fix(analyzer): route _analyze_company_safe through cache-aware path

_analyze_company_safe was calling SERP.query directly, bypassing the
SERP query cache in analyze_company. Now delegates fully to
analyze_company() and reads patent_count from the serp_queries cache.

.gitea/workflows/build.yaml

@@ -1,4 +1,4 @@
-name: Build and Push Docker Image
+name: Build and Push Docker Images
 
 on:
   push:
@@ -9,7 +9,7 @@ on:
   workflow_dispatch:
 
 jobs:
-  build-and-push:
+  build-api:
     runs-on: ubuntu-latest
     steps:
       - name: Install dependencies
@@ -20,43 +20,36 @@ jobs:
       - name: Checkout code
         shell: sh
         run: |
-          git clone https://gitea.leeworks.dev/${{ gitea.repository }}.git .
+          git clone http://gitea.gitea.svc.cluster.local/${{ gitea.repository }}.git .
           git checkout ${{ gitea.sha }}
 
       - name: Determine image tags
         id: tags
         shell: sh
         run: |
-          REGISTRY="gitea.leeworks.dev"
+          REGISTRY="gitea.gitea.svc.cluster.local:80"
           REPO_OWNER="${{ gitea.repository_owner }}"
           REPO_NAME="${{ gitea.repository }}"
 
-          # Extract repository name without owner
           REPO_NAME_ONLY=$(echo "$REPO_NAME" | cut -d'/' -f2)
-
-          # Convert to lowercase for Docker registry compatibility
           REPO_OWNER_LOWER=$(echo "$REPO_OWNER" | tr '[:upper:]' '[:lower:]')
           REPO_NAME_LOWER=$(echo "$REPO_NAME_ONLY" | tr '[:upper:]' '[:lower:]')
 
-          # Base image path
           IMAGE_BASE="${REGISTRY}/${REPO_OWNER_LOWER}/${REPO_NAME_LOWER}"
 
-          # Determine tag based on ref
           case "${{ gitea.ref }}" in
             refs/tags/*)
-              # Tag push - use the tag name
               TAG_NAME="${{ gitea.ref_name }}"
               echo "IMAGE_TAG=${IMAGE_BASE}:${TAG_NAME}" >> $GITHUB_OUTPUT
               echo "PUSH_LATEST=true" >> $GITHUB_OUTPUT
               ;;
             refs/heads/main)
-              # Main branch - use commit SHA (shortened to 7 chars) and latest
+              TIMESTAMP=$(date -u +%Y%m%d%H%M%S)
               SHORT_SHA=$(echo "${{ gitea.sha }}" | cut -c1-7)
-              echo "IMAGE_TAG=${IMAGE_BASE}:${SHORT_SHA}" >> $GITHUB_OUTPUT
+              echo "IMAGE_TAG=${IMAGE_BASE}:${TIMESTAMP}-${SHORT_SHA}" >> $GITHUB_OUTPUT
               echo "PUSH_LATEST=true" >> $GITHUB_OUTPUT
               ;;
             *)
-              # Other branches - use branch name
               BRANCH_TAG=$(echo "${{ gitea.ref_name }}" | sed 's/\//-/g')
               echo "IMAGE_TAG=${IMAGE_BASE}:${BRANCH_TAG}" >> $GITHUB_OUTPUT
               echo "PUSH_LATEST=false" >> $GITHUB_OUTPUT
@@ -68,15 +61,15 @@ jobs:
       - name: Login to registry
         shell: sh
         run: |
-          echo "${{ secrets.PERSONAL_TOKEN }}" | docker login gitea.leeworks.dev -u "${{ gitea.actor }}" --password-stdin
+          echo "${{ secrets.PERSONAL_TOKEN }}" | docker login gitea.gitea.svc.cluster.local:80 -u "${{ gitea.actor }}" --password-stdin
 
-      - name: Build and push with Docker
+      - name: Build and push API image
         shell: sh
         run: |
-          echo "Building image..."
+          echo "Building API image..."
           docker build -t ${{ steps.tags.outputs.IMAGE_TAG }} .
 
-          echo "Pushing image..."
+          echo "Pushing API image..."
           docker push ${{ steps.tags.outputs.IMAGE_TAG }}
 
           if [ "${{ steps.tags.outputs.PUSH_LATEST }}" = "true" ]; then
@@ -85,5 +78,75 @@ jobs:
             docker push ${{ steps.tags.outputs.IMAGE_LATEST }}
           fi
 
-          echo "Build and push completed successfully!"
-          echo "Image available at ${{ steps.tags.outputs.IMAGE_TAG }}"
+          echo "API image available at ${{ steps.tags.outputs.IMAGE_TAG }}"
+
+  build-frontend:
+    runs-on: ubuntu-latest
+    steps:
+      - name: Install dependencies
+        shell: sh
+        run: |
+          apk add --no-cache git docker-cli
+
+      - name: Checkout code
+        shell: sh
+        run: |
+          git clone http://gitea.gitea.svc.cluster.local/${{ gitea.repository }}.git .
+          git checkout ${{ gitea.sha }}
+
+      - name: Determine image tags
+        id: tags
+        shell: sh
+        run: |
+          REGISTRY="gitea.gitea.svc.cluster.local:80"
+          REPO_OWNER="${{ gitea.repository_owner }}"
+          REPO_NAME="${{ gitea.repository }}"
+
+          REPO_NAME_ONLY=$(echo "$REPO_NAME" | cut -d'/' -f2)
+          REPO_OWNER_LOWER=$(echo "$REPO_OWNER" | tr '[:upper:]' '[:lower:]')
+          REPO_NAME_LOWER=$(echo "$REPO_NAME_ONLY" | tr '[:upper:]' '[:lower:]')
+
+          IMAGE_BASE="${REGISTRY}/${REPO_OWNER_LOWER}/${REPO_NAME_LOWER}"
+
+          case "${{ gitea.ref }}" in
+            refs/tags/*)
+              TAG_NAME="${{ gitea.ref_name }}"
+              echo "IMAGE_TAG=${IMAGE_BASE}:frontend-${TAG_NAME}" >> $GITHUB_OUTPUT
+              echo "PUSH_LATEST=true" >> $GITHUB_OUTPUT
+              ;;
+            refs/heads/main)
+              TIMESTAMP=$(date -u +%Y%m%d%H%M%S)
+              SHORT_SHA=$(echo "${{ gitea.sha }}" | cut -c1-7)
+              echo "IMAGE_TAG=${IMAGE_BASE}:frontend-${TIMESTAMP}-${SHORT_SHA}" >> $GITHUB_OUTPUT
+              echo "PUSH_LATEST=true" >> $GITHUB_OUTPUT
+              ;;
+            *)
+              BRANCH_TAG=$(echo "${{ gitea.ref_name }}" | sed 's/\//-/g')
+              echo "IMAGE_TAG=${IMAGE_BASE}:frontend-${BRANCH_TAG}" >> $GITHUB_OUTPUT
+              echo "PUSH_LATEST=false" >> $GITHUB_OUTPUT
+              ;;
+          esac
+
+          echo "IMAGE_LATEST=${IMAGE_BASE}:frontend-latest" >> $GITHUB_OUTPUT
+
+      - name: Login to registry
+        shell: sh
+        run: |
+          echo "${{ secrets.PERSONAL_TOKEN }}" | docker login gitea.gitea.svc.cluster.local:80 -u "${{ gitea.actor }}" --password-stdin
+
+      - name: Build and push frontend image
+        shell: sh
+        run: |
+          echo "Building frontend image..."
+          docker build -t ${{ steps.tags.outputs.IMAGE_TAG }} ./frontend
+
+          echo "Pushing frontend image..."
+          docker push ${{ steps.tags.outputs.IMAGE_TAG }}
+
+          if [ "${{ steps.tags.outputs.PUSH_LATEST }}" = "true" ]; then
+            echo "Tagging and pushing frontend-latest..."
+            docker tag ${{ steps.tags.outputs.IMAGE_TAG }} ${{ steps.tags.outputs.IMAGE_LATEST }}
+            docker push ${{ steps.tags.outputs.IMAGE_LATEST }}
+          fi
+
+          echo "Frontend image available at ${{ steps.tags.outputs.IMAGE_TAG }}"

