PatSnap/Pharma-Intelligence
1
1#!/usr/bin/env python32"""3PatSnap Pharma Intelligence — Hugging Face Space Demo4Product-grade multi-module AI agent for life science intelligence.5 6Modules:7 - Agent Chat (natural language → tool orchestration → report)8 - Target Intelligence (靶点全景)9 - Drug Exploration (药物管线)10 - Disease Investigation (疾病格局)11 - Company Profiling (公司分析)12 - Clinical Trials (临床试验)13"""14 15import os, json, asyncio, time, re16from collections import Counter17from datetime import datetime18from typing import Optional, Dict, List, Tuple19 20import gradio as gr21 22# =============================================================================23# CONFIGURATION24# =============================================================================25 26API_KEY = os.getenv("PATSNAP_API_KEY", "")27HF_TOKEN = os.getenv("HF_TOKEN", "") # optional: enable LLM agent mode28SERVER_URL = f"https://connect.patsnap.com/096456/logic-mcp?apikey={API_KEY}"29 30# Module definitions31MODULES = {32 "target": {"icon": "🎯", "label": "Target Intelligence", "label_cn": "靶点全景",33 "tool": "ls_drug_search", "entity": "target",34 "desc": "Analyze any biomedical target — biology, drugs, pipeline, trials."},35 "drug": {"icon": "💊", "label": "Drug Exploration", "label_cn": "药物管线",36 "tool": "ls_drug_search", "entity": "drug",37 "desc": "Search drugs by target, disease, mechanism, or company."},38 "disease": {"icon": "🏥", "label": "Disease Investigation", "label_cn": "疾病格局",39 "tool": "ls_drug_search", "entity": "disease",40 "desc": "Understand disease landscape — epidemiology, treatments, pipeline."},41 "company": {"icon": "🏢", "label": "Company Profiling", "label_cn": "公司分析",42 "tool": "ls_organization_pipeline_fetch", "entity": "company",43 "desc": "Profile pharma companies — pipeline, deals, therapeutic focus."},44 "trial": {"icon": "🧪", "label": "Clinical Trials", "label_cn": "临床试验",45 "tool": "ls_clinical_trial_search", "entity": "trial",46 "desc": "Explore clinical trials by target, disease, phase, or sponsor."},47}48 49# Language: 'en' or 'zh'50DEFAULT_LANG = "en"51 52# =============================================================================53# MOCK DATA — Comprehensive, module-specific54# =============================================================================55 56MOCK_TARGETS = {57 "EGFR": {58 "name": "EGFR", "full_name": "Epidermal Growth Factor Receptor",59 "family": "ErbB/HER receptor tyrosine kinase",60 "class": "Kinase", "indication_count": 12, "drug_count": 28,61 "pdb_ids": ["1M17", "1XKK", "2J5F"],62 "pathways": ["RAS/RAF/MEK/ERK", "PI3K/AKT/mTOR", "JAK/STAT"],63 "mutation_hotspots": ["L858R (exon 21)", "exon 19 deletion", "T790M (exon 20)", "C797S (exon 20)"],64 "approved_drugs": [65 {"name": "Osimertinib", "type": "Small molecule", "gen": "3rd-gen TKI",66 "year": 2015, "indications": ["NSCLC (T790M+)", "NSCLC (1L EGFRm)", "NSCLC (adjuvant)"],67 "company": "AstraZeneca"},68 {"name": "Gefitinib", "type": "Small molecule", "gen": "1st-gen TKI",69 "year": 2003, "indications": ["NSCLC (EGFRm)"], "company": "AstraZeneca"},70 {"name": "Erlotinib", "type": "Small molecule", "gen": "1st-gen TKI",71 "year": 2004, "indications": ["NSCLC", "Pancreatic"], "company": "Roche/Genentech"},72 {"name": "Afatinib", "type": "Small molecule", "gen": "2nd-gen TKI",73 "year": 2013, "indications": ["NSCLC (EGFRm)"], "company": "Boehringer Ingelheim"},74 {"name": "Dacomitinib", "type": "Small molecule", "gen": "2nd-gen TKI",75 "year": 2018, "indications": ["NSCLC (EGFRm)"], "company": "Pfizer"},76 {"name": "Cetuximab", "type": "Monoclonal antibody", "gen": "mAb",77 "year": 2004, "indications": ["CRC", "HNSCC"], "company": "Merck KGaA / BMS"},78 {"name": "Panitumumab", "type": "Monoclonal antibody", "gen": "mAb",79 "year": 2006, "indications": ["CRC"], "company": "Amgen"},80 {"name": "Amivantamab", "type": "Bispecific antibody", "gen": "BsAb",81 "year": 2021, "indications": ["NSCLC (ex20ins)"], "company": "Janssen"},82 {"name": "Patritumab deruxtecan", "type": "ADC", "gen": "ADC (HER3)",83 "year": 2024, "indications": ["NSCLC (post-TKI)"], "company": "Daiichi Sankyo / Merck"},84 ],85 "pipeline_summary": {86 "phase_3": 45, "phase_2": 120, "phase_1": 85, "preclinical": 200,87 "hot_topics": ["4th-gen TKIs (C797S)", "Bispecific ADCs", "PROTAC degraders",88 "Combination with immunotherapy", "Brain-penetrant TKIs"],89 },90 "competitive_landscape": "Highly competitive — every major pharma has an EGFR asset. "91 "Innovation is focused on resistance mechanisms and next-gen modalities.",92 },93 "HER2": {94 "name": "HER2", "full_name": "Human Epidermal Growth Factor Receptor 2",95 "family": "ErbB/HER receptor tyrosine kinase",96 "class": "Kinase", "indication_count": 5, "drug_count": 15,97 "pdb_ids": ["1N8Z", "3PP0"],98 "pathways": ["RAS/RAF/MEK/ERK", "PI3K/AKT/mTOR"],99 "mutation_hotspots": ["Amplification (breast/gastric)", "Exon 20 mutations (NSCLC)"],100 "approved_drugs": [101 {"name": "Trastuzumab", "type": "Monoclonal antibody", "gen": "mAb",102 "year": 1998, "indications": ["HER2+ Breast Cancer", "HER2+ Gastric Cancer"], "company": "Roche"},103 {"name": "Trastuzumab deruxtecan", "type": "ADC", "gen": "ADC",104 "year": 2019, "indications": ["HER2+ Breast Cancer", "HER2-low Breast Cancer", "HER2+ Gastric Cancer",105 "HER2-mutant NSCLC"], "company": "Daiichi Sankyo / AstraZeneca"},106 {"name": "Pertuzumab", "type": "Monoclonal antibody", "gen": "mAb",107 "year": 2012, "indications": ["HER2+ Breast Cancer"], "company": "Roche"},108 {"name": "Lapatinib", "type": "Small molecule", "gen": "TKI",109 "year": 2007, "indications": ["HER2+ Breast Cancer"], "company": "Novartis"},110 {"name": "Tucatinib", "type": "Small molecule", "gen": "TKI",111 "year": 2020, "indications": ["HER2+ Breast Cancer (CNS mets)"], "company": "Seagen / Merck"},112 ],113 "pipeline_summary": {114 "phase_3": 25, "phase_2": 80, "phase_1": 55, "preclinical": 130,115 "hot_topics": ["HER2-low targeting", "Bispecific ADCs", "Brain metastasis"],116 },117 "competitive_landscape": "HER2 ADC space is the current battleground. Enhertu dominates; "118 "competitors focus on differentiation via payload, DAR, or epitope.",119 },120 "PD-L1": {121 "name": "PD-L1", "full_name": "Programmed Death-Ligand 1",122 "family": "B7 immune checkpoint",123 "class": "Immune checkpoint ligand", "indication_count": 20, "drug_count": 20,124 "approved_drugs": [125 {"name": "Atezolizumab", "type": "Monoclonal antibody", "gen": "Anti-PD-L1 mAb",126 "year": 2016, "indications": ["NSCLC", "SCLC", "Urothelial", "HCC"], "company": "Roche"},127 {"name": "Durvalumab", "type": "Monoclonal antibody", "gen": "Anti-PD-L1 mAb",128 "year": 2017, "indications": ["NSCLC (Stage III)", "SCLC", "Biliary Tract"], "company": "AstraZeneca"},129 {"name": "Avelumab", "type": "Monoclonal antibody", "gen": "Anti-PD-L1 mAb",130 "year": 2017, "indications": ["Merkel Cell", "Urothelial", "RCC"], "company": "Merck KGaA / Pfizer"},131 ],132 "pipeline_summary": {"phase_3": 35, "phase_2": 60, "phase_1": 40, "preclinical": 100},133 "competitive_landscape": "PD-L1 is a companion to PD-1 — the focus is on combination strategies "134 "and predictive biomarker development.",135 },136}137 138MOCK_COMPANIES = {139 "Roche": {140 "name": "Roche", "ticker": "ROG.SW",141 "headquarters": "Basel, Switzerland",142 "employees": "~100,000",143 "2024_revenue": "$65.4B",144 "therapeutic_areas": ["Oncology", "Neuroscience", "Ophthalmology", "Immunology", "Infectious Disease"],145 "flagship_drugs": [146 {"name": "Trastuzumab (Herceptin)", "target": "HER2", "sales": "$3.2B", "phase": "Approved"},147 {"name": "Atezolizumab (Tecentriq)", "target": "PD-L1", "sales": "$4.8B", "phase": "Approved"},148 {"name": "Bevacizumab (Avastin)", "target": "VEGF", "sales": "$2.1B", "phase": "Approved"},149 {"name": "Trastuzumab deruxtecan (co-developed)", "target": "HER2", "sales": "$3.5B", "phase": "Approved"},150 ],151 "pipeline_count": {"approved": 28, "phase_3": 15, "phase_2": 35, "phase_1": 22},152 "recent_deals": [153 "Acquired Carmot Therapeutics (obesity) — $2.7B upfront (2023)",154 "Acquired Telavant (IBD) — $7.1B (2023)",155 ],156 "strategy": "Roche combines strong internal R&D with strategic bolt-on acquisitions. "157 "Oncology remains the core, with growing investment in immunology and metabolic disease.",158 },159 "AstraZeneca": {160 "name": "AstraZeneca", "ticker": "AZN.L",161 "headquarters": "Cambridge, UK",162 "employees": "~90,000",163 "2024_revenue": "$54.1B",164 "therapeutic_areas": ["Oncology", "CVRM", "Respiratory & Immunology", "Rare Disease"],165 "flagship_drugs": [166 {"name": "Osimertinib (Tagrisso)", "target": "EGFR", "sales": "$6.8B", "phase": "Approved"},167 {"name": "Durvalumab (Imfinzi)", "target": "PD-L1", "sales": "$4.5B", "phase": "Approved"},168 {"name": "Trastuzumab deruxtecan (co-developed)", "target": "HER2", "sales": "$3.5B", "phase": "Approved"},169 {"name": "Dapagliflozin (Farxiga)", "target": "SGLT2", "sales": "$7.1B", "phase": "Approved"},170 ],171 "pipeline_count": {"approved": 22, "phase_3": 12, "phase_2": 28, "phase_1": 18},172 "recent_deals": [173 "Acquired Gracell Biotechnologies (CAR-T) — $1.2B (2023)",174 "Acquired Icosavax (RSV/hMPV vaccine) — $1.1B (2023)",175 "Acquired Fusion Pharmaceuticals (radiopharma) — $2.4B (2024)",176 ],177 "strategy": "AZ's oncology portfolio is anchored by Tagrisso, Imfinzi, and Enhertu. "178 "Actively expanding into cell therapy, radiopharmaceuticals, and ADCs.",179 },180}181 182MOCK_DISEASES = {183 "NSCLC": {184 "name": "Non-Small Cell Lung Cancer",185 "global_incidence": "~2.2M new cases/year (2024)",186 "mortality": "~1.8M deaths/year",187 "5yr_survival": "Stage I: 65%, Stage IV: 8%",188 "major_mutations": ["EGFR (15-20%)", "KRAS (25-30%)", "ALK (3-5%)",189 "ROS1 (1-2%)", "BRAF (1-3%)", "MET exon 14 (3%)", "RET (1-2%)"],190 "approved_drugs_count": 45,191 "drug_classes": ["TKIs (1st/2nd/3rd gen)", "Immune checkpoint inhibitors",192 "ADCs", "Bispecific antibodies", "Chemotherapy"],193 "key_drugs": [194 {"name": "Osimertinib", "target": "EGFR", "setting": "1L EGFRm ± adjuvant"},195 {"name": "Pembrolizumab", "target": "PD-1", "setting": "1L PD-L1 ≥50% ± chemo"},196 {"name": "Amivantamab", "target": "EGFR/MET", "setting": "ex20ins"},197 {"name": "Sotorasib", "target": "KRAS G12C", "setting": "2L+"},198 {"name": "Lorlatinib", "target": "ALK", "setting": "1L ALK+"},199 ],200 "market_size": "$32B (2024), projected $48B by 2030",201 "pipeline": {"phase_3": 85, "phase_2": 150, "phase_1": 100},202 "key_trends": ["Perioperative immunotherapy", "MRD-guided adjuvant therapy",203 "Antibody-drug conjugates expanding", "Bispecifics entering 1L"],204 },205 "Breast Cancer": {206 "name": "Breast Cancer",207 "global_incidence": "~2.3M new cases/year (2024)",208 "mortality": "~685K deaths/year",209 "subtypes": ["HR+/HER2- (70%)", "HER2+ (15-20%)", "TNBC (10-15%)"],210 "approved_drugs_count": 52,211 "key_drugs": [212 {"name": "Trastuzumab deruxtecan", "target": "HER2", "setting": "HER2+ and HER2-low"},213 {"name": "Sacituzumab govitecan", "target": "Trop-2", "setting": "TNBC, HR+/HER2-"},214 {"name": "Palbociclib", "target": "CDK4/6", "setting": "HR+/HER2- 1L"},215 {"name": "Olaparib", "target": "PARP", "setting": "BRCA1/2-mutated"},216 ],217 "market_size": "$28B (2024)",218 "key_trends": ["CDK4/6 moving to adjuvant", "ADCs dominating HER2 space",219 "Immunotherapy for TNBC", "Oral SERDs entering market"],220 },221}222 223MOCK_DRUG_SEARCH = {224 # Returned by any drug search; keyed by target/disease225 "default": [226 {"name": "Osimertinib", "target": "EGFR", "type": "Small molecule",227 "highest_phase": "Approved", "first_approved": "2015-11-13",228 "indications": ["NSCLC"], "company": "AstraZeneca"},229 {"name": "Gefitinib", "target": "EGFR", "type": "Small molecule",230 "highest_phase": "Approved", "first_approved": "2003-05-05",231 "indications": ["NSCLC"], "company": "AstraZeneca"},232 {"name": "Cetuximab", "target": "EGFR", "type": "Monoclonal antibody",233 "highest_phase": "Approved", "first_approved": "2004-02-12",234 "indications": ["CRC", "HNSCC"], "company": "Merck KGaA / BMS"},235 {"name": "Trastuzumab deruxtecan", "target": "HER2", "type": "ADC",236 "highest_phase": "Approved", "first_approved": "2019-12-20",237 "indications": ["Breast Cancer", "Gastric Cancer", "NSCLC"],238 "company": "Daiichi Sankyo / AstraZeneca"},239 {"name": "Pembrolizumab", "target": "PD-1", "type": "Monoclonal antibody",240 "highest_phase": "Approved", "first_approved": "2014-09-04",241 "indications": ["Melanoma", "NSCLC", "HNSCC", "cHL"], "company": "Merck (MSD)"},242 {"name": "Amivantamab", "target": "EGFR/MET", "type": "Bispecific antibody",243 "highest_phase": "Approved", "first_approved": "2021-05-21",244 "indications": ["NSCLC (ex20ins)"], "company": "Janssen"},245 {"name": "Sotorasib", "target": "KRAS G12C", "type": "Small molecule",246 "highest_phase": "Approved", "first_approved": "2021-05-28",247 "indications": ["NSCLC (KRAS G12C)"], "company": "Amgen"},248 {"name": "Sacituzumab govitecan", "target": "Trop-2", "type": "ADC",249 "highest_phase": "Approved", "first_approved": "2020-04-22",250 "indications": ["TNBC", "HR+/HER2- Breast Cancer"], "company": "Gilead"},251 ]252}253 254# =============================================================================255# CSS — Product-grade styling256# =============================================================================257 258CUSTOM_CSS = """259/* ===== Design Tokens ===== */260:root {261 --navy-950: #020617;262 --navy-900: #0a1628;263 --navy-800: #122540;264 --navy-700: #1a3050;265 --navy-600: #1e3a5f;266 --teal-500: #0891b2;267 --teal-400: #06b6d4;268 --teal-50: #ecfeff;269 --emerald-500: #10b981;270 --emerald-50: #ecfdf5;271 --surface: #ffffff;272 --bg: #f8fafc;273 --bg-alt: #f1f5f9;274 --text: #1e293b;275 --text-secondary: #64748b;276 --text-tertiary: #94a3b8;277 --border: #e2e8f0;278 --border-light: #f1f5f9;279 --radius-sm: 6px;280 --radius: 10px;281 --radius-lg: 14px;282 --radius-xl: 20px;283 --shadow-xs: 0 1px 2px rgba(15,23,42,0.04);284 --shadow-sm: 0 1px 3px rgba(15,23,42,0.06);285 --shadow: 0 1px 3px rgba(15,23,42,0.08), 0 1px 2px rgba(15,23,42,0.04);286 --shadow-md: 0 4px 6px -1px rgba(15,23,42,0.08), 0 2px 4px -2px rgba(15,23,42,0.04);287 --font: 'Inter', -apple-system, BlinkMacSystemFont, 'SF Pro Display', 'Segoe UI', system-ui, sans-serif;288 --font-mono: 'SF Mono', 'JetBrains Mono', 'Fira Code', monospace;289}290 291@import url('https://fonts.googleapis.com/css2?family=Inter:wght@400;500;600;700&display=swap');292 293.gradio-container {294 max-width: 100% !important;295 padding: 0 32px !important;296 margin: 0 auto !important;297 font-family: var(--font) !important;298 background: var(--bg) !important;299 -webkit-font-smoothing: antialiased;300 -moz-osx-font-smoothing: grayscale;301}302 303/* Force ALL text dark — override Gradio Soft theme everywhere */304.gradio-container .prose,305.gradio-container .prose *,306.gradio-container .md,307.gradio-container .md *,308.gradio-container [class*="markdown"],309.gradio-container [class*="markdown"] * {310 color: var(--text) !important;311}312.gradio-container .prose h2, .gradio-container .md h2,313.gradio-container [class*="markdown"] h2 {314 color: var(--navy-950) !important;315 font-weight: 700 !important;316}317.gradio-container .prose strong, .gradio-container .md strong,318.gradio-container [class*="markdown"] strong {319 color: var(--navy-900) !important;320}321.gradio-container label, .gradio-container .label-text,322.gradio-container .label-container span {323 color: var(--text) !important;324 font-weight: 500 !important;325}326.gradio-container td { color: var(--text) !important; }327.gradio-container th { color: var(--text-secondary) !important; }328 329/* Ensure header text is always white and visible */330.header-container,331.header-container * {332 color: #ffffff !important;333}334.header-title {335 color: #ffffff !important;336}337.header-subtitle {338 color: rgba(255, 255, 255, 0.85) !important;339}340 341/* ===== Header ===== */342.header-container {343 background: linear-gradient(135deg, var(--navy-950) 0%, var(--navy-900) 30%, var(--navy-700) 70%, var(--navy-600) 100%);344 padding: 40px 48px 34px;345 margin-bottom: 28px;346 color: white;347 position: relative;348 overflow: hidden;349 border-bottom: 1px solid rgba(255,255,255,0.06);350}351.header-container::before {352 content: '';353 position: absolute;354 top: -120px;355 right: -80px;356 width: 420px;357 height: 420px;358 background: radial-gradient(circle, rgba(6,182,212,0.13) 0%, rgba(6,182,212,0.04) 40%, transparent 70%);359 border-radius: 50%;360 pointer-events: none;361}362.header-container::after {363 content: '';364 position: absolute;365 bottom: -60px;366 left: 30%;367 width: 300px;368 height: 300px;369 background: radial-gradient(circle, rgba(16,185,129,0.08) 0%, transparent 70%);370 border-radius: 50%;371 pointer-events: none;372}373.header-row {374 display: flex;375 align-items: flex-start;376 justify-content: space-between;377 position: relative;378 z-index: 1;379}380.header-title {381 font-size: 26px;382 font-weight: 700;383 margin: 0 0 10px 0;384 letter-spacing: -0.4px;385 line-height: 1.2;386 color: #ffffff;387}388.header-subtitle {389 font-size: 14px;390 opacity: 0.68;391 margin: 0;392 font-weight: 400;393 line-height: 1.6;394 max-width: 520px;395 color: rgba(255,255,255,0.85);396}397.header-badges {398 display: flex;399 gap: 8px;400 margin-top: 16px;401 flex-wrap: