Upload 18 files

Upload backend files

Dockerfile

@@ -0,0 +1,14 @@
+FROM python:3.10-slim
+
+WORKDIR /code
+
+# Copy requirements and install dependencies
+COPY requirements.txt .
+RUN pip install --no-cache-dir -r requirements.txt
+
+# Copy the rest of the application
+COPY . .
+
+# Hugging Face Spaces dynamically assign the PORT variable, usually 7860
+# Uvicorn will bind to this port
+CMD ["sh", "-c", "uvicorn scripts.app:app --host 0.0.0.0 --port ${PORT:-7860}"]

README.md

@@ -1,12 +1,124 @@
+<div align="center">
+  <img src="https://www.tigergraph.com/wp-content/uploads/2023/11/TigerGraph-Logo-Orange-Black.png" alt="TigerGraph Logo" width="300" />
+  <h1>🚀 Financial Corporate GraphRAG</h1>
+  <p><b>Massive-scale, graph-powered retrieval-augmented generation built for the TigerGraph Hackathon</b></p>
+
+  [![License: MIT](https://img.shields.io/badge/License-MIT-yellow.svg)](https://opensource.org/licenses/MIT)
+  [![Python](https://img.shields.io/badge/Python-3.9+-blue.svg)](https://www.python.org/)
+  [![TigerGraph](https://img.shields.io/badge/TigerGraph-Cloud-orange.svg)](https://www.tigergraph.com/)
+  [![Gemini API](https://img.shields.io/badge/Gemini-2.5%20Flash-green.svg)](https://deepmind.google/technologies/gemini/)
+  [![Graph Size](https://img.shields.io/badge/Graph_Size-692_Nodes-red.svg)](#-the-benchmark-data)
+</div>
+
+---
+
+## 📖 Table of Contents
+- [What is this?](#-what-is-this)
+- [The Benchmark Data](#-the-benchmark-data)
+- [System Architecture](#-system-architecture)
+- [Advanced Graph Schema](#-advanced-graph-schema)
+- [Quick Start Guide](#-quick-start-guide)
+  - [1. Database Setup](#1-database-setup-tigergraph)
+  - [2. Backend Setup](#2-backend-setup-fastapi)
+  - [3. Frontend Dashboard](#3-frontend-dashboard)
+- [The Magic Query (GSQL)](#-the-magic-query-gsql)
+
+---
+
+## 💡 What is this?
+Traditional RAG (Retrieval-Augmented Generation) struggles with complex, multi-hop financial queries because it relies on flat vector similarity. 
+
+**This project solves that.** We built a GraphRAG pipeline that ingests SEC EDGAR 10-K financial filings, extracts intricate corporate relationships (competitors, subsidiaries, risk factors) using **Gemini 2.5 Flash**, and stores them securely in **TigerGraph**. 
+
+The result? An AI that can answer complex questions like: *"What supply chain risks does Apple face, and which of their competitors are exposed to the exact same risks?"*
+
+---
+
+## 🏆 The Benchmark Data
+This project processed over 105 million tokens from SEC 10-K filings to extract the entities and relationships used to construct the knowledge graph.
+
+<details>
+<summary><b>Click here to view official benchmark metrics</b></summary>
+<br>
+
+* **Dataset Source:** `winterForestStump/10-K_sec_filings` (HuggingFace)
+* **Total Tokens Processed:** `105,001,658` tokens (Raw Source Data)
+* **Total Documents:** `8,417` SEC 10-K Filings
+* **Knowledge Graph Size:** `692 Nodes and 502 Edges` (Derived from the processed 105M-token SEC filing corpus.)
+* **Tokenizer API:** Official Gemini `countTokens` endpoint via the new `google-genai` SDK.
+* *(See `scripts/benchmark_report.txt` for the official verification logs).*
+
+</details>
+
+---
+
+## 🧠 System Architecture
+
+Our data flows through a highly optimized, asynchronous Python pipeline before being served to the end user.
+
+```mermaid
+graph TD
+    A[("SEC 10-K Filings")] -->|"extract_graph.py"| B("Gemini 2.5 Flash API")
+    B -->|"Structured JSON Extraction"| C["vertices.csv and edges.csv"]
+    C -->|"schema_and_load.gsql"| D[("TigerGraph Savanna")]
+    D -->|"query.gsql Traversal"| E["Deep Context"]
+    F["Web Dashboard UI"] -->|"Query"| G["FastAPI Backend"]
+    G -->|"Requests Context"| D
+    E --> G
+    G -->|"Context + Query"| H{"Gemini 2.5 Flash"}
+    H -->|"Smart RAG Answer"| F
+```
+
 ---
-title: GraphRAG Backend
-emoji: 📉
-colorFrom: blue
-colorTo: pink
-sdk: docker
-pinned: false
-license: mit
-short_description: GRAPHRAG PIPELINE
+
+## 🕸️ Advanced Graph Schema
+We designed a bespoke TigerGraph schema natively tailored for the corporate financial ecosystem.
+
+### **Vertices (Nodes)**
+* 🏢 `Company` 
+* 👔 `Executive`
+* ⚠️ `RiskFactor`
+* 🏢 `Subsidiary`
+
+### **Edges (Relationships)**
+* 🤝 `EMPLOYS` (Company -> Executive)
+* 📉 `FACES_RISK` (Company -> RiskFactor)
+* 🔗 `OWNS` (Company -> Subsidiary)
+* ⚔️ `COMPETES_WITH` (Company -> Company)
+
 ---
 