.gitlab-ci.yml

@@ -1,33 +0,0 @@
-stages:
-  - build
-
-variables:
-  DOCKER_DRIVER: overlay2
-  DOCKER_TLS_CERTDIR: "/certs"
-  IMAGE_TAG: $CI_REGISTRY_IMAGE:$CI_COMMIT_REF_SLUG
-  LATEST_TAG: $CI_REGISTRY_IMAGE:latest
-
-build-and-push:
-  stage: build
-  image: docker:24-cli
-  services:
-    - docker:24-dind
-  before_script:
-    - echo "Logging into GitLab Container Registry..."
-    - docker login -u $CI_REGISTRY_USER -p $CI_REGISTRY_PASSWORD $CI_REGISTRY
-  script:
-    - echo "Building Docker image..."
-    - docker build -t $IMAGE_TAG -t $LATEST_TAG .
-    - echo "Pushing Docker image to registry..."
-    - docker push $IMAGE_TAG
-    - docker push $LATEST_TAG
-    - echo "Build and push completed successfully!"
-    - echo "Image available at $IMAGE_TAG"
-  rules:
-    - if: $CI_COMMIT_BRANCH == "main"
-      when: always
-    - if: $CI_COMMIT_TAG
-      when: always
-    - when: manual
-  tags:
-    - docker

README.md

@@ -17,7 +17,7 @@ SPARC automatically collects, parses, and analyzes patents from companies to pro
 - **Portfolio Analysis**: Evaluates multiple patents holistically for comprehensive insights
 - **Batch Processing**: Analyze multiple companies concurrently with progress tracking
 - **REST API**: FastAPI web service with async job support
-- **Dashboard**: Interactive Streamlit visualization dashboard
+- **Dashboard**: React TypeScript web dashboard with authentication
 - **Robust Testing**: 40 tests covering all major functionality
 
 ## Architecture
@@ -27,7 +27,9 @@ SPARC/
 ├── serp_api.py       # Patent retrieval and PDF parsing
 ├── llm.py            # Claude AI integration via OpenRouter
 ├── analyzer.py       # High-level orchestration
-├── api.py            # FastAPI web service
+├── api.py            # FastAPI web service with auth endpoints
+├── auth.py           # JWT authentication module
+├── database.py       # PostgreSQL storage with caching
 ├── types.py          # Data models
 └── config.py         # Environment configuration
 ```
@@ -48,7 +50,7 @@ docker-compose up -d
 
 # Access the services
 # - API: http://localhost:8000
-# - Dashboard: http://localhost:8501
+# - Dashboard: http://localhost:8080
 # - API Docs: http://localhost:8000/docs
 ```
 
@@ -186,21 +188,22 @@ curl -X POST http://localhost:8000/analyze/batch/async \
   -d '{"companies": ["nvidia", "amd", "intel", "qualcomm"]}'
 ```
 
-### Visualization Dashboard
+### Web Dashboard
 
-Launch the interactive Streamlit dashboard:
+The React dashboard is included in Docker Compose:
 
