Sharad Khare Learn
Domain

DeepForge AI

Most quick AI answers are shallow. DeepForge is designed for topics that need depth: - multi-source web exploration - synthesis across findings - structured report generation - enhancement with examples, implications, and context You provide a topic, DeepForge runs a two-stage agent workflow, and you receive a downloadable markdown report.

Python Streamlit
← All projects

DeepForge AI

Deep web research assistant by Sharad Khare.

DeepForge AI helps you turn any research question into a comprehensive, citation-ready report using OpenAI Agents SDK and Firecrawl deep research.

Built and maintained by Sharad Khare — AI strategist, full-stack developer, and creator of practical AI workflow tools.

What this project does

Most quick AI answers are shallow. DeepForge is designed for topics that need depth:

  • multi-source web exploration
  • synthesis across findings
  • structured report generation
  • enhancement with examples, implications, and context

You provide a topic, DeepForge runs a two-stage agent workflow, and you receive a downloadable markdown report.

Why Sharad Khare built DeepForge AI

Research-heavy work (market analysis, technical scouting, policy review, competitive intelligence) needs more than one-pass chat responses.

DeepForge demonstrates a production-style pattern:

1. Research agent gathers evidence with Firecrawl deep research

2. Elaboration agent expands the draft into an actionable long-form report

This gives teams a reusable blueprint for deep-research copilots.

How it works

User Topic
   │
   ▼
Research Agent + Firecrawl Deep Research
   │
   ▼
Initial Structured Report
   │
   ▼
Elaboration Agent (context, examples, implications)
   │
   ▼
Enhanced Markdown Report + Download

Pipeline stages

| Stage | Component | Output |

|------|-----------|--------|

| 1. Input | Streamlit UI | Topic + API keys |

| 2. Deep research | Firecrawl + research agent | Multi-source findings |

| 3. Draft synthesis | Research agent | Initial report |

| 4. Enhancement | Elaboration agent | Expanded report |

| 5. Delivery | Streamlit UI | View + markdown download |

Features

  • Deep web research with Firecrawl (max_depth, time_limit, max_urls)
  • Two-agent workflow: research + elaboration
  • Real-time research activity updates in UI
  • Initial and enhanced report views
  • One-click markdown export
  • Modular code structure (deepforge/services.py)

Quick start

cd deepforge-ai
pip install -r requirements.txt
streamlit run app.py

Required API keys

  • OpenAI API key
  • Firecrawl API key

Enter both keys in the sidebar, provide a topic, and click Start Research.

Example research topics

  • "Latest developments in agentic AI for enterprise workflows"
  • "Impact of climate policy on renewable infrastructure investment"
  • "State of open-source LLM tooling for production deployments"
  • "Security risks in AI browser automation systems"
  • "Emerging trends in multimodal model adoption"

Project structure

deepforge-ai/
├── app.py
├── deepforge/
│   ├── config.py
│   └── services.py
├── requirements.txt
├── pyproject.toml
└── README.md

Who this is for

  • Analysts producing deep topic briefs
  • Founders doing market and competitor research
  • Consultants building research copilots
  • Developers learning OpenAI + Firecrawl agent patterns

Use cases

  • Generate long-form research reports from one prompt
  • Build internal knowledge briefs for strategy teams
  • Automate technical landscape scans for new domains
  • Create downloadable research artifacts for stakeholders

Author

License

MIT © Sharad Khare