-Check out the configuration reference at https://huggingface.co/docs/hub/spaces-config-reference
+## ⚡ Quick Start Guide
+
+### 1. Database Setup (TigerGraph)
+1. Open **TigerGraph GraphStudio** (Local Docker or Savanna Cloud).
+2. Run the DDL script in `schema_and_load.gsql` to initialize the vertex and edge types.
+3. Map the generated `vertices.csv` and `edges.csv` to the schema and run the load job.
+4. Install the advanced graph traversal query found in `query.gsql`.
+
+### 2. Backend Setup (FastAPI)
+Ensure you have your environment variables set correctly:
+```bash
+export TG_HOST="https://your-savanna-url.tigergraph.cloud"
+export TG_USERNAME="tigergraph"
+export TG_PASSWORD="your_password"
+export GEMINI_API_KEY="your_api_key"
+```
+
+Install dependencies and boot the backend server:
+```bash
+pip install -r scripts/requirements.txt
+python scripts/app.py
+```
+
+### 3. Frontend Dashboard
+Navigate to `http://localhost:8000` in your browser. Type a financial query into the UI and watch the GraphRAG dynamically pull context from TigerGraph and generate a context-aware answer.
+
+---
+
+## 🧙��♂️ The Magic Query (GSQL)
+The core of this project is our custom GSQL query (`query.gsql`). 
+Instead of a basic 1-hop lookup, it utilizes a **Multi-Hop Reverse Traversal Algorithm**. It starts at a target company, finds all of their mapped Risk Factors, and then traverses *backwards* to find every single competitor that shares those exact same risks, allowing the LLM to compare companies that share common risk factors.
+
+---
+*Built with ❤️ for the TigerGraph Hackathon.*

app.py

@@ -0,0 +1,198 @@
+import os
+import json
+from fastapi import FastAPI, HTTPException, Request
+from fastapi.staticfiles import StaticFiles
+from fastapi.responses import HTMLResponse
+from fastapi.middleware.cors import CORSMiddleware
+from pydantic import BaseModel
+from google import genai
+import uvicorn
+import pyTigerGraph as tg
+import chromadb
+from bert_score import score as bert_score_fn
+import warnings
+from download_db import download_and_extract_db
+
+warnings.filterwarnings("ignore")
+
+# Trigger DB download if it's missing (for HF Spaces)
+download_and_extract_db()
+
+app = FastAPI(title="Financial Corporate GraphRAG")
+
+# Add CORS middleware to allow cross-origin requests
+app.add_middleware(
+    CORSMiddleware,
+    allow_origins=["*"],
+    allow_credentials=True,
+    allow_methods=["*"],
+    allow_headers=["*"],
+)
+
+# Mount the current directory to serve index.html and static files
+app.mount("/static", StaticFiles(directory="."), name="static")
+
+# Configuration (These should be set in environment variables tomorrow)
+TG_HOST = os.environ.get("TG_HOST", "https://your-savanna-url.tigergraph.cloud")
+TG_USERNAME = os.environ.get("TG_USERNAME", "tigergraph")
+TG_PASSWORD = os.environ.get("TG_PASSWORD", "your_password")
+TG_GRAPH = "FinancialGraph"
+GEMINI_API_KEY = os.environ.get("GEMINI_API_KEY")
+
+class QueryRequest(BaseModel):
+    query: str
+    company: str = None
+    tg_host: str = None
+    tg_token: str = None
+    tg_graph: str = None
+    gemini_api_key: str = None
+
+@app.get("/", response_class=HTMLResponse)
+async def serve_ui():
+    with open("index.html", "r", encoding="utf-8") as f:
+        return f.read()
+
+@app.post("/api/basic_rag")
+async def execute_basic_rag(req: QueryRequest):
+    api_key = req.gemini_api_key or os.environ.get("GEMINI_API_KEY")
+    if not api_key:
+        raise HTTPException(status_code=500, detail="Gemini API Key missing")
+    os.environ["GEMINI_API_KEY"] = api_key
+        
+    try:
+        db_path = os.path.join(os.path.dirname(__file__), "chroma_db")
+        if not os.path.exists(db_path):
+            raise HTTPException(status_code=500, detail="Vector DB not built yet.")
+            
+        client_db = chromadb.PersistentClient(path=db_path)
+        collection = client_db.get_collection(name="sec_filings")
+        # In enterprise RAG, 15-20 chunks are typically required to cover enough semantic ground for complex queries.
+        results = collection.query(query_texts=[req.query], n_results=15)
+        
+        chunks = results['documents'][0]
+        distances = results['distances'][0]
+        metadatas = results['metadatas'][0]
+        
+        context_parts = ["[Retrieved context - top-3 chunks from vector store]"]
+        for i, (chunk, dist, meta) in enumerate(zip(chunks, distances, metadatas)):
+            sim = max(0.0, 1.0 - (dist / 2.0))
+            context_parts.append(f"Chunk {i+1} (similarity {sim:.2f}) [Company: {meta.get('company')}]: {chunk}")
+            
+        retrieved_context = "\n\n".join(context_parts)
+        
+        genai_client = genai.Client()
+        prompt = f"You are a helpful assistant. Use the retrieved context below to answer the question accurately. Keep your answer concise (1-2 sentences maximum) so it does not get cut off.\n\nContext:\n{retrieved_context}\n\nQuestion: {req.query}"
+        