 ```bash
-streamlit run dashboard.py
+docker-compose up -d
 ```
 
 Dashboard features:
+- **Authentication**: User registration, login, and JWT-based sessions
 - **Company Analysis**: Analyze individual companies with real-time results
-- **Batch Analysis**: Process multiple companies with progress tracking and charts
-- **Analytics**: View historical analysis data and trends (requires database mode)
-- **System Status**: Monitor database and analyzer health
+- **Batch Analysis**: Process multiple companies with progress tracking
+- **Analytics**: View historical analysis data and trends
+- **Admin Panel**: User management for administrators
 
-The dashboard runs at `http://localhost:8501` by default.
+The dashboard runs at `http://localhost:8080` when using Docker Compose.
 
 ## Running Tests
 
@@ -280,4 +283,4 @@ For open source projects, say how it is licensed.
 
 Core functionality complete. Ready for production use with API keys configured.
 
-All major features implemented: REST API, Streamlit dashboard, Docker containerization, database storage, and multi-company batch processing.
+All major features implemented: REST API, React dashboard with authentication, Docker containerization, database storage with caching, and multi-company batch processing.

SPARC/analyzer.py

@@ -4,26 +4,33 @@ This module ties together patent retrieval, parsing, and LLM analysis
 to provide company performance estimation based on patent portfolios.
 """
 
+import hashlib
 from concurrent.futures import ThreadPoolExecutor, as_completed
 from typing import Callable
 
+from SPARC import config
+from SPARC.database import DatabaseClient
 from SPARC.serp_api import SERP
 from SPARC.llm import LLMAnalyzer
-from SPARC.types import Patent, CompanyAnalysisResult, BatchAnalysisResult
+from SPARC.types import Patent, Patents, CompanyAnalysisResult, BatchAnalysisResult
 
 
 class CompanyAnalyzer:
     """Orchestrates end-to-end company performance analysis via patents."""
 
-    def __init__(self, openrouter_api_key: str | None = None):
+    def __init__(self, openrouter_api_key: str | None = None, db_client: DatabaseClient | None = None):
         """Initialize the company analyzer.
 
         Args:
           openrouter_api_key: Optional OpenRouter API key. If None, loads from config.
+          db_client: Optional DatabaseClient for patent caching. Created automatically if None.
         """
         self.llm_analyzer = LLMAnalyzer(api_key=openrouter_api_key)
+        self.db = db_client or DatabaseClient(config.database_url)
+        self.db.connect()
+        self.db.initialize_schema()
 
-    def analyze_company(self, company_name: str) -> str:
+    def analyze_company(self, company_name: str, patents: "Patents | None" = None) -> str:
         """Analyze a company's performance based on their patent portfolio.
 
         This is the main entry point that orchestrates the full pipeline:
@@ -35,40 +42,52 @@ class CompanyAnalyzer:
 
         Args:
           company_name: Name of the company to analyze
+          patents: Optional pre-fetched Patents result to avoid duplicate API calls
 
         Returns:
           Comprehensive analysis of company's innovation and performance outlook
         """
-        print(f"Retrieving patents for {company_name}...")
-        patents = SERP.query(company_name)
+        if patents is None:
+            # Check SERP query cache first
+            query_hash = hashlib.sha256(company_name.lower().encode()).hexdigest()
+            cached_ids = self.db.get_cached_serp_query(query_hash)
+            if cached_ids is not None:
+                print(f"Using cached SERP results for {company_name} ({len(cached_ids)} patents)")
+                patents = Patents(patents=[
+                    Patent(patent_id=pid, pdf_link="")
+                    for pid in cached_ids
+                ])
+            else:
+                print(f"Retrieving patents for {company_name}...")
+                patents = SERP.query(company_name)
+                # Cache the SERP results
+                if patents.patents:
+                    self.db.store_serp_query(
+                        company_name=company_name,
+                        query_hash=query_hash,
+                        patent_ids=[p.patent_id for p in patents.patents],
+                    )
 
         if not patents.patents:
             return f"No patents found for {company_name}"
 
         print(f"Found {len(patents.patents)} patents. Processing...")
 
-        # Download and parse each patent
+        # Download, parse, and minimize patents in parallel
         processed_patents = []
-        for idx, patent in enumerate(patents.patents, 1):
-            print(f"Processing patent {idx}/{len(patents.patents)}: {patent.patent_id}")
-
-            try:
-                # Download PDF
-                patent = SERP.save_patents(patent)
-
-                # Parse sections from PDF
-                sections = SERP.parse_patent_pdf(patent.pdf_path)
-
-                # Minimize for LLM (remove bloat)
-                minimized_content = SERP.minimize_patent_for_llm(sections)
-
-                processed_patents.append(
-                    {"patent_id": patent.patent_id, "content": minimized_content}
-                )
-
-            except Exception as e:
-                print(f"Warning: Failed to process {patent.patent_id}: {e}")
-                continue
+        with ThreadPoolExecutor(max_workers=config.patent_thread_workers) as executor:
+            future_to_patent = {
+                executor.submit(self._process_single_patent, patent, company_name, self.db): patent
+                for patent in patents.patents
+            }
+            for future in as_completed(future_to_patent):
+                patent = future_to_patent[future]
+                try:
+                    result = future.result()
+                    if result:
+                        processed_patents.append(result)
+                except Exception as e:
+                    print(f"Warning: Failed to process {patent.patent_id}: {e}")
 
         if not processed_patents:
             return f"Failed to process any patents for {company_name}"
@@ -113,6 +132,46 @@ class CompanyAnalyzer:
         except Exception as e:
             return f"Failed to analyze patent {patent_id}: {e}"
 
+    @staticmethod
+    def _process_single_patent(
+        patent: Patent,
+        company_name: str = "",
+        db: DatabaseClient | None = None,
+    ) -> dict | None:
+        """Download, parse, and minimize a single patent. Thread-safe.
+
+        Checks DB cache before downloading. Stores results after processing.
+
+        Returns:
+            Dict with patent_id and minimized content, or None on failure.
+        """
+        try:
+            # Check DB cache first
+            if db:
+                cached = db.get_cached_patent(patent.patent_id)
+                if cached and cached.get("minimized_content"):
+                    return {"patent_id": patent.patent_id, "content": cached["minimized_content"]}
+
+            # Full processing: download, parse, minimize
+            patent = SERP.save_patents(patent)
+            sections = SERP.parse_patent_pdf(patent.pdf_path)
+            minimized_content = SERP.minimize_patent_for_llm(sections)
+
+            # Store in DB cache
+            if db:
+                db.store_patent(
+                    patent_id=patent.patent_id,
+                    company_name=company_name,
+                    pdf_link=patent.pdf_link,
+                    raw_sections=sections,
+                    minimized_content=minimized_content,
+                )
+
+            return {"patent_id": patent.patent_id, "content": minimized_content}
+        except Exception as e:
+            print(f"Warning: Failed to process {patent.patent_id}: {e}")
+            return None
+
     def _analyze_company_safe(self, company_name: str) -> CompanyAnalysisResult:
         """Internal wrapper that catches exceptions and returns structured result.
 
@@ -123,11 +182,14 @@ class CompanyAnalyzer:
             CompanyAnalysisResult with success/failure status
         """
         try:
-            patents = SERP.query(company_name)
-            patent_count = len(patents.patents) if patents.patents else 0
-
+            # Delegate to analyze_company which handles SERP/patent caching
             analysis = self.analyze_company(company_name)
 
+            # Determine patent count from cached SERP query
+            query_hash = hashlib.sha256(company_name.lower().encode()).hexdigest()
+            cached_ids = self.db.get_cached_serp_query(query_hash)
+            patent_count = len(cached_ids) if cached_ids else 0
+
             # Check if analysis indicates failure
             if analysis.startswith("No patents found") or analysis.startswith(
                 "Failed to process"

SPARC/api.py

@@ -161,6 +161,7 @@ app = FastAPI(
     description="Semiconductor Patent & Analytics Report Core - Patent portfolio analysis using AI",
     version="1.0.0",
     lifespan=lifespan,
+    root_path=config.root_path,
 )
 
 # Add CORS middleware for React frontend

SPARC/config.py

@@ -25,3 +25,11 @@ use_cache = os.getenv("USE_CACHE", "true").lower() in ("true", "1", "yes")
 # Legacy compatibility - USE_DATABASE is deprecated, database is always used
 # This variable is kept for backwards compatibility but has no effect
 use_database = os.getenv("USE_DATABASE", "false").lower() in ("true", "1", "yes")
+
+# Patent search configuration
+patent_search_days = int(os.getenv("PATENT_SEARCH_DAYS", "90"))
+patent_thread_workers = int(os.getenv("PATENT_THREAD_WORKERS", "5"))
+
+# Root path for running behind a reverse proxy (e.g., "/api" when served at /api/)
+# This ensures OpenAPI docs work correctly when accessed via the proxy
+root_path = os.getenv("ROOT_PATH", "")

SPARC/database.py

@@ -1,9 +1,11 @@
 """Database client for storing and retrieving LLM messages and user authentication."""
 
+import contextlib
 import psycopg2
+from psycopg2.pool import ThreadedConnectionPool
 from psycopg2.extras import RealDictCursor
 from typing import Dict, List, Optional
-from datetime import datetime
+from datetime import datetime, timedelta
 import json
 import hashlib
 import bcrypt
@@ -12,24 +14,49 @@ import bcrypt
 class DatabaseClient:
     """Handles database operations for message storage and retrieval."""
 
-    def __init__(self, database_url: str):
