upload: quickstart.ipynb (RUN_005 HF1)

quickstart.ipynb

@@ -0,0 +1,233 @@
+{
+ "nbformat": 4,
+ "nbformat_minor": 5,
+ "metadata": {
+  "kernelspec": {
+   "display_name": "Python 3",
+   "language": "python",
+   "name": "python3"
+  },
+  "language_info": {
+   "name": "python",
+   "version": "3.12.0"
+  }
+ },
+ "cells": [
+  {
+   "cell_type": "markdown",
+   "id": "cell-01",
+   "metadata": {},
+   "source": [
+    "# USDA Phytochemical & Ethnobotanical Database — Enriched v2.0\n",
+    "\n",
+    "**104,388 records · 24,771 compounds · 2,315 species · 4 enrichment layers**\n",
+    "\n",
+    "| Tier | Price | Includes |\n",
+    "|------|-------|----------|\n",
+    "| **Single Entity** | €699 | JSON + Parquet + SHA-256 Manifest |\n",
+    "| **Team** | €1,349 | + `duckdb_queries.sql` (20 queries) + `compound_priority_score.py` + 4 Pre-computed Views |\n",
+    "| **Enterprise** | €1,699 | + `snowflake_load.sql` + `chromadb_ingest.py` + `pinecone_ingest.py` + `embedding_guide.md` + Opportunity Matrix |\n",
+    "\n",
+    "**Full dataset:** [ethno-api.com](https://ethno-api.com) · ",
+    "**This notebook runs on the free 400-row sample.**\n"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "id": "cell-02",
+   "metadata": {},
+   "outputs": [],
+   "execution_count": null,
+   "source": [
+    "import pandas as pd\n",
+    "import duckdb\n",
+    "\n",
+    "# Load the free 400-row sample\n",
+    "# To use full dataset: replace path with 'ethno_dataset_2026_v2.json'\n",
+    "SAMPLE = 'ethno_sample_400.json'\n",
+    "\n",
+    "df = pd.read_json(SAMPLE)\n",
+    "print(f'Shape: {df.shape}')\n",
+    "print(f'Columns: {list(df.columns)}')\n",
+    "df.head(3)"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "id": "cell-03",
+   "metadata": {},
+   "outputs": [],
+   "execution_count": null,
+   "source": [
+    "# Dataset statistics\n",
+    "print(f'Unique compounds:  {df[\"chemical\"].nunique():,}')\n",
+    "print(f'Unique species:    {df[\"plant_species\"].nunique():,}')\n",
+    "print(f'Application coverage: {df[\"application\"].notna().mean()*100:.1f}%')\n",
+    "print(f'Dosage coverage:      {df[\"dosage\"].notna().mean()*100:.1f}%')\n",
+    "print(f'\\nTop compound by PubMed evidence:')\n",
+    "print(df.groupby(\"chemical\")[\"pubmed_mentions_2026\"].max().sort_values(ascending=False).head(5).to_string())"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "id": "cell-04",
+   "metadata": {},
+   "outputs": [],
+   "execution_count": null,
+   "source": [
+    "# Q05: Composite evidence score (weighted: PubMed 30% | Trials 35% | ChEMBL 20% | Patents 15%)\n",
+    "# Source: duckdb_queries.sql — available in Team + Enterprise tier\n",
+    "result = duckdb.sql(f\"\"\"\n",
+    "    SELECT\n",
+    "        chemical,\n",
+    "        MAX(pubmed_mentions_2026)       AS pubmed,\n",
+    "        MAX(clinical_trials_count_2026) AS trials,\n",
+    "        MAX(chembl_bioactivity_count)   AS bioassays,\n",
+    "        MAX(patent_count_since_2020)    AS patents,\n",
+    "        ROUND(\n",
+    "            (MAX(pubmed_mentions_2026)       * 0.30) +\n",
+    "            (MAX(clinical_trials_count_2026) * 0.35) +\n",
+    "            (MAX(chembl_bioactivity_count)   * 0.20) +\n",
+    "            (MAX(patent_count_since_2020)    * 0.15),\n",
+    "        2) AS composite_score\n",
+    "    FROM read_json_auto('{SAMPLE}')\n",
+    "    GROUP BY chemical\n",
+    "    ORDER BY composite_score DESC\n",
+    "    LIMIT 10\n",
+    "\"\"\").df()\n",
+    "print('Top 10 by composite evidence score:')\n",
+    "result"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "id": "cell-05",
+   "metadata": {},
+   "outputs": [],
+   "execution_count": null,
+   "source": [
+    "# Q17: IP whitespace candidates — high research signal, low patent activity\n",
+    "# Use case: freedom-to-operate screening\n",
+    "# Source: duckdb_queries.sql (Team + Enterprise tier)\n",
+    "whitespace = duckdb.sql(f\"\"\"\n",
+    "    SELECT\n",
+    "        chemical,\n",
+    "        MAX(pubmed_mentions_2026)    AS pubmed,\n",
+    "        MAX(patent_count_since_2020) AS patents_since_2020,\n",
+    "        ROUND(MAX(pubmed_mentions_2026)::DOUBLE / NULLIF(MAX(patent_count_since_2020),0), 1)\n",
+    "            AS research_to_patent_ratio\n",
+    "    FROM read_json_auto('{SAMPLE}')\n",
+    "    GROUP BY chemical\n",
+    "    HAVING MAX(pubmed_mentions_2026) > 50 AND MAX(patent_count_since_2020) < 5\n",
+    "    ORDER BY research_to_patent_ratio DESC\n",
+    "    LIMIT 10\n",
+    "\"\"\").df()\n",
+    "print('IP Whitespace Candidates (high research, low patents):')\n",