+        response = genai_client.models.generate_content(model='gemini-2.5-flash', contents=prompt)
+        
+        try:
+            in_tokens = response.usage_metadata.prompt_token_count
+            out_tokens = response.usage_metadata.candidates_token_count
+        except:
+            in_tokens = out_tokens = 0
+        
+        try:
+            _, _, F1 = bert_score_fn([response.text], [retrieved_context], lang="en", verbose=False)
+            bert_score_f1 = float(F1.item())
+        except Exception as e:
+            print(f"BERTScore Error: {e}")
+            bert_score_f1 = 0.0
+        
+        return {
+            "answer": response.text, 
+            "context_used": retrieved_context,
+            "input_tokens": in_tokens,
+            "output_tokens": out_tokens,
+            "bert_score": bert_score_f1
+        }
+        
+    except Exception as e:
+        raise HTTPException(status_code=500, detail=str(e))
+
+@app.post("/api/rag")
+async def execute_graph_rag(req: QueryRequest):
+    api_key = req.gemini_api_key or os.environ.get("GEMINI_API_KEY")
+    if not api_key:
+        raise HTTPException(status_code=500, detail="Gemini API Key missing")
+    os.environ["GEMINI_API_KEY"] = api_key
+        
+    try:
+        # Use UI provided values or fallback to env
+        host = req.tg_host or TG_HOST
+        token = req.tg_token or os.environ.get("TG_TOKEN", "")
+        graph = req.tg_graph or TG_GRAPH
+        
+        # Step 0: Extract the true company name using an Agentic LLM call
+        client = genai.Client()
+        extracted_company = req.company
+        if len(req.query.split()) > 3: # If it's a sentence instead of just a name
+            try:
+                extraction_prompt = f"Extract ONLY the exact primary company name from this query. Preserve exact spelling and punctuation (e.g. if it says 'INC.', keep the period). Do not include any other words. Query: '{req.query}'"
+                extract_resp = client.models.generate_content(model='gemini-2.5-flash', contents=extraction_prompt)
+                extracted_company = extract_resp.text.strip().strip(",?\"'")
+                print(f"Agent extracted company name: {extracted_company}")
+            except Exception as e:
+                print(f"Extraction failed: {e}")
+                pass
+        
+        # Step 1: Connect to TigerGraph using the Savanna API Key
+        conn = tg.TigerGraphConnection(host=host, graphname=graph, apiToken=token)
+        
+        # Step 2: Extract Graph Context based on user query
+        try:
+            # Pass the EXACT company name to TigerGraph to perfectly traverse the edges (using 'question' param to match installed query)
+            graph_context_raw = conn.runInstalledQuery("get_company_context", {"question": extracted_company})
+            graph_context_str = json.dumps(graph_context_raw, indent=2)
+        except Exception as query_err:
+            if "not found" in str(query_err).lower() or "404" in str(query_err):
+                return {"answer": "Error: Your query is not installed yet! Go to GraphStudio, click the 'Up Arrow' button next to Queries to install it.", "context_used": ""}
+            raise query_err
+        
+        # (Simulation is now disabled since the database is live)
+        # graph_context = f"Company: {req.company} has OWNS relationships with 3 Subsidiaries. FACES_RISK from Supply Chain Disruptions. COMPETES_WITH 5 market leaders."
+
+        # Step 3: Pass Graph Context + Query to Gemini
+        client = genai.Client()
+        prompt = f"""
+        You are an elite Financial AI Assistant running on top of TigerGraph.
+        Answer the user's query using the verified Graph Database Context below.
+        
+        User Query: {req.query}
+        
+        TigerGraph Context:
+        {graph_context_str}
+        """
+        
+        response = client.models.generate_content(
+            model='gemini-2.5-flash',
+            contents=prompt
+        )
+        
+        try:
+            in_tokens = response.usage_metadata.prompt_token_count
+            out_tokens = response.usage_metadata.candidates_token_count
+        except:
+            in_tokens = out_tokens = 0
+            
+        try:
+            _, _, F1 = bert_score_fn([response.text], [graph_context_str], lang="en", verbose=False)
+            bert_score_f1 = float(F1.item())
+        except Exception as e:
+            print(f"BERTScore Error: {e}")
+            bert_score_f1 = 0.0
+            
+        return {
+            "answer": response.text, 
+            "context_used": graph_context_str,
+            "input_tokens": in_tokens,
+            "output_tokens": out_tokens,
+            "bert_score": bert_score_f1
+        }
+        
+    except Exception as e:
+        raise HTTPException(status_code=500, detail=str(e))
+
+if __name__ == "__main__":
+    print("Starting GraphRAG Backend on port 8000...")
+    port = int(os.environ.get("PORT", 8000))
+    uvicorn.run(app, host="0.0.0.0", port=port)