+    def __init__(self, database_url: str, minconn: int = 2, maxconn: int = 10):
         """Initialize the database client.
 
         Args:
             database_url: PostgreSQL connection string
+            minconn: Minimum connections in the pool
+            maxconn: Maximum connections in the pool
         """
         self.database_url = database_url
+        self._pool: ThreadedConnectionPool | None = None
+        self._minconn = minconn
+        self._maxconn = maxconn
+        # Legacy single connection kept for backwards compatibility
         self.conn = None
 
+    def _ensure_pool(self):
+        """Create the connection pool if it doesn't exist yet."""
+        if self._pool is None or self._pool.closed:
+            self._pool = ThreadedConnectionPool(
+                self._minconn, self._maxconn, self.database_url
+            )
+
+    @contextlib.contextmanager
+    def get_conn(self):
+        """Check out a connection from the pool. Returns it on exit."""
+        self._ensure_pool()
+        conn = self._pool.getconn()
+        try:
+            yield conn
+        finally:
+            self._pool.putconn(conn)
+
     def connect(self):
-        """Establish database connection."""
+        """Establish database connection (legacy single-connection path)."""
         if not self.conn or self.conn.closed:
             self.conn = psycopg2.connect(self.database_url)
 
     def close(self):
-        """Close database connection."""
+        """Close database connection and pool."""
         if self.conn and not self.conn.closed:
             self.conn.close()
+        if self._pool and not self._pool.closed:
+            self._pool.closeall()
 
     def initialize_schema(self):
         """Create database tables if they don't exist."""
@@ -110,6 +137,40 @@ class DatabaseClient:
                 ON users(email)
             """)
 
+            # Create patents cache table
+            cursor.execute("""
+                CREATE TABLE IF NOT EXISTS patents (
+                    patent_id VARCHAR(64) PRIMARY KEY,
+                    company_name VARCHAR(255),
+                    pdf_link TEXT,
+                    raw_sections JSONB,
+                    minimized_content TEXT,
+                    created_at TIMESTAMP DEFAULT CURRENT_TIMESTAMP
+                )
+            """)
+
+            cursor.execute("""
+                CREATE INDEX IF NOT EXISTS idx_patents_company
+                ON patents(company_name)
+            """)
+
+            # Create SERP query cache table
+            cursor.execute("""
+                CREATE TABLE IF NOT EXISTS serp_queries (
+                    id SERIAL PRIMARY KEY,
+                    company_name VARCHAR(255),
+                    query_hash VARCHAR(64) UNIQUE,
+                    result_patent_ids TEXT[],
+                    expires_at TIMESTAMP NOT NULL,
+                    created_at TIMESTAMP DEFAULT CURRENT_TIMESTAMP
+                )
+            """)
+
+            cursor.execute("""
+                CREATE INDEX IF NOT EXISTS idx_serp_queries_hash
+                ON serp_queries(query_hash)
+            """)
+
             self.conn.commit()
 
     @staticmethod
@@ -320,6 +381,87 @@ class DatabaseClient:
                 "period_days": days,
             }
 
+    # Patent Cache Methods
+
+    def get_cached_patent(self, patent_id: str) -> Optional[Dict]:
+        """Look up a cached patent by ID.
+
+        Returns:
+            Dict with raw_sections and minimized_content, or None.
+        """
+        with self.get_conn() as conn:
+            with conn.cursor(cursor_factory=RealDictCursor) as cursor:
+                cursor.execute(
+                    "SELECT * FROM patents WHERE patent_id = %s",
+                    (patent_id,),
+                )
+                row = cursor.fetchone()
+                return dict(row) if row else None
+
+    def store_patent(
+        self,
+        patent_id: str,
+        company_name: str,
+        pdf_link: str,
+        raw_sections: Dict,
+        minimized_content: str,
+    ) -> None:
+        """Store a processed patent in the cache."""
+        with self.get_conn() as conn:
+            with conn.cursor() as cursor:
+                cursor.execute(
+                    """
+                    INSERT INTO patents (patent_id, company_name, pdf_link, raw_sections, minimized_content)
+                    VALUES (%s, %s, %s, %s, %s)
+                    ON CONFLICT (patent_id) DO UPDATE SET
+                        raw_sections = EXCLUDED.raw_sections,
+                        minimized_content = EXCLUDED.minimized_content
+                    """,
+                    (patent_id, company_name, pdf_link, json.dumps(raw_sections), minimized_content),
+                )
+            conn.commit()
+
+    def get_cached_serp_query(self, query_hash: str) -> Optional[List[str]]:
+        """Look up cached SERP query results.
+
+        Returns:
+            List of patent IDs if cache hit and not expired, None otherwise.
+        """
+        with self.get_conn() as conn:
+            with conn.cursor(cursor_factory=RealDictCursor) as cursor:
+                cursor.execute(
+                    """
+                    SELECT result_patent_ids FROM serp_queries
+                    WHERE query_hash = %s AND expires_at > NOW()
+                    """,
+                    (query_hash,),
+                )
+                row = cursor.fetchone()
+                return row["result_patent_ids"] if row else None
+
+    def store_serp_query(
+        self,
+        company_name: str,
+        query_hash: str,
+        patent_ids: List[str],
+        ttl_hours: int = 24,
+    ) -> None:
+        """Store SERP query results in the cache."""
+        expires_at = datetime.now() + timedelta(hours=ttl_hours)
+        with self.get_conn() as conn:
+            with conn.cursor() as cursor:
+                cursor.execute(
+                    """
+                    INSERT INTO serp_queries (company_name, query_hash, result_patent_ids, expires_at)
+                    VALUES (%s, %s, %s, %s)
+                    ON CONFLICT (query_hash) DO UPDATE SET
+                        result_patent_ids = EXCLUDED.result_patent_ids,
+                        expires_at = EXCLUDED.expires_at
+                    """,
+                    (company_name, query_hash, patent_ids, expires_at),
+                )
+            conn.commit()
+
     # User Authentication Methods
 
     @staticmethod

SPARC/serp_api.py

@@ -1,17 +1,20 @@
+import os
 import serpapi
 from SPARC import config
 import re
 import pdfplumber  # pip install pdfplumber
 import requests
+from datetime import datetime, timedelta
 from typing import Dict
 from SPARC.types import Patents, Patent
 
 class SERP:
-  def query(company: str) -> Patents:
+  def query(company: str, days_back: int = None) -> Patents:
     """Query Google Patents for a company's recent patents.
 
     Args:
         company: Name of the company to search for
+        days_back: Number of days to look back for patents (default from config)
 
     Returns:
         Patents object containing list of patents with PDF links
@@ -23,13 +26,19 @@ class SERP:
         patents with restricted access). The returned count may be lower
         than the requested number of results.
     """
+    if days_back is None:
+        days_back = config.patent_search_days
+    end_date = datetime.now()
+    start_date = end_date - timedelta(days=days_back)
+    date_filter = f"cdr:1,cd_min:{start_date.strftime('%-m/%-d/%Y')},cd_max:{end_date.strftime('%-m/%-d/%Y')}"
+
     # Make API call
     params = {
       "engine": "google_patents",
       "q": company,
       "num": 10,
       "filter": 1,
-      "tbs": "cdr:1,cd_min:10/28/2025,cd_max:11/4/2025",
+      "tbs": date_filter,
       "api_key": config.api_key,
     }
     search = serpapi.search(params)
@@ -46,20 +55,23 @@ class SERP:
 
   def save_patents(patent: Patent) -> Patent:
     """
-    Save the patent PDF to the patents folder
-    
+    Save the patent PDF to the patents folder, skipping download if already cached.
+
     Args:
       patent: Patent object
 
     Returns:
       Patent object with updated PDF path
     """
-    response = requests.get(patent.pdf_link)
-    print(patent.pdf_link)
-    with open(f"patents/{patent.patent_id}.pdf", "wb") as f:
-      f.write(response.content)
-    
-    patent.pdf_path = f"patents/{patent.patent_id}.pdf"
+    pdf_path = f"patents/{patent.patent_id}.pdf"
+    os.makedirs("patents", exist_ok=True)
+
+    if not (os.path.exists(pdf_path) and os.path.getsize(pdf_path) > 0):
+      response = requests.get(patent.pdf_link)
+      with open(pdf_path, "wb") as f:
+        f.write(response.content)
+
+    patent.pdf_path = pdf_path
     return patent
 
   def parse_patent_pdf(pdf_path: str) -> Dict:

docker-compose.yml

@@ -38,6 +38,7 @@ services:
       DATABASE_URL: postgresql://postgres:postgres@postgres:5432/sparc
       USE_CACHE: "true"
       JWT_SECRET: ${JWT_SECRET:-sparc-secret-key-change-in-production}
+      ROOT_PATH: /api
     ports:
       - "8000:8000"
     depends_on:

docs/DATABASE_MODE.md

@@ -1,16 +1,19 @@
-# Database Mode for Testing and Analytics
+# Database Storage and Caching
 
-This document explains how to use SPARC's database mode for storing LLM messages for testing and analytics purposes.
+This document explains how SPARC uses PostgreSQL for storing LLM messages, enabling response caching and analytics.
 
 ## Overview
 
-SPARC supports two modes of operation:
+SPARC stores all LLM interactions in PostgreSQL, providing:
 
-1. **API Mode** (default): Messages are sent to OpenRouter's API and you receive real LLM responses
-2. **Database Mode**: Messages are stored in a PostgreSQL database without making API calls, useful for:
-   - Testing the application without consuming API credits
-   - Collecting analytics on message patterns and usage
-   - Development and debugging
+- **Response Caching**: Avoid redundant API calls for previously analyzed patents
+- **Analytics**: Track usage patterns, token consumption, and analysis history
+- **Persistence**: Maintain analysis history across sessions
+
+SPARC supports two cache modes:
+
+1. **Cache Mode** (default, `USE_CACHE=true`): Check database for cached responses before making API calls
+2. **Fresh Mode** (`USE_CACHE=false`): Always make fresh API calls (still stores results in database)
 
 ## Setup
 
@@ -45,43 +48,43 @@ cp .env.example .env
 Edit `.env` and set:
 