wrap;402}403.header-badge {404 display: inline-flex;405 align-items: center;406 gap: 5px;407 background: rgba(255,255,255,0.08);408 backdrop-filter: blur(12px);409 -webkit-backdrop-filter: blur(12px);410 padding: 5px 16px;411 border-radius: 20px;412 font-size: 11.5px;413 font-weight: 500;414 letter-spacing: 0.02em;415 border: 1px solid rgba(255,255,255,0.10);416 color: rgba(255,255,255,0.8);417}418 419/* ===== Tabs ===== */420.tabs { border: none !important; }421.tab-nav {422 background: transparent !important;423 border-bottom: 1px solid var(--border) !important;424 border-radius: 0 !important;425 padding: 0 4px !important;426 box-shadow: none !important;427 margin-bottom: 24px !important;428 gap: 4px !important;429}430.tab-nav button {431 border-radius: 8px 8px 0 0 !important;432 padding: 10px 20px !important;433 font-size: 13.5px !important;434 font-weight: 500 !important;435 border: none !important;436 color: #475569 !important;437 transition: color 0.15s ease, background 0.15s ease !important;438 background: transparent !important;439 border-bottom: 2px solid transparent !important;440 margin-bottom: -1px !important;441 cursor: pointer !important;442}443.tab-nav button:hover {444 color: var(--text) !important;445 background: var(--bg-alt) !important;446}447.tab-nav button.selected {448 background: transparent !important;449 color: #020617 !important;450 border-bottom: 2px solid #020617 !important;451 box-shadow: none !important;452 font-weight: 600 !important;453}454.tab-nav button:focus-visible {455 outline: 2px solid var(--navy-600) !important;456 outline-offset: -2px !important;457 border-radius: 8px 8px 0 0 !important;458}459 460/* Ensure all text elements are readable */461.gradio-container * {462 color: inherit !important;463}464.gradio-container button:not(.btn-primary):not(.example-chip) {465 color: #475569 !important;466}467.gradio-container button:not(.btn-primary):not(.example-chip):hover {468 color: #1e293b !important;469}470 471/* ===== Input ===== */472.agent-input textarea, .agent-input input {473 border-radius: var(--radius) !important;474 border: 1.5px solid var(--border) !important;475 padding: 14px 16px !important;476 font-size: 14.5px !important;477 line-height: 1.6 !important;478 transition: border-color 0.15s ease, box-shadow 0.15s ease !important;479 resize: none !important;480 background: var(--surface) !important;481 color: var(--text) !important;482}483.agent-input textarea:focus, .agent-input input:focus {484 border-color: var(--navy-800) !important;485 box-shadow: 0 0 0 3px rgba(18,37,64,0.08) !important;486 outline: none !important;487}488 489/* Example chips */490.example-chip {491 border-radius: 20px !important;492 padding: 6px 16px !important;493 font-size: 12.5px !important;494 font-weight: 500 !important;495 border: 1px solid var(--border) !important;496 background: var(--surface) !important;497 color: var(--text-secondary) !important;498 cursor: pointer !important;499 transition: all 0.15s ease !important;500 white-space: nowrap !important;501 min-width: unset !important;502 height: auto !important;503 line-height: 1.4 !important;504}505.example-chip:hover {506 border-color: var(--teal-400) !important;507 color: var(--teal-500) !important;508 background: var(--teal-50) !important;509}510.example-chip:focus-visible {511 outline: 2px solid var(--teal-400) !important;512 outline-offset: 1px !important;513}514 515/* Primary button */516.btn-primary {517 background: var(--navy-950) !important;518 color: white !important;519 border: none !important;520 border-radius: var(--radius) !important;521 padding: 11px 32px !important;522 font-size: 14px !important;523 font-weight: 600 !important;524 cursor: pointer !important;525 transition: background 0.15s ease, box-shadow 0.15s ease !important;526 letter-spacing: 0.01em;527}528.btn-primary:hover {529 background: var(--navy-800) !important;530 box-shadow: var(--shadow-md);531}532.btn-primary:focus-visible {533 outline: 2px solid var(--navy-600) !important;534 outline-offset: 2px !important;535}536 537/* Secondary button */538.btn-secondary {539 background: var(--bg-alt) !important;540 color: #475569 !important;541 border: 1px solid var(--border) !important;542 border-radius: var(--radius) !important;543 padding: 10px 24px !important;544 font-size: 14px !important;545 font-weight: 500 !important;546 cursor: pointer !important;547 transition: all 0.15s ease !important;548}549.btn-secondary:hover {550 background: var(--border) !important;551 color: #1e293b !important;552}553.btn-secondary:focus-visible {554 outline: 2px solid var(--navy-600) !important;555 outline-offset: 2px !important;556}557 558/* ===== Thinking Steps ===== */559.thinking-steps {560 background: linear-gradient(135deg, #f8faff, #f0fdfa);561 border: 1px solid #ccfbf1;562 border-radius: var(--radius);563 padding: 14px 18px;564 margin: 16px 0;565}566.thinking-step {567 padding: 3px 0;568 color: #115e59;569 font-size: 13px;570 line-height: 1.6;571}572 573/* ===== Cards ===== */574.card {575 background: var(--surface);576 border-radius: var(--radius-lg);577 padding: 24px;578 box-shadow: var(--shadow-xs);579 border: 1px solid var(--border);580 margin-bottom: 16px;581 transition: box-shadow 0.2s ease;582}583.card:hover { box-shadow: var(--shadow-sm); }584 585/* ===== Report Content ===== */586.report { color: var(--text); font-size: 14.5px; line-height: 1.75; }587.report h2 {588 font-size: 20px;589 font-weight: 700;590 margin-top: 28px;591 margin-bottom: 12px;592 color: var(--navy-950);593 letter-spacing: -0.3px;594}595.report h3 {596 font-size: 15px;597 font-weight: 600;598 margin-top: 20px;599 margin-bottom: 8px;600 color: var(--text);601}602.report table {603 width: 100%;604 border-collapse: collapse;605 margin: 14px 0;606 font-size: 13.5px;607 border-radius: var(--radius-sm);608 