company research agent

Report Abuse

Basic Information

This repository implements an agentic platform that produces comprehensive company research reports by orchestrating a pipeline of specialized AI agents. It is built to gather, curate, synthesize, and format information about a company from multiple sources such as company websites, news articles, financial reports, and industry analyses. The system is implemented as a backend API with WebSocket streaming and a modern React frontend for real-time progress and report viewing. Core components include distinct research analyzers and processing nodes that run sequentially to collect, filter, brief, and edit content. The project integrates third-party APIs and model backends and can be run locally, with Docker, or deployed to common cloud platforms. It requires API keys for Tavily, Gemini, and OpenAI and supports optional MongoDB persistence. The codebase and documentation provide setup scripts, manual install steps, and deployment guidance.

Links

Categorization

App Details

Features
The project centers on a multi-agent research pipeline with named analysis nodes such as CompanyAnalyzer, IndustryAnalyzer, FinancialAnalyst, and NewsScanner plus processing nodes like Collector, Curator, Briefing, and Editor. It performs multi-source research and content curation using Tavily relevance scoring with a configurable threshold. A dual-model architecture uses Gemini 2.0 Flash for high-context synthesis and GPT-4.1-mini for precise formatting, deduplication, and markdown report compilation. Real-time progress and streaming report chunks are delivered over WebSocket endpoints implemented with FastAPI. The frontend is a responsive React UI that subscribes to status types including query generation, document curation, briefing start/complete, and report chunks. Utilities include Docker compose files, setup scripts for dependency installation, and optional MongoDB persistence.
Use Cases
This tool speeds and standardizes company diligence by automating data collection, relevance-based filtering, synthesis, and formatted report generation. Analysts, researchers, or teams can use the pipeline to consolidate disparate public sources into structured briefings and a final report without manual aggregation. Real-time WebSocket streaming provides visibility into progress and intermediate results during long research jobs. The modular node design enables customization or extension of analyzers and processing steps. Local development, Docker, and cloud deployment options make it adaptable to different environments. Built-in content deduplication and relevance scoring reduce noise and help focus the output on high-value documents.

Please fill the required fields*