Ecommerce competitor price, margin and stock monitor

Autonomous price monitoring pack for small ecommerce operators who need a repeatable way to detect price drops, competitor pressure, margin risk, and repricing actions before revenue is lost.

This pack is offline-first. It can run from a local JSON export, cron job, ERP dump, Shopify export, marketplace scrape, or any connector you add later. If a local coding-agent CLI is available, it can add a short reasoning summary, but the core scan is deterministic and works without any LLM key.

Who It Is For

• Store operators tracking 10 to 5,000 SKUs across marketplaces.
• Agencies managing repricing and margin alerts for several clients.
• LLM operators who need a safe, auditable pricing assistant.
• Developers who want a clean starter for plugging in Shopify, Amazon, Stripe, Airtable, or crawler data.

Quick Start

cd agent-ecommerce-price-monitor-v1
python3 agent.py --help
python3 agent.py scan --input sample_products.json --output report.json
python3 agent.py status

Expected output: report.json with product-level alerts, recommended actions, and a summary by severity.

What It Does

• Reads a product feed from JSON.
• Computes price-drop percentage, competitor delta, gross margin, and inventory pressure.
• Flags below-MAP pricing, negative margin, stale data, stockout risk, and undercut opportunities.
• Produces deterministic actions such as raise_price, match_competitor, hold_price, review_cost, and restock_watch.
• Triage output (canonical contract) uses the buckets critical_alerts, high_alerts, manual_review, and watchlist, with mode set to deterministic and per-product keys sku, severity, action, reason.
• Optionally asks a BYO-LLM CLI to produce the triage object with the prompt stored in PROMPTS.md; the LLM output must match the same five-root contract (mode, critical_alerts, high_alerts, manual_review, watchlist) or it is discarded.

Architecture

 product export / connector
          |
          v
   sample_products.json
          |
          v
   agent.py scan
          |
          +--> deterministic pricing rules
          |
          +--> optional local LLM summary
          |
          v
      report.json
          |
          v
   operator action queue

Honest Limits

• It does not scrape websites by default.
• It does not change prices automatically unless you add a connector.
• It is not financial advice and does not guarantee higher revenue.
• It is built to surface reviewable pricing signals, not to run a black-box repricing engine.

Files

• agent.py - CLI scanner and deterministic pricing engine.
• sample_products.json - offline feed for tests and demos.
• PROMPTS.md - named prompts loaded by agent.py.
• CHECKLIST.md - deployment checklist with commands.
• CUSTOMIZE.md - where to plug your own data and actions.

Use a local model (Ollama)

The optional LLM step can also run on your own machine through Ollama, so no text leaves your server. No account and no API key.

ollama pull <model>               # any chat model you like
export PACK_LLM=ollama            # prefer Ollama even if a coding-agent CLI is installed
export PACK_OLLAMA_MODEL=<model>  # optional: default is the first chat model the server lists
python3 -c "import llm_adapter; print(llm_adapter.detect())"   # prints: ollama

• PACK_OLLAMA_HOST points to another server (default http://127.0.0.1:11434); OLLAMA_HOST is also read when PACK_LLM=ollama.
• PACK_OLLAMA_NUM_CTX sets the context size if your reports or prompts are long.
• Without PACK_LLM, a local Ollama server is used only when no Claude Code, Kimi Code or Codex CLI is found.
• If the server is down or the model is missing, the pack keeps working in its deterministic mode.

Learning memory (local and private)

Every LLM call this pack makes is journaled in ./.pack_memory/ next to the code, so the pack can improve for you. Nothing is ever sent anywhere.

• What is kept: the task name, the masked input, the masked answer, status and latency. Keys, tokens, e-mails, phone numbers, IP addresses, card and IBAN numbers and personal paths are masked before writing; the system prompt is never stored.
• Teach it: python3 pack_memory.py feedback last ok confirms a result, python3 pack_memory.py feedback last bad --fix "the right answer" corrects it. Confirmed and corrected runs are reused as validated examples when a similar input comes in. Unrated runs are never injected into a prompt.
• Find similar cases: python3 pack_memory.py similar "some text" and python3 pack_memory.py clusters (groups of near-identical failures) work in plain Python, no model needed. Optional meaning-based matching uses an embedding model you already run locally: python3 pack_memory.py enable-semantic (Ollama on this machine only).
• Share, only if you want: python3 pack_memory.py export writes an anonymised summary (counts, statuses, size buckets, no text) for you to review.
• Control: PACK_MEMORY=0 turns it off, PACK_MEMORY_STORE_TEXT=0 keeps fingerprints only, python3 pack_memory.py purge --yes wipes it.