Docs Lookup Pipeline

Docs Lookup Pipeline

Documentation lookup, caching, and freshness pipeline for AI agent real-time doc retrieval. Multi-source cascade (Context7, DeepWiki, direct URL) with TTL-based caching and stale pattern detection.

Quick Start

pip install -e .
docs-lookup lookup next app-router --version 14
docs-lookup check src/page.tsx next
docs-lookup stats --cache-dir .cache/docs

Architecture

graph TD
    A[Lookup Request] --> B{Cache Check}
    B -->|Hit & Fresh| C[Return Cached]
    B -->|Miss/Stale| D[Context7]
    D -->|Found| E[Cache & Return]
    D -->|Miss| F[DeepWiki]
    F -->|Found| E
    F -->|Miss| G[Direct URL]
    G --> E
    E --> H[DocEntry]
    H --> I[Freshness Score]

Usage

from docs_lookup_pipeline.core import DocLookupPipeline, detect_stale_patterns

pipeline = DocLookupPipeline()
result = pipeline.lookup("next", "app-router", version="14")
if result.found:
    print(result.best_entry.content)

# Detect stale API usage
findings = detect_stale_patterns(code, "next")
for f in findings:
    print(f"Line {f['line']}: {f['recommendation']}")

Features

• Multi-source cascade with configurable priority
• TTL-based LRU cache with disk persistence
• Freshness levels: FRESH / STALE / EXPIRED
• Stale API pattern detection for Next.js, React
• Batch lookup for multiple libraries

Part of Agentic Development Series

Post 61 of "Agentic Development: 21 Lessons from 8,481 AI Coding Sessions" by Nick Krzemienski.