 ```env
-# For database mode (testing/analytics)
-USE_DATABASE=true
+# Database connection (required)
 DATABASE_URL=postgresql://postgres:postgres@localhost:5432/sparc
 
-# For API mode (production)
-USE_DATABASE=false
+# Cache mode: use cached responses when available
+USE_CACHE=true
+
+# API key for fresh LLM calls
 OPENROUTER_API_KEY=your_openrouter_key_here
 ```
 
 ## Usage
 
-### Running in Database Mode
+### Running with Cache Mode (Default)
 
-Set `USE_DATABASE=true` in your `.env` file, then run the application normally:
+Set `USE_CACHE=true` in your `.env` file, then run the application normally:
 
 ```bash
 python main.py
 ```
 
-Instead of sending messages to OpenRouter, the application will:
-- Store all prompts in the database
-- Return a placeholder response
-- Log metadata (company name, analysis type, timestamps)
+The application will:
+- Check the database for cached responses matching the request
+- If found, return the cached response (no API call)
+- If not found, make an API call and store the response for future use
 
-### Running in API Mode
+### Running with Fresh Mode
 
-Set `USE_DATABASE=false` in your `.env` file, then run the application normally:
+Set `USE_CACHE=false` in your `.env` file to always get fresh responses:
 
 ```bash
 python main.py
 ```
 
-The application will send messages to OpenRouter and return real LLM responses.
-
-### Hybrid Mode (Optional)
-
-You can also enable database logging while still using the API by initializing the database client in your code. The `LLMAnalyzer` will automatically log all API calls to the database if a database client is available.
+The application will:
+- Always send messages to OpenRouter for real LLM responses
+- Store all responses in the database
+- Useful when you need the latest analysis or want to refresh cached data
 
 ## Viewing Analytics
 
@@ -195,16 +198,16 @@ docker-compose down -v
 
 ## Toggling Between Modes
 
-You can easily switch between modes by changing the `USE_DATABASE` environment variable:
+You can easily switch between modes by changing the `USE_CACHE` environment variable:
 
-### Quick Toggle (temporary, for testing)
+### Quick Toggle (temporary)
 
 ```bash
-# Run in database mode
-USE_DATABASE=true python main.py
+# Run with caching enabled
+USE_CACHE=true python main.py
 
-# Run in API mode
-USE_DATABASE=false python main.py
+# Run with fresh API calls
+USE_CACHE=false python main.py
 ```
 
 ### Persistent Toggle
@@ -212,38 +215,48 @@ USE_DATABASE=false python main.py
 Edit your `.env` file:
 
 ```env
-# For testing/analytics
-USE_DATABASE=true
+# Use cached responses when available (recommended for most use)
+USE_CACHE=true
 
-# For production use
-USE_DATABASE=false
+# Always make fresh API calls
+USE_CACHE=false
 ```
 
 ## Use Cases
 
-### Testing Without API Costs
+### Cost Optimization with Caching
 
-During development, enable database mode to test the full application flow without consuming API credits:
+Cache mode reduces API costs by reusing previous analysis results:
 
 ```bash
-USE_DATABASE=true python main.py
+USE_CACHE=true python main.py
+```
+
+If the same company/patent combination was analyzed before, the cached response is returned instantly.
+
+### Fresh Analysis
+
+When you need the latest LLM analysis (e.g., after model updates):
+
+```bash
+USE_CACHE=false python main.py
 ```
 
 ### Collecting Usage Analytics
 
-Enable database mode in a test environment to collect analytics on:
+The database stores all interactions, enabling analytics on:
 - Which companies are analyzed most frequently
 - Types of analyses performed
-- Prompt patterns and lengths
-- Usage over time
+- Token usage and costs over time
+- Response caching hit rates
 
 ### Development and Debugging
 
-Database mode is useful for:
-- Testing patent parsing logic without API calls
+Database storage is useful for:
+- Reviewing actual prompts sent to the LLM
+- Analyzing response patterns
 - Debugging the full pipeline end-to-end
-- Collecting sample prompts for optimization
-- Understanding token usage patterns (when in API mode with logging)
+- Understanding token usage patterns
 
 ## Troubleshooting
 

docs/DEPLOYMENT.md

@@ -64,7 +64,7 @@ docker-compose ps
 # You should see:
 # - sparc-postgres (healthy)
 # - sparc-api (running on port 8000)
-# - sparc-dashboard (running on port 8501)
+# - sparc-dashboard (running on port 8080)
 ```
 