+    "whitespace"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "id": "cell-06",
+   "metadata": {},
+   "outputs": [],
+   "execution_count": null,
+   "source": [
+    "# Q06: Evidence tier classification (PLATINUM / GOLD / SILVER / BRONZE)\n",
+    "# Source: duckdb_queries.sql (Team + Enterprise tier)\n",
+    "tiers = duckdb.sql(f\"\"\"\n",
+    "    SELECT evidence_tier, COUNT(*) AS compound_count\n",
+    "    FROM (\n",
+    "        SELECT chemical,\n",
+    "            CASE\n",
+    "                WHEN MAX(pubmed_mentions_2026)>5000 AND MAX(clinical_trials_count_2026)>100 THEN 'PLATINUM'\n",
+    "                WHEN MAX(pubmed_mentions_2026)>1000 AND MAX(clinical_trials_count_2026)>20  THEN 'GOLD'\n",
+    "                WHEN MAX(pubmed_mentions_2026)>100 THEN 'SILVER'\n",
+    "                ELSE 'BRONZE'\n",
+    "            END AS evidence_tier\n",
+    "        FROM read_json_auto('{SAMPLE}')\n",
+    "        GROUP BY chemical\n",
+    "    ) sub\n",
+    "    GROUP BY evidence_tier\n",
+    "    ORDER BY evidence_tier\n",
+    "\"\"\").df()\n",
+    "print('Evidence tier distribution:')\n",
+    "tiers"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "id": "cell-07",
+   "metadata": {},
+   "source": [
+    "## What's Included in Each Tier\n",
+    "\n",
+    "### Team License (€1,349) — adds to Single:\n",
+    "| File | Description |\n",
+    "|------|-------------|\n",
+    "| `duckdb_queries.sql` | 20 production-ready queries across 5 categories (compound discovery, evidence scoring, species analysis, application intelligence, IP/pipeline) |\n",
+    "| `compound_priority_score.py` | Combined evidence score calculator across all 4 layers |\n",
+    "| `top500_by_pubmed.parquet` | Pre-computed Top 500 compounds by PubMed score |\n",
+    "| `top500_by_trials.parquet` | Pre-computed Top 500 by ClinicalTrials count |\n",
+    "| `top500_by_patent_density.parquet` | Pre-computed Top 500 by patent density |\n",
+    "| `anti_inflammatory_panel.parquet` | All anti-inflammatory application records |\n",
+    "\n",
+    "### Enterprise License (€1,699) — adds to Team:\n",
+    "| File | Description |\n",
+    "|------|-------------|\n",
+    "| `snowflake_load.sql` | Drop-in Snowflake import script (Stage + COPY INTO + Verify) |\n",
+    "| `chromadb_ingest.py` | ChromaDB vector ingest with batch upload, --resume, verification |\n",
+    "| `pinecone_ingest.py` | Pinecone ingest supporting OpenAI, sentence-transformers, Pinecone inference |\n",
+    "| `embedding_guide.md` | ClinicalBERT, RAG pipeline templates, cost benchmarks |\n",
+    "| `compound_opportunity_matrix.csv` | Ranked compound candidates: high bioactivity × low patent density |\n",
+    "| `clinical_pipeline_gaps.csv` | High-PubMed, low-trial compounds: drug discovery pipeline gaps |\n",
+    "| `ethno_rag_chunks.jsonl` | Pre-chunked JSONL for direct LLM embedding (ChromaDB / Pinecone) |\n",
+    "\n",
+    "→ **Purchase full dataset:** [ethno-api.com](https://ethno-api.com)\n"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "id": "cell-08",
+   "metadata": {},
+   "outputs": [],
+   "execution_count": null,
+   "source": [
+    "# Load via HuggingFace Datasets\n",
+    "# from datasets import load_dataset\n",
+    "# ds = load_dataset(\n",
+    "#     'wirthal1990-tech/USDA-Phytochemical-Database-JSON',\n",
+    "#     split='sample',\n",
+    "#     trust_remote_code=False\n",
+    "# )\n",
+    "# df_hf = ds.to_pandas()\n",
+    "# print(f'Records: {len(df_hf)} | Columns: {list(df_hf.columns)}')\n",
+    "print('HuggingFace repo: https://huggingface.co/datasets/wirthal1990-tech/USDA-Phytochemical-Database-JSON')"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "id": "cell-09",
+   "metadata": {},
+   "source": [
+    "## Citation\n",
+    "\n",
+    "```bibtex\n",
+    "@misc{ethno_api_v2_2026,\n",
+    "  title     = {USDA Phytochemical \\& Ethnobotanical Database --- Enriched v2.0},\n",
+    "  author    = {Wirth, Alexander},\n",
+    "  year      = {2026},\n",
+    "  publisher = {Ethno-API},\n",
+    "  url       = {https://ethno-api.com},\n",
+    "  note      = {104,388 records, 24,771 unique chemicals, 2,315 plant species,\n",
+    "               8-column schema with PubMed, ClinicalTrials, ChEMBL, and PatentsView enrichment}\n",
+    "}\n",
+    "```\n",
+    "\n",
+    "**Links:** [ethno-api.com](https://ethno-api.com) · ",
+    "[GitHub](https://github.com/wirthal1990-tech/USDA-Phytochemical-Database-JSON) · ",
+    "[HuggingFace](https://huggingface.co/datasets/wirthal1990-tech/USDA-Phytochemical-Database-JSON)\n"
+   ]
+  }
+ ]
+}