How milwinantany/agentic-researcher works
This repository implements a command-line research assistant: a user supplies a topic, and the orchestrator plans sub-questions, finds and crawls web pages, indexes their text, synthesizes findings with an LLM, and writes a Markdown brief. The graph distinguishes orchestration from the domain stages and shows the search, web, vector-store, and model integrations. The README describes interchangeable LLM providers, but the inspected LLM utility specifically calls Ollama; other provider wiring is not shown.
Main entry points: main.py CLI orchestrator · reason.py Research synthesizer · llm.py LLM utility · planner.py Question planner
Components 14
Research workflow
- CLI orchestrator
main.py - Question planner
core/planner.py - Research synthesizer
core/reason.py
Language model
- LLM utility
utils/llm.py
Index and reports
- Document indexer
core/indexer.py - ChromaDB collection
- Markdown writer
core/output.py - Research reports
outputs
External sources
- Web search
core/search.py - Page crawler
core/crawler.py
Other components
- Researcher
- DDGS search
- Web pages
- Ollama model
Connections 15
- Researcher → CLI orchestrator: submits topic
README.md - CLI orchestrator → Question planner: plans questions
main.py - Question planner → LLM utility: requests questions
core/planner.py - LLM utility → Ollama model: sends prompt
utils/llm.py - CLI orchestrator → Web search: searches topic
main.py - Web search → DDGS search: queries results
core/search.py - CLI orchestrator → Page crawler: crawls URLs
main.py - Page crawler → Web pages: fetches pages
core/crawler.py - CLI orchestrator → Document indexer: indexes documents
main.py - Document indexer → ChromaDB collection: stores documents
core/indexer.py - CLI orchestrator → Research synthesizer: synthesizes research
main.py - Research synthesizer → ChromaDB collection: queries context
core/reason.py - Research synthesizer → LLM utility: requests synthesis
core/reason.py - CLI orchestrator → Markdown writer: writes brief
main.py - Markdown writer → Research reports: writes Markdown
core/output.py