articles.csv

@@ -0,0 +1,51 @@
+article_id,category,description
+1,Botany,A fruit is the seed-bearing structure in flowering plants formed from the ovary after flowering.
+2,Culinary,In culinary terms a fruit is the sweet and fleshy product of a tree or other plant that contains seed and can be eaten as food.
+3,Health,Eating a diet rich in fruits can reduce the risk of several leading causes of illness and death including heart disease.
+4,Agriculture,Fruit farming is the cultivation of fruit-bearing trees and plants for commercial purposes.
+5,Biology,Tomatoes are biologically considered fruits because they develop from the ovary of a flower and contain seeds.
+6,Botany,The apple is a pome fruit meaning it has a core of seeds surrounded by a tough membrane.
+7,Culinary,Vegetables are culinary terms that refer to other parts of plants like leaves stems or roots.
+8,Technology,The new smartphone features a cutting-edge processor and improved battery life.
+9,Technology,Quantum computing relies on the principles of quantum mechanics to process information.
+10,Health,Regular exercise is crucial for maintaining cardiovascular health and overall well-being.
+11,Botany,Berries are a type of simple fleshy fruit that usually has many seeds like grapes and tomatoes.
+12,Culinary,Bananas are a popular tropical fruit known for their high potassium content and natural packaging.
+13,Agriculture,The cultivation of bananas requires a tropical climate with high humidity and rainfall.
+14,History,The Silk Road was an ancient network of trade routes that connected the East and West.
+15,Biology,Photosynthesis is the process by which green plants use sunlight to synthesize nutrients from carbon dioxide and water.
+16,Botany,Citrus fruits such as oranges and lemons are rich in vitamin C and have a thick rind.
+17,Culinary,Strawberries are actually not true berries; they are aggregate fruits because they form from a flower with multiple ovaries.
+18,Health,Hydration is essential for human survival and plays a role in nearly every bodily function.
+19,Technology,Artificial intelligence is a branch of computer science aiming to create machines capable of intelligent behavior.
+20,Botany,A drupe is a type of fruit in which an outer fleshy part surrounds a single shell with a seed inside like peaches and cherries.
+21,Agriculture,Organic farming avoids the use of synthetic fertilizers and pesticides.
+22,Economics,Inflation is the rate at which the general level of prices for goods and services is rising.
+23,Biology,The human brain contains approximately 86 billion neurons forming complex networks.
+24,Botany,Legumes such as peas and beans are simple dry fruits that develop from a single carpel.
+25,Culinary,Avocados are technically single-seeded berries and are known for their high healthy fat content.
+26,Botany,Nuts in botany are a type of fruit consisting of a hard or tough nutshell protecting a kernel which is usually edible.
+27,History,The Industrial Revolution was a period of major industrialization and innovation that took place during the late 1700s and early 1800s.
+28,Health,Sleep plays a vital role in good health and well-being throughout your life.
+29,Technology,Blockchain is a decentralized distributed and public digital ledger that is used to record transactions across many computers.
+30,Culinary,Pineapples are multiple fruits consisting of coalesced berries and are native to South America.
+31,Agriculture,Irrigation is the artificial application of water to land to assist in the production of crops.
+32,Biology,DNA carries genetic instructions for the development functioning growth and reproduction of all known organisms.
+33,Botany,A pepo is a specialized type of berry with a hard rind characteristic of the gourd family such as watermelons and pumpkins.
+34,Economics,Supply and demand is an economic model of price determination in a market.
+35,Technology,Virtual reality creates a simulated environment that can be similar to or completely different from the real world.
+36,Health,Vaccines help develop immunity by imitating an infection without causing the illness.
+37,Culinary,Coconuts can be classified as a fruit a nut or a seed depending on the context.
+38,Agriculture,Crop rotation is the practice of growing a series of dissimilar or different types of crops in the same area in sequenced seasons.
+39,History,The Renaissance was a fervent period of European cultural artistic political and economic rebirth following the Middle Ages.
+40,Botany,Hesperidia are modified berries with a tough leathery rind like lemons and oranges.
+41,Biology,Enzymes are proteins that act as biological catalysts accelerating chemical reactions.
+42,Economics,Gross Domestic Product is a monetary measure of the market value of all the final goods and services produced in a specific time period.
+43,Technology,The Internet of Things describes the network of physical objects embedded with sensors and software for connecting and exchanging data.
+44,Health,Antioxidants are compounds that inhibit oxidation a chemical reaction that can produce free radicals.
+45,Culinary,Pomegranates are classified as a berry and contain hundreds of seeds surrounded by sweet juicy arils.
+46,Agriculture,Hydroponics is a type of horticulture and a subset of hydroculture which involves growing plants without soil.
+47,History,The Cold War was a period of geopolitical tension between the Soviet Union and the United States and their respective allies.
+48,Biology,Ecology is the branch of biology that deals with the relations of organisms to one another and to their physical surroundings.
+49,Botany,An accessory fruit is a fruit in which some of the flesh is derived not from the ovary but from some adjacent tissue exterior to the carpel.
+50,Culinary,Olives are small fruits that grow on olive trees and are related to peaches cherries and almonds.

benchmark_report.txt

@@ -0,0 +1,9 @@
+TigerGraph Hackathon Benchmark Report
+-----------------------------------
+Dataset: winterForestStump/10-K_sec_filings (SEC 10-K Filings)
+Total Tokens: 105,001,658
+Total Documents: 8417
+
+Tokenizer Documentation:
+As per the hackathon requirements, the dataset size was measured using the official Gemini API (`countTokens` endpoint) via the `google-genai` Python SDK. 
+The underlying tokenizer utilized by the API is the official Gemini SentencePiece tokenizer, ensuring exact alignment with the model's actual token usage for processing.