 The database is automatically initialized by the `init-db` service.
@@ -116,11 +116,13 @@ docker-compose up -d postgres
 # Wait for database to be healthy, then initialize
 python scripts/init_database.py
 
-# Terminal 1: Start FastAPI backend
+# Start FastAPI backend
 uvicorn SPARC.api:app --host 0.0.0.0 --port 8000 --reload
 
-# Terminal 2: Start Streamlit dashboard
-streamlit run dashboard.py --server.port 8501 --server.address 0.0.0.0
+# For the React frontend (separate terminal)
+cd frontend
+npm install
+npm run dev
 ```
 
 ---
@@ -141,7 +143,7 @@ Access the services:
 |---------|-----|
 | REST API | http://localhost:8000 |
 | API Documentation (Swagger) | http://localhost:8000/docs |
-| Dashboard (Web UI) | http://localhost:8501 |
+| Dashboard (Web UI) | http://localhost:8080 |
 
 ---
 
@@ -149,16 +151,17 @@ Access the services:
 
 ### Via Dashboard (Web UI)
 
-1. Open http://localhost:8501
-2. Select **"Company Analysis"** from the sidebar
-3. Enter a company name (e.g., "Intel")
-4. Click **"Analyze"**
+1. Open http://localhost:8080
+2. Register a new account or login (default admin: `admin` / `admin`)
+3. Navigate to **"Analysis"** from the sidebar
+4. Enter a company name (e.g., "Intel")
+5. Click **"Analyze"**
 
 This will:
 - Query SerpAPI for recent patents
 - Download and parse patent PDFs
 - Send patent content to Claude for analysis
-- Store prompt/response in PostgreSQL
+- Store prompt/response in PostgreSQL (with caching)
 - Display results in the dashboard
 
 ### Via REST API
@@ -233,12 +236,12 @@ docker exec -it sparc-postgres psql -U postgres -d sparc -c \
 
 | Component | Purpose |
 |-----------|---------|
-| **Dashboard** | Streamlit web UI for interactive analysis |
-| **FastAPI** | REST API for programmatic access |
+| **Dashboard** | React TypeScript web UI with authentication |
+| **FastAPI** | REST API with JWT authentication |
 | **Analyzer** | Orchestrates patent retrieval and LLM analysis |
 | **SerpAPI** | Retrieves patent data from Google Patents |
 | **OpenRouter** | Routes requests to Claude for AI analysis |
-| **PostgreSQL** | Stores prompts, responses, and analytics |
+| **PostgreSQL** | Stores prompts, responses, users, and cached results |
 
 ---
 
@@ -248,10 +251,9 @@ docker exec -it sparc-postgres psql -U postgres -d sparc -c \
 |----------|----------|---------|-------------|
 | `API_KEY` | Yes | - | SerpAPI key for patent search |
 | `OPENROUTER_API_KEY` | Yes | - | OpenRouter API key for Claude access |
-| `DATABASE_URL` | Yes* | - | PostgreSQL connection string |
-| `USE_DATABASE` | No | `false` | Set to `true` to enable database storage |
-
-*Required when `USE_DATABASE=true`
+| `DATABASE_URL` | Yes | - | PostgreSQL connection string |
+| `USE_CACHE` | No | `true` | Check database for cached responses before API calls |
+| `JWT_SECRET` | Yes | - | Secret key for JWT authentication (change in production!) |
 
 ### Database URL Format
 
@@ -273,9 +275,9 @@ The `docker-compose.yml` includes all services needed for production:
 | Service | Container | Port | Description |
 |---------|-----------|------|-------------|
 | `postgres` | sparc-postgres | 5432 | PostgreSQL database |
-| `init-db` | sparc-init-db | - | One-time database initialization |
-| `api` | sparc-api | 8000 | FastAPI REST API |
-| `dashboard` | sparc-dashboard | 8501 | Streamlit web UI |
+| `init-db` | sparc-init-db | - | One-time database initialization (seeds admin user) |
+| `api` | sparc-api | 8000 | FastAPI REST API with JWT auth |
+| `dashboard` | sparc-dashboard | 8080 | React TypeScript web UI |
 
 ### Common Docker Compose Commands
 
@@ -382,11 +384,11 @@ cp .env.example .env
 docker-compose up -d postgres
 python scripts/init_database.py
 uvicorn SPARC.api:app --reload &
-streamlit run dashboard.py
+cd frontend && npm install && npm run dev &
 
 # Check status
 curl http://localhost:8000/health
-open http://localhost:8501
+open http://localhost:8080
 
 # View data
 python scripts/view_analytics.py

frontend/Dockerfile

@@ -21,8 +21,11 @@ FROM nginx:alpine
 # Copy built files
 COPY --from=build /app/dist /usr/share/nginx/html
 
-# Copy nginx config
-COPY nginx.conf /etc/nginx/conf.d/default.conf
+# Copy nginx template (processed at startup with envsubst)
+COPY nginx.conf.template /etc/nginx/templates/default.conf.template
+
+# Default API URL (override with -e API_URL=...)
+ENV API_URL=http://api:8000/
 
 EXPOSE 80
 

frontend/nginx.conf.template

@@ -15,7 +15,7 @@ server {
 
     # Proxy API requests to backend
     location /api/ {
-        proxy_pass http://api:8000/;
+        proxy_pass ${API_URL}/;
         proxy_http_version 1.1;
         proxy_set_header Upgrade $http_upgrade;
         proxy_set_header Connection 'upgrade';

tests/test_analyzer.py

@@ -1,11 +1,22 @@
 """Tests for the high-level company analyzer orchestration."""
 
 import pytest
-from unittest.mock import Mock, patch, call
+from unittest.mock import Mock, patch, call, MagicMock
 from SPARC.analyzer import CompanyAnalyzer
 from SPARC.types import Patent, Patents, CompanyAnalysisResult, BatchAnalysisResult
 
 
+@pytest.fixture(autouse=True)
+def mock_db(mocker):
+    """Mock DatabaseClient for all tests so no real DB connection is needed."""
+    mock_db_cls = mocker.patch("SPARC.analyzer.DatabaseClient")
+    mock_db_instance = MagicMock()
+    mock_db_instance.get_cached_patent.return_value = None
+    mock_db_instance.get_cached_serp_query.return_value = None
+    mock_db_cls.return_value = mock_db_instance
+    return mock_db_instance
+
+
 class TestCompanyAnalyzer:
     """Test the CompanyAnalyzer orchestration logic."""
 
@@ -17,7 +28,7 @@ class TestCompanyAnalyzer:
 
         mock_llm.assert_called_once_with(api_key="test-key")
 
-    def test_analyze_company_full_pipeline(self, mocker):
+    def test_analyze_company_full_pipeline(self, mocker, mock_db):
         """Test complete company analysis pipeline."""
         # Mock all the dependencies
         mock_query = mocker.patch("SPARC.analyzer.SERP.query")
@@ -178,6 +189,180 @@ class TestCompanyAnalyzer:
         assert "PDF not found" in result
 
 
+class TestSingleQueryBugFix:
+    """Test that SERP.query is only called once per company analysis."""
+
+    def test_analyze_company_safe_calls_query_once(self, mocker, mock_db):
