iil-researchfw

Platform research framework — async search, citations, analysis, export.

Features

• Async-first: asyncio.gather() für parallele API-Calls (arXiv, Semantic Scholar, PubMed, OpenAlex)
• Brave Search: Web-Search via Brave API
• Citations: APA, MLA, Chicago, Harvard, IEEE, Vancouver + BibTeX/RIS
• AI Summary: LLM-agnostisch via LLMCallable Protocol
• Export: Markdown, LaTeX, DOCX
• Rate Limiting: Token-Bucket per API-Endpoint, per Quelle konfigurierbar (AcademicSearchService(rate_limits={...}), Defaults arXiv 3.0s / S2 1.0s / OpenAlex 0.1s / PubMed 0.34s — parallel über Quellen, seriell innerhalb einer Quelle, auch beim Retry)
• Retry: tenacity Exponential Backoff
• Prompts injizierbar: SmartSearchService/AISummaryService nehmen eigene SmartSearchPrompts/SummaryPrompts an (ADR-204-Brücke für Konsumenten mit eigener Prompt-Verwaltung)

Installation

pip install iil-researchfw
pip install iil-researchfw[export]     # + python-docx, markdown
pip install iil-researchfw[scraping]   # + beautifulsoup4, playwright
pip install iil-researchfw[all]        # alles

Quick Start

import asyncio
from iil_researchfw.search.academic import AcademicSearchService
from iil_researchfw.search.brave import BraveSearchService
from iil_researchfw.citations.formatter import CitationService, CitationStyle

async def main():
    # Academic Search
    academic = AcademicSearchService()
    papers = await academic.search("machine learning transformers", max_results=10)
    for p in papers:
        print(f"{p.title} — {p.source}")

    # Web Search
    brave = BraveSearchService(api_key="your-key")  # or BRAVE_API_KEY env var
    results = await brave.search("Python asyncio best practices")

    # Citations
    citations = CitationService()
    citation = await citations.from_doi("10.1145/3290605.3300233")
    print(citation.format(CitationStyle.APA))

asyncio.run(main())

Architecture

See ADR-105.

License

MIT