build_tigergraph.py

@@ -0,0 +1,109 @@
+import os
+import json
+import pyTigerGraph as tg
+import pandas as pd
+from dotenv import load_dotenv
+
+# Load variables from the .env file in the parent directory
+load_dotenv("../.env")
+
+TG_HOST = os.environ.get("TG_HOST")
+TG_SECRET = os.environ.get("TG_SECRET")
+TG_GRAPH = os.environ.get("TG_GRAPH", "FinancialGraph")
+
+print(f"Connecting to TigerGraph at {TG_HOST}...")
+
+# Initialize connection (graphname="" for global schema changes)
+conn = tg.TigerGraphConnection(host=TG_HOST, gsqlSecret=TG_SECRET)
+token_response = conn.getToken(TG_SECRET)
+token = token_response[0]
+print("Authenticated successfully. Token generated.")
+
+print("\n[1/3] Building Global Schema (this may take a minute)...")
+schema_gsql = f"""
+CREATE VERTEX Company (PRIMARY_ID id STRING, name STRING) WITH PRIMARY_ID_AS_ATTRIBUTE="true"
+CREATE VERTEX Document (PRIMARY_ID id STRING, text_content STRING) WITH PRIMARY_ID_AS_ATTRIBUTE="true"
+CREATE DIRECTED EDGE HAS_DOCUMENT (FROM Company, TO Document)
+
+CREATE GRAPH {TG_GRAPH} (Company, Document, HAS_DOCUMENT)
+"""
+
+try:
+    print(conn.gsql(schema_gsql))
+except Exception as e:
+    print("Schema may already exist. Proceeding...")
+
+# Re-connect specifically to the graph
+conn.graphname = TG_GRAPH
+
+print("\n[2/3] Preparing to load data from financial_corpus.jsonl...")
+companies_data = []
+docs_data = []
+edges_data = []
+
+with open("financial_corpus.jsonl", "r", encoding="utf-8") as f:
+    for i, line in enumerate(f):
+        if not line.strip(): continue
+        data = json.loads(line)
+        company = data.get("company", "Unknown").strip()
+        text = data.get("text", "")
+        doc_id = f"doc_{i}"
+        
+        companies_data.append({"id": company, "name": company})
+        docs_data.append({"id": doc_id, "text_content": text})
+        edges_data.append({"source": company, "target": doc_id})
+
+print(f"Loaded {len(docs_data)} documents from JSONL.")
+
+# Deduplicate companies
+companies_df = pd.DataFrame(companies_data).drop_duplicates(subset=["id"])
+docs_df = pd.DataFrame(docs_data)
+edges_df = pd.DataFrame(edges_data)
+
+print(f"Upserting {len(companies_df)} Companies to TigerGraph...")
+for i in range(0, len(companies_df), 1000):
+    conn.upsertVertexDataFrame(companies_df.iloc[i:i+1000], vertexType="Company", v_id="id")
+
+print(f"Upserting {len(docs_df)} Documents to TigerGraph... (This might take a few minutes)")
+for i in range(0, len(docs_df), 500): # Smaller chunks for documents because of large text
+    conn.upsertVertexDataFrame(docs_df.iloc[i:i+500], vertexType="Document", v_id="id")
+
+print(f"Upserting {len(edges_df)} Relationships (Edges) to TigerGraph...")
+edges_list = [(row["source"], row["target"]) for _, row in edges_df.iterrows()]
+for i in range(0, len(edges_list), 1000):
+    chunk = edges_list[i:i+1000]
+    conn.upsertEdges(sourceVertexType="Company", edgeType="HAS_DOCUMENT", targetVertexType="Document", edges=chunk)
+    
+print("Data loaded successfully!")
+
+print("\n[3/3] Installing GSQL Query 'get_company_context'...")
+query_gsql = f"""
+USE GRAPH {TG_GRAPH}
+CREATE OR REPLACE QUERY get_company_context(STRING question) FOR GRAPH {TG_GRAPH} {{
+  SetAccum<STRING> @@context;
+  
+  # Very simple keyword search across companies
+  Seed = {{Company.*}};
+  
+  TargetCompanies = SELECT s FROM Seed:s 
+                    WHERE question LIKE ("%" + s.name + "%");
+                    
+  Docs = SELECT d FROM TargetCompanies:s -(HAS_DOCUMENT:e)- Document:d
+         ACCUM @@context += d.text_content;
+         
+  PRINT @@context as results;
+}}
+INSTALL QUERY get_company_context
+"""
+print("Running GSQL Query installation (this usually takes 2-3 minutes)...")
+try:
+    print(conn.gsql(query_gsql))
+except Exception as e:
+    print(f"Query installation error: {e}")
+
+print("\n" + "="*50)
+print("ALL DONE! TigerGraph is now fully operational.")
+print("="*50)
+print("IMPORTANT: Copy this Bearer Token to use in your Dashboard:")
+print(f"{token}")
+print("="*50)

build_vector_db.py

@@ -0,0 +1,59 @@
+import json
+import os
+import chromadb
+from tqdm import tqdm
+
+DB_DIR = "./chroma_db"
+CORPUS_FILE = "financial_corpus.jsonl"
+LIMIT = int(os.environ.get("EMBED_LIMIT", "1000"))  # Configurable limit for local dev vs production
+
+def build_vector_db():
+    print("Initializing ChromaDB (this may download the embedding model if first run)...")
+    client = chromadb.PersistentClient(path=DB_DIR)
+    collection = client.get_or_create_collection(name="sec_filings")
+
+    print(f"Loading up to {LIMIT} documents from {CORPUS_FILE}...")
+    documents = []
+    metadatas = []
+    ids = []
+
+    count = 0
+    with open(CORPUS_FILE, 'r', encoding='utf-8') as f:
+        for line in f:
+            try:
+                data = json.loads(line)
+                # Filter out very short or empty chunks
+                text = data.get('text', '').strip()
+                if len(text) < 50:
+                    continue
+                    
+                documents.append(text)
+                metadatas.append({
+                    "company": data.get('company', 'Unknown'), 
+                    "year": data.get("year", "Unknown")
+                })
+                ids.append(f"doc_{count}")
+                count += 1
+                
+                if LIMIT > 0 and count >= LIMIT:
+                    break
+            except Exception as e:
+                print(f"Error parsing line: {e}")
+                continue
+
+    print(f"Adding {count} documents to Vector Store (this will automatically generate embeddings)...")
+    batch_size = 100
+    for i in tqdm(range(0, len(documents), batch_size)):
+        collection.add(
+            documents=documents[i:i+batch_size],
+            metadatas=metadatas[i:i+batch_size],
+            ids=ids[i:i+batch_size]
+        )
+
+    print("\n✅ Vector database built successfully! Ready for Basic RAG.")
+
+if __name__ == "__main__":
+    if not os.path.exists(CORPUS_FILE):
+        print(f"Error: {CORPUS_FILE} not found. Please ensure it's in the same directory.")
+    else:
+        build_vector_db()