+        """_analyze_company_safe should call SERP.query exactly once."""
+        mock_query = mocker.patch("SPARC.analyzer.SERP.query")
+        mock_save = mocker.patch("SPARC.analyzer.SERP.save_patents")
+        mock_parse = mocker.patch("SPARC.analyzer.SERP.parse_patent_pdf")
+        mock_minimize = mocker.patch("SPARC.analyzer.SERP.minimize_patent_for_llm")
+        mock_llm = mocker.patch("SPARC.analyzer.LLMAnalyzer")
+
+        patent = Patent(patent_id="US123", pdf_link="http://example.com/test.pdf")
+        mock_query.return_value = Patents(patents=[patent])
+
+        def save_side_effect(p):
+            p.pdf_path = "patents/US123.pdf"
+            return p
+
+        mock_save.side_effect = save_side_effect
+        mock_parse.return_value = {"abstract": "Test"}
+        mock_minimize.return_value = "Content"
+
+        mock_llm_instance = Mock()
+        mock_llm_instance.analyze_patent_portfolio.return_value = "Analysis"
+        mock_llm.return_value = mock_llm_instance
+
+        analyzer = CompanyAnalyzer()
+        analyzer._analyze_company_safe("TestCorp")
+
+        # The key assertion: SERP.query called exactly once, not twice
+        mock_query.assert_called_once_with("TestCorp")
+
+    def test_analyze_company_with_prefetched_patents_skips_query(self, mocker):
+        """analyze_company should not call SERP.query when patents are provided."""
+        mock_query = mocker.patch("SPARC.analyzer.SERP.query")
+        mock_save = mocker.patch("SPARC.analyzer.SERP.save_patents")
+        mock_parse = mocker.patch("SPARC.analyzer.SERP.parse_patent_pdf")
+        mock_minimize = mocker.patch("SPARC.analyzer.SERP.minimize_patent_for_llm")
+        mock_llm = mocker.patch("SPARC.analyzer.LLMAnalyzer")
+
+        patent = Patent(patent_id="US123", pdf_link="http://example.com/test.pdf")
+        prefetched = Patents(patents=[patent])
+
+        def save_side_effect(p):
+            p.pdf_path = "patents/US123.pdf"
+            return p
+
+        mock_save.side_effect = save_side_effect
+        mock_parse.return_value = {"abstract": "Test"}
+        mock_minimize.return_value = "Content"
+
+        mock_llm_instance = Mock()
+        mock_llm_instance.analyze_patent_portfolio.return_value = "Analysis"
+        mock_llm.return_value = mock_llm_instance
+
+        analyzer = CompanyAnalyzer()
+        analyzer.analyze_company("TestCorp", patents=prefetched)
+
+        # SERP.query should never be called
+        mock_query.assert_not_called()
+
+
+class TestPatentCaching:
+    """Test patent-level DB caching in the pipeline."""
+
+    def test_process_single_patent_uses_db_cache(self, mocker, mock_db):
+        """_process_single_patent returns cached content when available."""
+        mock_save = mocker.patch("SPARC.analyzer.SERP.save_patents")
+
+        mock_db.get_cached_patent.return_value = {
+            "patent_id": "US123",
+            "minimized_content": "Cached minimized content",
+        }
+
+        patent = Patent(patent_id="US123", pdf_link="http://example.com/test.pdf")
+        result = CompanyAnalyzer._process_single_patent(patent, "TestCorp", mock_db)
+
+        assert result == {"patent_id": "US123", "content": "Cached minimized content"}
+        # Should NOT download since cache hit
+        mock_save.assert_not_called()
+
+    def test_process_single_patent_stores_to_db_cache(self, mocker, mock_db):
+        """_process_single_patent stores result in DB after processing."""
+        mock_save = mocker.patch("SPARC.analyzer.SERP.save_patents")
+        mock_parse = mocker.patch("SPARC.analyzer.SERP.parse_patent_pdf")
+        mock_minimize = mocker.patch("SPARC.analyzer.SERP.minimize_patent_for_llm")
+
+        # No cache hit
+        mock_db.get_cached_patent.return_value = None
+
+        patent = Patent(patent_id="US123", pdf_link="http://example.com/test.pdf")
+
+        def save_side_effect(p):
+            p.pdf_path = "patents/US123.pdf"
+            return p
+
+        mock_save.side_effect = save_side_effect
+        mock_parse.return_value = {"abstract": "Test abstract"}
+        mock_minimize.return_value = "Minimized content"
+
+        result = CompanyAnalyzer._process_single_patent(patent, "TestCorp", mock_db)
+
+        assert result == {"patent_id": "US123", "content": "Minimized content"}
+        mock_db.store_patent.assert_called_once_with(
+            patent_id="US123",
+            company_name="TestCorp",
+            pdf_link="http://example.com/test.pdf",
+            raw_sections={"abstract": "Test abstract"},
+            minimized_content="Minimized content",
+        )
+
+    def test_serp_query_cache_hit_skips_api(self, mocker, mock_db):
+        """When SERP query is cached, API call is skipped."""
+        mock_query = mocker.patch("SPARC.analyzer.SERP.query")
+        mock_save = mocker.patch("SPARC.analyzer.SERP.save_patents")
+        mock_parse = mocker.patch("SPARC.analyzer.SERP.parse_patent_pdf")
+        mock_minimize = mocker.patch("SPARC.analyzer.SERP.minimize_patent_for_llm")
+        mock_llm = mocker.patch("SPARC.analyzer.LLMAnalyzer")
+
+        # Simulate SERP cache hit
+        mock_db.get_cached_serp_query.return_value = ["US123"]
+        # Simulate patent cache hit too
+        mock_db.get_cached_patent.return_value = {
+            "patent_id": "US123",
+            "minimized_content": "Cached content",
+        }
+
+        mock_llm_instance = Mock()
+        mock_llm_instance.analyze_patent_portfolio.return_value = "Analysis"
+        mock_llm.return_value = mock_llm_instance
+
+        analyzer = CompanyAnalyzer()
+        result = analyzer.analyze_company("TestCorp")
+
+        assert result == "Analysis"
+        # SERP.query should NOT be called
+        mock_query.assert_not_called()
+        # No downloads should happen
+        mock_save.assert_not_called()
+
+    def test_serp_query_cache_miss_stores_result(self, mocker, mock_db):
+        """When SERP query cache misses, result is stored after API call."""
+        mock_query = mocker.patch("SPARC.analyzer.SERP.query")
+        mock_save = mocker.patch("SPARC.analyzer.SERP.save_patents")
+        mock_parse = mocker.patch("SPARC.analyzer.SERP.parse_patent_pdf")
+        mock_minimize = mocker.patch("SPARC.analyzer.SERP.minimize_patent_for_llm")