overflow: hidden;609 border: 1px solid var(--border);610}611.report th {612 background: var(--bg-alt);613 padding: 10px 14px;614 text-align: left;615 font-weight: 600;616 font-size: 11.5px;617 color: var(--text-secondary);618 text-transform: uppercase;619 letter-spacing: 0.05em;620 border-bottom: 1px solid var(--border);621}622.report td {623 padding: 10px 14px;624 border-bottom: 1px solid var(--border-light);625 color: var(--text);626}627.report tr:last-child td { border-bottom: none; }628.report tr:hover td { background: #fafcff; }629 630/* ===== Status Badge ===== */631.badge {632 display: inline-block;633 padding: 2px 10px;634 border-radius: 10px;635 font-size: 11px;636 font-weight: 600;637 letter-spacing: 0.02em;638}639.badge-approved { background: #dcfce7; color: #166534; }640.badge-phase3 { background: #dbeafe; color: #1e40af; }641.badge-phase2 { background: #fef3c7; color: #92400e; }642.badge-phase1 { background: #fce7f3; color: #9d174d; }643 644/* ===== Footer ===== */645.footer {646 text-align: center;647 padding: 28px 16px;648 color: var(--text-tertiary);649 font-size: 12px;650 border-top: 1px solid var(--border-light);651 margin-top: 48px;652 letter-spacing: 0.02em;653}654.footer strong { color: var(--text-secondary); font-weight: 600; }655 656/* ===== Animations ===== */657@keyframes fadeIn {658 from { opacity: 0; transform: translateY(8px); }659 to { opacity: 1; transform: translateY(0); }660}661.fade-in { animation: fadeIn 0.35s ease-out; }662 663@media (prefers-reduced-motion: reduce) {664 *, *::before, *::after {665 animation-duration: 0.01ms !important;666 animation-iteration-count: 1 !important;667 transition-duration: 0.01ms !important;668 }669 .fade-in { animation: none; }670}671 672/* Misc */673footer { display: none !important; }674::-webkit-scrollbar { width: 6px; }675::-webkit-scrollbar-track { background: transparent; }676::-webkit-scrollbar-thumb { background: var(--border); border-radius: 3px; }677"""678 679# =============================================================================680# MCP CLIENT681# =============================================================================682 683_mcp_tools_cache: Optional[List[str]] = None684 685async def get_mcp_tools() -> List[str]:686 """Fetch available MCP tool names with caching."""687 global _mcp_tools_cache688 if _mcp_tools_cache is not None:689 return _mcp_tools_cache690 if not API_KEY:691 return []692 try:693 from mcp import ClientSession694 from mcp.client.streamable_http import streamablehttp_client695 async with streamablehttp_client(SERVER_URL, timeout=15, sse_read_timeout=15) as (read, write, _):696 async with ClientSession(read, write) as session:697 await session.initialize()698 result = await session.list_tools()699 _mcp_tools_cache = [t.name for t in result.tools]700 return _mcp_tools_cache701 except Exception as e:702 print(f"[MCP] Tool discovery failed: {e}")703 return []704 705 706async def call_mcp_tool(tool_name: str, args: dict) -> Optional[dict]:707 """Call a specific MCP tool. Returns parsed JSON or None on failure."""708 if not API_KEY:709 return None710 try:711 from mcp import ClientSession712 from mcp.client.streamable_http import streamablehttp_client713 async with streamablehttp_client(SERVER_URL, timeout=30, sse_read_timeout=30) as (read, write, _):714 async with ClientSession(read, write) as session:715 await session.initialize()716 result = await session.call_tool(tool_name, arguments=args)717 if result.content:718 text = result.content[0].text719 return json.loads(text) if isinstance(text, str) else text720 except Exception as e:721 print(f"[MCP] {tool_name} failed: {e}")722 return None723 724 725# =============================================================================726# AGENT ENGINE — Intent Parsing & Tool Routing727# =============================================================================728 729# Entity extraction patterns730TARGET_KEYWORDS = {731 "egfr": "EGFR", "her2": "HER2", "her-2": "HER2", "erbb2": "HER2",732 "pd-l1": "PD-L1", "pdl1": "PD-L1", "pd-1": "PD-1", "pd1": "PD-1",733 "braf": "BRAF", "alk": "ALK", "ros1": "ROS1", "kras": "KRAS",734 "vegf": "VEGF", "vegfr": "VEGFR", "ctla-4": "CTLA-4", "ctla4": "CTLA-4",735 "stat3": "STAT3", "met": "c-MET", "c-met": "c-MET", "ret": "RET",736 "ntrk": "NTRK", "fgfr": "FGFR", "parp": "PARP", "cdk4": "CDK4/6",737 "cdk4/6": "CDK4/6", "btk": "BTK", "jak": "JAK", "flt3": "FLT3",738 "idh1": "IDH1", "idh2": "IDH2", "tp53": "TP53", "p53": "TP53",739 "brca1": "BRCA1", "brca2": "BRCA2", "trop-2": "Trop-2", "trop2": "Trop-2",740 "claudin": "Claudin 18.2", "cldn18": "Claudin 18.2",741}742 743DISEASE_KEYWORDS = {744 "non-small cell lung cancer": "NSCLC", "nsclc": "NSCLC", "lung cancer": "Lung Cancer",745 "breast cancer": "Breast Cancer", "triple-negative breast": "Triple-Negative Breast Cancer",746 "tnbc": "Triple-Negative Breast Cancer",747 "colorectal": "Colorectal Cancer", "crc": "Colorectal Cancer",748 "melanoma": "Melanoma",749 "pancreatic": "Pancreatic Cancer",750 "hepatocellular": "Hepatocellular Carcinoma", "hcc": "Hepatocellular Carcinoma",751 "liver cancer": "Hepatocellular Carcinoma",752 "gastric": "Gastric Cancer", "stomach cancer": "Gastric Cancer",753 "leukemia": "Leukemia", "lymphoma": "Lymphoma",754 "ovarian": "Ovarian Cancer", "prostate": "Prostate Cancer",755 "multiple myeloma": "Multiple Myeloma", "mm": "Multiple Myeloma",756 "renal cell": "Renal Cell Carcinoma", "rcc": "Renal Cell Carcinoma",757 "bladder": "Urothelial Carcinoma", "urothelial": "Urothelial Carcinoma",758 "head and neck": "Head and Neck Cancer", "hnscc": "Head and Neck Cancer",759 "glioblastoma": "Glioblastoma", "gbm": "Glioblastoma",760 "alzheimer": "Alzheimer's Disease", "parkinson": "Parkinson's Disease",761 "diabetes": "Diabetes", "obesity": "Obesity",762}763 764COMPANY_KEYWORDS = {765 "roche": "Roche", "genentech": "Roche",766 "novartis": "Novartis",767 "pfizer": "Pfizer",768 "merck": "Merck (MSD)", "msd": "Merck (MSD)",769 "bristol-myers": "Bristol-Myers Squibb", "bms": "Bristol-Myers Squibb",770 "astrazeneca": "AstraZeneca", "az": "AstraZeneca",771 "johnson": "Johnson & Johnson", "jnj": "Johnson & Johnson", "janssen": "Johnson & Johnson",772 "sanofi": "Sanofi",773 "gsk": "GlaxoSmithKline",774 "abbvie": "AbbVie",775 "amgen": "Amgen",776 "gilead": "Gilead",777 "lilly": "Eli Lilly", "eli lilly": "Eli Lilly",778 "moderna": "Moderna",779 "biontech": "BioNTech",780 "daiichi": "Daiichi Sankyo",781 "beigene": "BeiGene", "百济": "BeiGene",782 "innovent": "Innovent", "信达": "Innovent",783 "hengrui": "Hengrui", "恒瑞": "Hengrui",784 "akebia": "Akebia",785}786 787PHASE_KEYWORDS = {788 "phase 3": "phase_3", "phase iii": "phase_3", "phase iii": "phase_3", "pivotal": "phase_3",789 "phase 2": "phase_2", "phase ii": "phase_2",790 "phase 1": "phase_1", "phase i": "phase_1", "first-in-human": "phase_1", "fih": "phase_1",791 "approved": "approved", "marketed": "approved", "launched": "approved",792 "preclinical": "preclinical",793}794 795MODULE_KEYWORDS = {796 "target": ["target", "靶点", "receptor", "kinase", "protein", "gene", "mutation", "pathway",797 "inhibitor target", "antibody target", "drug target"],798 "drug": ["drug", "药物", "medicine", "inhibitor", "antibody", "therapy", "treatment regimen",799 "approved drug", "pipeline drug", "molecule", "compound", "modality"],800 "disease": ["disease", "疾病", "cancer", "tumor", "indication", "适应症", "epidemiology",801 "patients", "prevalence", "incidence", "mortality"],802 "company": ["company", "公司", "pharma", "biotech", "pipeline of", "portfolio",803 "acquisition", "merger", "partner", "revenue"],804 "trial": ["trial", "试验", "clinical", "nct", "enrollment", "endpoint", "randomized",805 "double-blind", "phase 3 trial", "phase 2 trial", "phase 1 trial"],806}807 808 809def parse_intent(query: str) -> Dict:810 """811 Parse a natural language query to determine:812 - module (target/drug/disease/company/trial)813 - entities (targets, diseases, companies, phases)814 - mcp_args (for direct MCP call)815 - confidence816 """817 lower = query.lower()818 result = {819 "module": "target", # default820 "entities": {"targets": [], "diseases": [], "companies": [], "phases": []},821 "mcp_args": {"limit": 10},822 "confidence": 0.0,823 "thinking": [],824 }825 826 # Extract entities827 for kw, val in TARGET_KEYWORDS.items():828 if kw in lower and val not in result["entities"]["targets"]:829 result["entities"]["targets"].append(val)830 831 for kw, val in DISEASE_KEYWORDS.items():832 if kw in lower and val not in result["entities"]["diseases"]:833 result["entities"]["diseases"].append(val)834 835 for kw, val in COMPANY_KEYWORDS.items():836 if kw in lower and val not in result["entities"]["companies"]:837 result["entities"]["companies"].append(val)838 839 for kw, val in PHASE_KEYWORDS.items():840 if kw in lower:841 result["entities"]["phases"].append(val)842 843 # Determine module by scoring keyword matches844 scores = {m: 0 for m in MODULE_KEYWORDS}845 for module, keywords in MODULE_KEYWORDS.items():846 for kw in keywords:847 if kw in lower:848 scores[module] += 1849 850 best_module = max(scores, key=scores.get)851 max_score = scores[best_module]852 853 # Heuristic overrides based on entities found854 if result["entities"]["targets"] and not result["entities"]["diseases"] and not result["entities"]["companies"]:855 if max_score == 0 or best_module == "drug":856 result["module"] = "target" if scores["target"] >= scores["drug"] else "drug"857 else:858 result["module"] = best_module859 elif result["entities"]["diseases"] and not result["entities"]["targets"]:860 result["module"] = "disease"861 elif result["entities"]["companies"] and not result["entities"]["targets"]:862 result["module"] = "company"863 elif "trial" in lower or "clinical" in lower or "enrollment" in lower:864 result["module"] = "trial"865 else:866 result["module"] = best_module if max_score > 0 else "target"867 868 # Build MCP args based on module869 mod = result["module"]870 if mod == "target" and result["entities"]["targets"]:871 result["mcp_args"] = {"target": result["entities"]["targets"][:3], "limit": 10}872 result["thinking"].append(f"🔍 Detected target query → searching for: {', '.join(result['entities']['targets'][:3])}")873 elif mod == "drug":874 args = {"limit": 10}875 if result["entities"]["targets"]:876 args["target"] = result["entities"]["targets"][:3]877 if result["entities"]["diseases"]:878 args["disease"] = result["entities"]["diseases"][:3]879 if result["entities"]["phases"]:880 args["highest_phase"] = result["entities"]["phases"][:3]881 result["mcp_args"] = args882 result["thinking"].append(f"🔍 Detected drug query → searching with {json.dumps(args)}")883 elif mod == "disease" and result["entities"]["diseases"]:884 result["mcp_args"] = {"disease": result["entities"]["diseases"][:3], "limit": 10}885 result["thinking"].append(f"🔍 Detected disease query → searching for: {', '.join(result['entities']['diseases'][:3])}")886 elif mod == "company" and result["entities"]["companies"]:887 result["mcp_args"] = {"company": result["entities"]["companies"][:3], "limit": 10}888 result["thinking"].append(f"🔍 Detected company query → searching for: {', '.join(result['entities']['companies'][:3])}")889 elif mod == "trial":890 args = {"limit": 10}891 if result["entities"]["targets"]:892 args["target"] = result["entities"]["targets"][:3]893 if result["entities"]["diseases"]:894 args["disease"] = result["entities"]["diseases"][:3]895 if result["entities"]["phases"]:896 args["phase"] = result["entities"]["phases"][:3]897 result["mcp_args"] = args898 result["thinking"].append(f"🔍 Detected trial query → searching with {json.dumps(args)}")899 