count_tokens.py

@@ -0,0 +1,131 @@
+import os
+import json
+import time
+from datasets import load_dataset
+from google import genai
+
+def main():
+    if "GEMINI_API_KEY" not in os.environ:
+        print("ERROR: GEMINI_API_KEY environment variable not set.")
+        print("Please set it before running this script.")
+        return
+
+    client = genai.Client()
+    
+    TARGET_TOKENS = 105_000_000
+    current_tokens = 0
+    document_count = 0
+    
+    output_file = "financial_corpus.jsonl"
+    
+    print(f"Starting token counting using Gemini 'countTokens' API...")
+    print(f"Targeting {TARGET_TOKENS:,} tokens.")
+    
+    # Resume Logic
+    if os.path.exists(output_file):
+        print(f"Found existing {output_file}. Resuming...")
+        with open(output_file, 'r', encoding='utf-8') as f:
+            for line in f:
+                if not line.strip(): continue
+                data = json.loads(line)
+                current_tokens += data.get("tokens", 0)
+                document_count += 1
+        print(f"Resumed from {document_count} documents with {current_tokens:,} tokens.")
+    
+    if current_tokens >= TARGET_TOKENS:
+        print("Target already reached!")
+        return
+
+    start_time = time.time()
+    splits = ['001', '002', '003', '004', '005']
+    
+    docs_to_skip = document_count
+    
+    with open(output_file, 'a', encoding='utf-8') as f:
+        for split_id in splits:
+            if current_tokens >= TARGET_TOKENS:
+                break
+                
+            print(f"Loading split {split_id}...")
+            try:
+                dataset = load_dataset('winterForestStump/10-K_sec_filings', split=split_id, streaming=True)
+            except Exception as e:
+                print(f"Error loading dataset split {split_id}: {e}")
+                continue
+                
+            for row in dataset:
+                # Fast forward
+                if docs_to_skip > 0:
+                    docs_to_skip -= 1
+                    continue
+                    
+                if current_tokens >= TARGET_TOKENS:
+                    break
+                    
+                company = row.get("company_name", "Unknown Company")
+                year = str(row.get("filing_date", "Unknown Year"))[:4] if row.get("filing_date") else "Unknown Year"
+                
+                text_sections = []
+                if row.get("Business"): text_sections.append(str(row["Business"]))
+                if row.get("Risk Factors"): text_sections.append(str(row["Risk Factors"]))
+                
+                for k, v in row.items():
+                    if "Management" in k and "Discussion" in k and v:
+                        text_sections.append(str(v))
+                
+                combined_text = "\n\n".join(text_sections).strip()
+                
+                if not combined_text:
+                    continue
+                    
+                try:
+                    response = client.models.count_tokens(
+                        model='gemini-2.5-flash',
+                        contents=combined_text
+                    )
+                    tokens = response.total_tokens
+                    
+                    corpus_item = {
+                        "company": company,
+                        "year": year,
+                        "text": combined_text,
+                        "tokens": tokens
+                    }
+                    f.write(json.dumps(corpus_item) + "\n")
+                    f.flush()
+                    
+                    current_tokens += tokens
+                    document_count += 1
+                    
+                    if document_count % 10 == 0:
+                        print(f"Processed {document_count} documents. Total Tokens: {current_tokens:,}")
+                        
+                except Exception as e:
+                    print(f"API Error counting tokens for {company}: {e}")
+                    time.sleep(1) # Backoff
+
+    elapsed = time.time() - start_time
+    
+    print("\n--- TOKEN COUNTING COMPLETE ---")
+    print(f"Total Tokens: {current_tokens:,}")
+    print(f"Total Documents: {document_count}")
+    print(f"Time Taken for this run: {elapsed:.2f} seconds")
+    print(f"Corpus saved to: {output_file}")
+    
+    report = f"""TigerGraph Hackathon Benchmark Report
+-----------------------------------
+Dataset: winterForestStump/10-K_sec_filings (SEC 10-K Filings)
+Total Tokens: {current_tokens:,}
+Total Documents: {document_count}
+
+Tokenizer Documentation:
+As per the hackathon requirements, the dataset size was measured using the official Gemini API (`countTokens` endpoint) via the `google-genai` Python SDK. 
+The underlying tokenizer utilized by the API is the official Gemini SentencePiece tokenizer, ensuring exact alignment with the model's actual token usage for processing.
+"""
+    with open("benchmark_report.txt", "w") as f:
+        f.write(report)
+        
+    print("benchmark_report.txt generated successfully!")
+
+if __name__ == "__main__":
+    main()

download_db.py

@@ -0,0 +1,32 @@
+import os
+import zipfile
+import urllib.request
+
+DB_DIR = os.path.join(os.path.dirname(__file__), "chroma_db")
+# You must set DB_ZIP_URL as an environment variable in Hugging Face!
+ZIP_URL = os.environ.get("DB_ZIP_URL", "")
+
+def download_and_extract_db():
+    if os.path.exists(DB_DIR):
+        print("ChromaDB already exists locally. Skipping download.")
+        return
+        
+    if not ZIP_URL:
+        print("Warning: DB_ZIP_URL is not set. Cannot download Vector DB.")
+        return
+
+    print(f"Downloading ChromaDB ZIP from {ZIP_URL}...")
+    zip_path = "chroma_db.zip"
+    
+    try:
+        urllib.request.urlretrieve(ZIP_URL, zip_path)
+        print("Download complete. Extracting...")
+        with zipfile.ZipFile(zip_path, 'r') as zip_ref:
+            zip_ref.extractall(os.path.dirname(__file__) or ".")
+        print("Extraction complete. ChromaDB is ready.")
+        os.remove(zip_path)
+    except Exception as e:
+        print(f"Failed to download or extract DB: {e}")
+
+if __name__ == "__main__":
+    download_and_extract_db()