+        mock_llm = mocker.patch("SPARC.analyzer.LLMAnalyzer")
+
+        mock_db.get_cached_serp_query.return_value = None
+
+        patent = Patent(patent_id="US123", pdf_link="http://example.com/test.pdf")
+        mock_query.return_value = Patents(patents=[patent])
+
+        def save_side_effect(p):
+            p.pdf_path = "patents/US123.pdf"
+            return p
+
+        mock_save.side_effect = save_side_effect
+        mock_parse.return_value = {"abstract": "Test"}
+        mock_minimize.return_value = "Content"
+
+        mock_llm_instance = Mock()
+        mock_llm_instance.analyze_patent_portfolio.return_value = "Analysis"
+        mock_llm.return_value = mock_llm_instance
+
+        analyzer = CompanyAnalyzer()
+        analyzer.analyze_company("TestCorp")
+
+        mock_db.store_serp_query.assert_called_once()
+        call_kwargs = mock_db.store_serp_query.call_args[1]
+        assert call_kwargs["company_name"] == "TestCorp"
+        assert call_kwargs["patent_ids"] == ["US123"]
+
+
 class TestBatchProcessing:
     """Test multi-company batch processing functionality."""
 
@@ -316,7 +501,7 @@ class TestBatchProcessing:
 
         assert callback.call_count == 2
 
-    def test_company_analysis_result_structure(self, mocker):
+    def test_company_analysis_result_structure(self, mocker, mock_db):
         """Test CompanyAnalysisResult has correct structure."""
         mock_query = mocker.patch("SPARC.analyzer.SERP.query")
         mock_save = mocker.patch("SPARC.analyzer.SERP.save_patents")
@@ -327,6 +512,9 @@ class TestBatchProcessing:
         patent = Patent(patent_id="US123", pdf_link="http://example.com/test.pdf")
         mock_query.return_value = Patents(patents=[patent])
 
+        # Simulate DB caching: after store, subsequent get returns the IDs
+        mock_db.get_cached_serp_query.side_effect = [None, ["US123"]]
+
         def save_side_effect(p):
             p.pdf_path = "patents/US123.pdf"
             return p

tests/test_serp_api.py

@@ -1,7 +1,11 @@
 """Tests for SERP API patent retrieval and parsing functionality."""
 
+import os
 import pytest
+from unittest.mock import patch, Mock
+from datetime import datetime, timedelta
 from SPARC.serp_api import SERP
+from SPARC.types import Patent
 
 
 class TestTextCleaning:
@@ -176,3 +180,89 @@ class TestPatentMinimization:
 
         # Sections should be separated by double newlines
         assert "\n\n" in result
+
+
+class TestDynamicDateRange:
+    """Test dynamic date range computation in SERP.query."""
+
+    def test_query_uses_rolling_date_window(self, mocker):
+        """Verify the date filter uses a rolling window, not hardcoded dates."""
+        mock_search = mocker.patch("SPARC.serp_api.serpapi.search")
+        mock_search.return_value = {"organic_results": []}
+        mocker.patch("SPARC.serp_api.config.api_key", "fake-key")
+        mocker.patch("SPARC.serp_api.config.patent_search_days", 90)
+
+        SERP.query("TestCorp")
+
+        call_params = mock_search.call_args[0][0]
+        tbs = call_params["tbs"]
+        # Should contain "cdr:1,cd_min:" with a date, not the old hardcoded one
+        assert "cdr:1,cd_min:" in tbs
+        assert "10/28/2025" not in tbs  # old hardcoded date gone
+
+    def test_query_respects_days_back_param(self, mocker):
+        """Verify days_back parameter controls the date window."""
+        mock_search = mocker.patch("SPARC.serp_api.serpapi.search")
+        mock_search.return_value = {"organic_results": []}
+        mocker.patch("SPARC.serp_api.config.api_key", "fake-key")
+        mocker.patch("SPARC.serp_api.config.patent_search_days", 90)
+
+        now = datetime.now()
+        SERP.query("TestCorp", days_back=30)
+
+        call_params = mock_search.call_args[0][0]
+        tbs = call_params["tbs"]
+        expected_start = (now - timedelta(days=30)).strftime("%-m/%-d/%Y")
+        assert expected_start in tbs
+
+
+class TestFilesystemPDFCaching:
+    """Test that save_patents skips download for existing files."""
+
+    def test_save_patents_skips_download_when_cached(self, mocker, tmp_path):
+        """Already-downloaded PDFs should not be re-downloaded."""
+        mock_get = mocker.patch("SPARC.serp_api.requests.get")
+        mocker.patch("SPARC.serp_api.os.makedirs")
+
+        pdf_path = tmp_path / "US123.pdf"
+        pdf_path.write_bytes(b"%PDF-1.4 fake content")
+
+        mocker.patch("SPARC.serp_api.os.path.exists", return_value=True)
+        mocker.patch("SPARC.serp_api.os.path.getsize", return_value=100)
+
+        patent = Patent(patent_id="US123", pdf_link="http://example.com/test.pdf")
+        result = SERP.save_patents(patent)
+
+        mock_get.assert_not_called()
+        assert result.pdf_path == "patents/US123.pdf"
+
+    def test_save_patents_downloads_when_not_cached(self, mocker):
+        """Missing PDFs should be downloaded."""
+        mock_response = Mock()
+        mock_response.content = b"%PDF-1.4 content"
+        mock_get = mocker.patch("SPARC.serp_api.requests.get", return_value=mock_response)
+        mocker.patch("SPARC.serp_api.os.makedirs")
+        mocker.patch("SPARC.serp_api.os.path.exists", return_value=False)
+        mock_open = mocker.patch("builtins.open", mocker.mock_open())
+
+        patent = Patent(patent_id="US456", pdf_link="http://example.com/test.pdf")
+        result = SERP.save_patents(patent)
+
+        mock_get.assert_called_once_with("http://example.com/test.pdf")
+        assert result.pdf_path == "patents/US456.pdf"
+
+    def test_save_patents_redownloads_empty_files(self, mocker):
+        """Empty/corrupt PDFs (0 bytes) should be re-downloaded."""
+        mock_response = Mock()
+        mock_response.content = b"%PDF-1.4 content"
+        mock_get = mocker.patch("SPARC.serp_api.requests.get", return_value=mock_response)
+        mocker.patch("SPARC.serp_api.os.makedirs")
+        mocker.patch("SPARC.serp_api.os.path.exists", return_value=True)
+        mocker.patch("SPARC.serp_api.os.path.getsize", return_value=0)
+        mock_open = mocker.patch("builtins.open", mocker.mock_open())
+
+        patent = Patent(patent_id="US789", pdf_link="http://example.com/test.pdf")
+        result = SERP.save_patents(patent)
+
+        mock_get.assert_called_once()
+        assert result.pdf_path == "patents/US789.pdf"