900 # Confidence901 entity_count = sum(len(v) for v in result["entities"].values())902 result["confidence"] = min(entity_count * 0.25 + scores[result["module"]] * 0.15, 0.95)903 904 return result905 906 907# =============================================================================908# REPORT BUILDERS909# =============================================================================910 911def _badge(phase: str) -> str:912 """Generate an HTML badge for a drug phase."""913 phase_lower = phase.lower()914 if "approved" in phase_lower:915 cls = "badge-approved"916 elif "phase 3" in phase_lower or "phase iii" in phase_lower:917 cls = "badge-phase3"918 elif "phase 2" in phase_lower or "phase ii" in phase_lower:919 cls = "badge-phase2"920 elif "phase 1" in phase_lower or "phase i" in phase_lower:921 cls = "badge-phase1"922 else:923 cls = "badge-phase2"924 return f'<span class="badge {cls}">{phase}</span>'925 926 927def build_target_report(target_data: Dict) -> str:928 """Build a structured target intelligence report."""929 name = target_data.get("name", "Unknown")930 full = target_data.get("full_name", "")931 family = target_data.get("family", "N/A")932 cls = target_data.get("class", "N/A")933 drugs = target_data.get("approved_drugs", [])934 pipeline = target_data.get("pipeline_summary", {})935 pathways = target_data.get("pathways", [])936 mutations = target_data.get("mutation_hotspots", [])937 landscape = target_data.get("competitive_landscape", "")938 939 report = []940 report.append(f"## 🎯 {name} — Target Intelligence Report")941 report.append("")942 943 # Overview card944 report.append("### 📋 Overview")945 report.append(f"| Property | Value |")946 report.append(f"|----------|-------|")947 report.append(f"| **Full Name** | {full} |")948 report.append(f"| **Family** | {family} |")949 report.append(f"| **Class** | {cls} |")950 report.append(f"| **Approved Drugs** | {len(drugs)} |")951 report.append("")952 953 # Pathways954 if pathways:955 report.append("### 🧬 Signaling Pathways")956 for p in pathways:957 report.append(f"- {p}")958 report.append("")959 960 # Mutations961 if mutations:962 report.append("### 🔬 Key Mutations / Variants")963 for m in mutations:964 report.append(f"- {m}")965 report.append("")966 967 # Approved Drugs Table968 if drugs:969 report.append(f"### 💊 Approved Drugs ({len(drugs)})")970 report.append("| Drug | Type | Generation | Year | Indications | Company |")971 report.append("|------|------|-----------|------|-------------|---------|")972 for d in drugs:973 inds = ", ".join(d.get("indications", [])[:2])974 if len(d.get("indications", [])) > 2:975 inds += f" +{len(d['indications']) - 2} more"976 report.append(f"| {d['name']} | {d['type']} | {d.get('gen', '-')} | "977 f"{d['year']} | {inds} | {d.get('company', '-')} |")978 report.append("")979 980 # Pipeline981 if pipeline:982 report.append("### 🔬 Pipeline Overview")983 report.append(f"| Phase | Count |")984 report.append(f"|-------|-------|")985 for phase in ["phase_3", "phase_2", "phase_1", "preclinical"]:986 label = phase.replace("_", " ").title()987 report.append(f"| {label} | {pipeline.get(phase, 'N/A')} |")988 report.append("")989 hot = pipeline.get("hot_topics", [])990 if hot:991 report.append("**🔥 Hot Topics:**")992 for t in hot:993 report.append(f"- {t}")994 report.append("")995 996 # Competitive landscape997 if landscape:998 report.append("### 🏔️ Competitive Landscape")999 report.append(landscape)1000 report.append("")1001 1002 report.append("---")1003 report.append(f"*Report generated by PatSnap Pharma Intelligence Agent*")1004 return "\n".join(report)1005 1006 1007def _normalize_drug_item(item: Dict) -> Dict:1008 """Normalize server response fields to internal schema."""1009 if "display_name_en" in item or "drug_type_view" in item:1010 drug_types = item.get("drug_type_view", [])1011 type_str = ", ".join(d.get("display_name_en", "") for d in drug_types if d.get("display_name_en")) or "N/A"1012 1013 # Indications from research_disease_view (fetch) or indication_view (search)1014 indications = item.get("research_disease_view", []) or item.get("indication_view", [])1015 ind_list = [d.get("display_name_en", "") for d in indications if d.get("display_name_en")] or []1016 1017 # Company from originator_org_master_entity_id_view (fetch) or originator_view (search)1018 orgs = item.get("originator_org_master_entity_id_view", []) or item.get("originator_view", []) or item.get("organization_view", [])1019 company = ", ".join(d.get("display_name_en", "") for d in orgs if d.get("display_name_en")) or "N/A"1020 1021 # Phase from global_highest_dev_status_view (fetch) or highest_phase_view (search)1022 phase = item.get("global_highest_dev_status_view") or item.get("highest_phase_view", {})1023 phase_str = phase.get("display_name_en", "") if isinstance(phase, dict) else str(phase) if phase else "N/A"1024 1025 return {1026 "name": item.get("display_name_en") or item.get("display_name_cn") or "N/A",1027 "type": type_str,1028 "highest_phase": phase_str,1029 "first_approved": item.get("first_approved_date", ""),1030 "indications": ind_list,1031 "company": company,1032 }1033 return item1034 1035 1036def _normalize_items(items: List[Dict]) -> List[Dict]:1037 """Normalize a list of server response items."""1038 return [_normalize_drug_item(i) for i in items]1039 1040 1041# =============================================================================1042# ENTITY EXTRACTORS — Shape live MCP data into report-friendly dicts1043# =============================================================================1044 1045def _profile_text(item: Dict) -> str:1046 """Extract a profile description string from profile/profile_v2 fields."""1047 for key in ("profile_v2", "profile"):1048 val = item.get(key)1049 if isinstance(val, list) and val:1050 content = val[0].get("content", "")1051 if content:1052 return content1053 return ""1054 1055 1056def _aliases(item: Dict, max_n: int = 6) -> List[str]:1057 """Extract English aliases from a target/disease item."""1058 aliases = item.get("alias", []) or []1059 out, seen = [], set()1060 for a in aliases:1061 if isinstance(a, dict) and a.get("lang") == "EN":1062 n = a.get("name", "").strip()1063 if n and n not in seen:1064 seen.add(n)1065 out.append(n)1066 if len(out) >= max_n:1067 break1068 return out1069 1070 1071def extract_target(item: Dict) -> Dict:1072 """Extract structured target info from ls_target_fetch result."""1073 return {1074 "name": item.get("display_name_en") or item.get("display_name_cn", "N/A"),1075 "aliases": _aliases(item),1076 "profile": _profile_text(item),1077 "drug_count": item.get("drug_count_roll_up") or item.get("drug_count", 0),1078 "dev_drug_count": item.get("dev_drug_count_roll_up") or item.get("dev_drug_count", 0),1079 "disease_count": item.get("disease_count_roll_up") or item.get("disease_count", 0),1080 "uniprot_id": (item.get("uniprot_id") or [None])[0],1081 "hgnc_id": (item.get("hgnc_id") or [None])[0],1082 "chembl_id": (item.get("chembl_id") or [None])[0],1083 "organism": item.get("organisms") or item.get("source", ""),1084 }1085 1086 1087def extract_disease(item: Dict) -> Dict:1088 """Extract structured disease info from ls_disease_fetch result."""1089 return {1090 "name": item.get("display_name_en") or item.get("display_name_cn", "N/A"),1091 "aliases": _aliases(item),1092 "profile": _profile_text(item),1093 "dev_drug_count": item.get("dev_drug_count_roll_up") or item.get("dev_drug_count", 0),1094 "mesh_id": item.get("mesh_id"),1095 "umls_cui": (item.get("umls_cui") or [None])[0],1096 }1097 1098 1099def extract_organization(item: Dict) -> Dict:1100 """Extract structured org info from ls_organization_fetch result."""1101 country = item.get("country_id_view") or {}1102 fdate = item.get("founded_date")1103 founded_year = None1104 if fdate and isinstance(fdate, int):1105 s = str(fdate)1106 if len(s) >= 4:1107 founded_year = s[:4]1108 return {1109 "name": item.get("display_name_en") or item.get("name_en", "N/A"),1110 "description": item.get("short_description_en") or (item.get("long_description_en", "") or "")[:400],1111 "website": item.get("website", ""),1112 "country": country.get("display_name_en", "") if isinstance(country, dict) else "",1113 "founded": founded_year,1114 "employees": item.get("employee_number"),1115 "stock_exchange": item.get("stock_exchange_code", ""),1116 "stock_symbol": item.get("stock_symbol", ""),1117 "ownership": ", ".join(item.get("ownership_type", []) or []),1118 "drug_count": item.get("drug_count_roll_up") or item.get("drug_count", 0),1119 "dev_drug_count": item.get("dev_drug_count_roll_up") or item.get("dev_drug_count", 0),1120 "patent_count": item.get("patent_phs_count_roll_up") or item.get("patent_phs_count", 0),1121 }1122 1123 1124def extract_pipeline_drug(item: Dict) -> Dict:1125 """Extract a drug record from ls_organization_pipeline_fetch."""1126 targets = item.get("targets", []) or []1127 target_str = ", ".join(t.get("display_name_en", "") for t in targets if t.get("display_name_en")) or "—"1128 statuses = item.get("status_tables", []) or []1129 indications = []1130 phases_seen = set()1131 best_phase = ""1132 for s in statuses:1133 d = s.get("disease_id_view", {}) or {}1134 name = d.get("display_name_en", "")1135 if name and name not in indications:1136 indications.append(name)1137 ds = s.get("dev_status")1138 if isinstance(ds, list):1139 for entry in ds:1140 view = entry.get("dev_status_id_view", {}) or {}1141 ph = view.get("display_name_en", "")1142 if ph and ph not in phases_seen:1143 phases_seen.add(ph)1144 if not best_phase:1145 best_phase = ph1146 elif isinstance(ds, dict):1147 ph = ds.get("display_name_en", "")1148 if ph and ph not in phases_seen:1149 phases_seen.add(ph)1150 if not best_phase:1151 best_phase = ph1152 return {1153 "name": item.get("display_name_en") or item.get("display_name_cn", "N/A"),1154 "targets": target_str,1155 "indications": indications[:3],1156 "phase": best_phase or "—",1157 }1158 1159 1160def extract_trial(item: Dict) -> Dict:1161 """Extract structured trial info from ls_clinical_trial_fetch result."""1162 phase = item.get("clinical_phase") or {}1163 phase_str = phase.get("display_name_en", "") if isinstance(phase, dict) else str(phase)1164 sponsors = item.get("sponsor_organization", []) or []1165 sponsor_str = ", ".join(s.get("display_name_en", "") for s in sponsors if s.get("display_name_en"))1166 diseases = item.get("disease", []) or []1167 disease_str = ", ".join(d.get("display_name_en", "") for d in diseases if d.get("display_name_en"))1168 drugs = item.get("experiment_drug", []) or []1169 drug_str = ", ".join(d.get("display_name_en", "") for d in drugs if d.get("display_name_en"))1170 return {1171 "nct": item.get("registration_number", ""),1172 "title": item.get("trial_title", ""),1173 "status": item.get("trial_status", ""),1174 "phase": phase_str,1175 "sponsor": sponsor_str,1176 "disease": disease_str,1177 "drugs": drug_str,1178 }1179 1180 1181def build_drug_report(items: List[Dict], query_summary: str, total: int = 0) -> str:1182 """Build a drug pipeline report from search results."""1183 if not items:1184 return f"## 💊 Drug Search: {query_summary}\n\n📭 No results found. Try a different query."1185 1186 drug_types = Counter()1187 companies = Counter()1188 years = []1189 phases = Counter()1190 1191 for item in items:1192 drug_types[item.get("type", "Unknown")] += 11193 companies[item.get("company", "Unknown")] += 11194 phases[item.get("highest_phase", "Unknown")] += 11195 date = item.get("first_approved", "")1196 if date and date != "N/A":1197 try:1198 years.append(int(date.split("-")[0]))1199 except (ValueError, IndexError):1200 pass