edges.csv

@@ -0,0 +1,503 @@
+source,target,e_type
+TROPICAL SPORTSWEAR INTERNATIONAL CORP.,Savane International Corp.,OWNS
+DATAWATCH CORP,interest rate risk,FACES_RISK
+DATAWATCH CORP,foreign currency exchange rate risk,FACES_RISK
+DATAWATCH CORP,DATAWATCH CORP International Subsidiaries,OWNS
+GALEY & LORD INC.,"Galey & Lord Industries, Inc.",OWNS
+"Galey & Lord Industries, Inc.","G&L Service Company, North America, Inc.",OWNS
+"G&L Service Company, North America, Inc.","Dimmit Industries, S.A. de C.V.",OWNS
+GALEY & LORD INC.,Swift Denim Group,OWNS
+GALEY & LORD INC.,Klopman International Group,OWNS
+GALEY & LORD INC.,"Swift Textiles Europe, Ltd.",OWNS
+"Polymer Group, Inc.","DT Acquisition, Inc.",OWNS
+GALEY & LORD INC.,"competitive and economic factors in the textile, apparel and home furnishings markets",FACES_RISK
+GALEY & LORD INC.,raw materials and other costs,FACES_RISK
+GALEY & LORD INC.,weather-related delays,FACES_RISK
+GALEY & LORD INC.,general economic conditions,FACES_RISK
+MEDIQ Incorporated,"MEDIQ/PRN Life Support Services, Inc.",OWNS
+MEDIQ Incorporated,SpectraCair,OWNS
+MEDIQ Incorporated,"CH Industries, Inc.",OWNS
+MEDIQ Incorporated,"National Patient Care Systems, Inc.",OWNS
+"CH Industries, Inc.","CH Medical, Inc.",OWNS
+"Bruckmann, Rosser, Sherrill & Co., L.P.",MEDIQ Incorporated,OWNS
+"Ferrer Freeman Thompson & Co., LLC",MEDIQ Incorporated,OWNS
+"Galen Partners, III L.P.",MEDIQ Incorporated,OWNS
+Rotko Entities,MEDIQ Incorporated,OWNS
+MEDIQ Incorporated,InnoServ Technologies,OWNS
+MEDIQ Incorporated,Cost Containment Pressures,FACES_RISK
+IEC Electronics Corp.,"Calidad Electronics, Inc.",OWNS
+IEC Electronics Corp.,"IEC - Edinburg Texas, Inc.",OWNS
+IEC Electronics Corp.,"Accutek, Inc.",OWNS
+IEC Electronics Corp.,IEC Arab Alabama Inc.,OWNS
+IEC Electronics Corp.,IEC Electronics - Ireland Limited,OWNS
+MEDIQ Incorporated,"MEDIQ/PRN Life Support Services, Inc.",OWNS
+"MEDIQ/PRN Life Support Services, Inc.","MEDIQ Management Services, Inc.",OWNS
+"MEDIQ/PRN Life Support Services, Inc.",SpectraCair,OWNS
+"MEDIQ/PRN Life Support Services, Inc.","CH Medical, Inc.",OWNS
+"MEDIQ/PRN Life Support Services, Inc.",Increasing pressure to reduce operating costs and capital expenditures,FACES_RISK
+"MEDIQ/PRN Life Support Services, Inc.",Cost containment pressures on health care providers,FACES_RISK
+"MEDIQ/PRN Life Support Services, Inc.",Reduced reimbursement levels by Medicare and other third-party payors,FACES_RISK
+"MEDIQ/PRN Life Support Services, Inc.",Legislative proposals leading to health care consolidations and managed care acceleration,FACES_RISK
+INTERSTATE JOHNSON LANE INC.,CapTrust,OWNS
+INTERSTATE JOHNSON LANE INC.,Highly Competitive Environment,FACES_RISK
+INTERSTATE JOHNSON LANE INC.,Marketplace Trading Volumes Volatility,FACES_RISK
+INTERSTATE JOHNSON LANE INC.,Interest Rate Conditions Volatility,FACES_RISK
+INTERSTATE JOHNSON LANE INC.,Revenue and Earnings Variability,FACES_RISK
+INTERSTATE JOHNSON LANE INC.,Significant Incremental Cash Requirements,FACES_RISK
+INTERSTATE JOHNSON LANE INC.,"Regulatory Liquidity and Capital Requirements (SEC, CFTC, NYSE)",FACES_RISK
+CapTrust,"Regulatory Liquidity and Capital Requirements (SEC, NASD)",FACES_RISK
+INTERSTATE JOHNSON LANE INC.,Year 2000 Technological Challenge,FACES_RISK
+INTERSTATE JOHNSON LANE INC.,Dependence on Third-Party Systems for Year 2000 Compliance,FACES_RISK
+INTERSTATE JOHNSON LANE INC.,Inflationary Effects,FACES_RISK
+"Flanigan's Enterprises, Inc.","Flanigan's Management Services, Inc.",OWNS
+"Flanigan's Enterprises, Inc.","Flanigan's Enterprises, Inc. of Georgia",OWNS
+"Flanigan's Enterprises, Inc.",Seventh Street Corp.,OWNS
+"Flanigan's Enterprises, Inc.",Big Daddy's #48 Inc.,OWNS
+"Flanigan's Enterprises, Inc.","Flanigan's Enterprises, Inc. of Pa.",OWNS
+"Flanigan's Enterprises, Inc.",Bankruptcy,FACES_RISK
+Subchapter S corporation (franchisee),Chairman of the Board,EMPLOYS
+TCI INTERNATIONAL INC.,Broadcast and Communications Division,OWNS
+TCI INTERNATIONAL INC.,Signal Processing Division,OWNS
+TCI INTERNATIONAL INC.,Test and Measurement Division,OWNS
+TCI INTERNATIONAL INC.,Potential Loss from Operations in Fiscal 1999,FACES_RISK
+TCI INTERNATIONAL INC.,Contract Termination Risk,FACES_RISK
