Add R personal shopper using OpenAI/Ollama + SerpAPI

[charles-hebert]

Motivation

• Provide a CLI-driven personal shopper implemented in R that reads a text list of items to scout and automates query generation, multi-engine shopping searches, and suggestion ranking.
• Support either openai or ollama LLM providers to generate search queries and to rank results so users can choose local or hosted models.
• Normalize and aggregate shopping results from search APIs to produce a compact CSV/table report (price, store, link) for easy consumption.

Description

• Added personal_shopper.R, a new script with a CLI entrypoint and core functions: read_interest_items, generate_queries, search_serpapi, rank_suggestions, shop_item, and run_personal_shopper.
• Implemented provider adapters openai_chat_json and ollama_chat_json and a dispatcher llm_chat_json so the same workflow can use either OpenAI or Ollama models.
• Implemented SerpAPI-based scouting via search_serpapi that normalizes title, price, store, and link, deduplicates results, and extracts numeric prices with safe_num.
• Added interests_sample.txt as an example input and a CSV output path with write_csv for consolidated results; default search engines are google_shopping, bing_shopping, and ebay.

Testing

• Attempted a syntax/parse check using Rscript -e "parse(file='personal_shopper.R'); cat('OK\n')", but it failed because Rscript is not available in the environment, so no runtime validation was completed.
• No end-to-end runs against external services were performed because runtime dependencies and API keys (SERPAPI_KEY, OPENAI_API_KEY or reachable OLLAMA_HOST) were not present or the runtime was unavailable.

---

Codex Task