
Groq	deepseek-r1-distill-llama-70b	deepseek-r1-distill-llama-70b, llama-3.3-70b-versatile, llama3-70b-8192
SambaNova	DeepSeek-V3-0324	DeepSeek-V3-0324

Full Stack (Recommended) - Single Command:

python -m uvicorn app:app --host 0.0.0.0 --port 8000

The application will serve both the backend API and pre-built frontend at http://localhost:8000

Backend API Documentation: http://localhost:8000/docs

python benchmarks/run_research.py "Your research question" \ --provider openai --model o3-mini --max-loops 3

Navigate to http://localhost:8000 for interactive research with real-time progress tracking.

📚 Benchmarking & Development

DeepResearchBench: Comprehensive research evaluation
ResearchQA: Question-answering with citation verification
DeepConsult: Consulting-style analysis tasks

The EDR-200 dataset contains 201 complete agentic research trajectories generated by Enterprise Deep Research—99 queries from DeepResearch Bench and 102 queries from DeepConsult. Unlike prior benchmarks that only capture final outputs, these trajectories expose the full reasoning process across search, reflection, and synthesis steps, enabling fine-grained analysis of agentic planning and decision-making dynamics.

Refer to our detailed benchmarking guide.

# Testing python -m pytest tests/ python test_agents.py # Code quality black src/ services/ benchmarks/ mypy src/ services/ flake8 src/ services/ benchmarks/ # Development server python -m uvicorn app:app --reload --host 0.0.0.0 --port 8000 cd ai-research-assistant && npm run dev

enterprise-deep-research/ ├── ai-research-assistant/ # React frontend ├── benchmarks/ # Evaluation framework ├── src/ # Core research engine │ ├── agent_architecture.py # Multi-agent orchestration │ ├── graph.py # LangGraph workflow definitions │ ├── state.py # Research state management │ ├── simple_steering.py # Steering & task management │ ├── steering_integration.py # Steering integration layer │ ├── prompts.py # Agent prompts & templates │ ├── configuration.py # Agent configuration │ ├── utils.py # Utility functions │ ├── visualization_agent.py # Visualization generation │ └── tools/ # Research tools & MCP integration ├── services/ # Backend services (research, analysis, parsing) ├── routers/ # FastAPI endpoints ├── models/ # Data schemas ├── app.py # Main FastAPI application ├── llm_clients.py # LLM provider clients ├── session_store.py # Session management └── requirements.txt # Python dependencies

Licensed under Apache 2.0.

@article{prabhakar2025enterprisedeepresearch, title={Enterprise Deep Research: Steerable Multi-Agent Deep Research for Enterprise Analytics}, author={Prabhakar, Akshara and Ram, Roshan and Chen, Zixiang and Savarese, Silvio and Wang, Frank and Xiong, Caiming and Wang, Huan and Yao, Weiran}, journal={arXiv preprint arXiv:2510.17797}, year={2025} }

Acknowledgments: Built on LangGraph, Tavily, React, Tailwind CSS, and FastAPI.

Read Entire Article