+TCI INTERNATIONAL INC.,Reduced US Government Demand (Post-Cold War),FACES_RISK
+TCI INTERNATIONAL INC.,"Declining Economies in Asia, Africa, Middle East",FACES_RISK
+TCI INTERNATIONAL INC.,Strong US Dollar Impacting Competitiveness,FACES_RISK
+TCI INTERNATIONAL INC.,Delay/Loss of Key Program Awards,FACES_RISK
+TCI INTERNATIONAL INC.,Delays in System Acceptance due to Customer Turnover,FACES_RISK
+TCI INTERNATIONAL INC.,Foreign Competitors,COMPETES_WITH
+Spartech Corporation,Spartech Plastics,OWNS
+Spartech Corporation,Spartech Polycom,OWNS
+Spartech Corporation,Spartech Color,OWNS
+Spartech Corporation,Spartech Vy-Cal Plastics,OWNS
+Spartech Corporation,GenPak,OWNS
+Spartech Corporation,Hamelin Industries,OWNS
+Spartech Corporation,Spartech Profiles,OWNS
+Spartech Corporation,Product Components,OWNS
+Spartech Corporation,Pawnee Industries,OWNS
+Spartech Corporation,Portage Industries,OWNS
+Spartech Corporation,Hamelin Group,OWNS
+Spartech Corporation,Preferred Plastic Sheet Division of Echlin Inc.,OWNS
+Echlin Inc.,Preferred Plastic Sheet Division of Echlin Inc.,OWNS
+Spartech Corporation,"Polycom Huntsman, Inc.",OWNS
+Spartech Corporation,"Prismaplast Canada, Ltd.",OWNS
+Spartech Corporation,"Anjac-Doron Plastics, Inc.",OWNS
+"CMG Information Services, Inc.",ADSmart Corporation,OWNS
+"CMG Information Services, Inc.","Engage Technologies, Inc.",OWNS
+"CMG Information Services, Inc.","Accipiter, Inc.",OWNS
+"CMG Information Services, Inc.",InfoMation Publishing Corporation,OWNS
+"CMG Information Services, Inc.",NaviSite Internet Services Corporation,OWNS
+"CMG Information Services, Inc.",Planet Direct Corporation,OWNS
+"CMG Information Services, Inc.",Password Internet Publishing Corporation,OWNS
+"CMG Information Services, Inc.","Magnitude Network, LLC",OWNS
+"CMG Information Services, Inc.","Open Market, Inc.",OWNS
+"CMG Information Services, Inc.",NaviNet,OWNS
+"CMG Information Services, Inc.","CMG@Ventures I, LLC",OWNS
+"CMG@Ventures I, LLC","Blaxxun Interactive, Inc.",OWNS
+"CMG@Ventures I, LLC",GeoCities,OWNS
+"CMG@Ventures I, LLC","Lycos, Inc.",OWNS
+"CMG@Ventures I, LLC",Parable LLC,OWNS
+"CMG@Ventures I, LLC",Vicinity Corporation,OWNS
+"CMG Information Services, Inc.","CMG@Ventures II, LLC",OWNS
+"CMG@Ventures II, LLC",Chemdex Corporation,OWNS
+"CMG@Ventures II, LLC",Critical Path,OWNS
+"CMG@Ventures II, LLC",GeoCities,OWNS
+"CMG@Ventures II, LLC","KOZ, Inc.",OWNS
+"CMG@Ventures II, LLC","Mother Nature's General Store, Inc.",OWNS
+"CMG@Ventures II, LLC",Parable LLC,OWNS
+"CMG@Ventures II, LLC","Reel.com, Inc.",OWNS
+"CMG@Ventures II, LLC","Sage Enterprises, Inc.",OWNS
+"CMG@Ventures II, LLC","Silknet Software, Inc.",OWNS
+"CMG@Ventures II, LLC","Softway Systems, Inc.",OWNS
+"CMG@Ventures II, LLC",Speech Machines plc,OWNS
+"CMG@Ventures II, LLC",TicketsLive Corporation,OWNS
+"CMG@Ventures II, LLC",Universal Learning Technology,OWNS
+"CMG@Ventures II, LLC",Visto Corporation,OWNS
+"CMG Information Services, Inc.","CMG@Ventures III, LLC",OWNS
+"CMG Information Services, Inc.",SalesLink Corporation,OWNS
+"CMG Information Services, Inc.",InSolutions Incorporated,OWNS
+"CMG Information Services, Inc.","On-Demand Solutions, Inc.",OWNS
+SalesLink Corporation,Pacific Direct Marketing Corporation,OWNS
+"CMG Information Services, Inc.",CMG Direct Corporation,OWNS
+Planet Direct Corporation,InfoMation Publishing Corporation,OWNS
+"Amazon.com, Inc.","Sage Enterprises, Inc.",OWNS
+Hollywood Entertainment Corporation,"Reel.com, Inc.",OWNS
+Salex Holding Corporation,Pershing Sun,EMPLOYS
+"Sun Associates, LLC",Betty Sun,EMPLOYS
+Salex Holding Corporation,competition in the automobile asset management business,FACES_RISK
+Salex Holding Corporation,inability to obtain additional financing,FACES_RISK
+Salex Holding Corporation,general business and economic conditions,FACES_RISK
+Salex Holding Corporation,PH&H Corp.,COMPETES_WITH
+Salex Holding Corporation,GE Capital,COMPETES_WITH
+Franklin Receivables LLC,Franklin Auto Trust 1998-1,OWNS
+IEC Electronics Corp.,"Calidad Electronics, Inc.",OWNS
+IEC Electronics Corp.,"IEC - Edinburg Texas, Inc.",OWNS
+IEC Electronics Corp.,"Accutek, Inc.",OWNS
+IEC Electronics Corp.,IEC Arab Alabama Inc.,OWNS
+IEC Electronics Corp.,IEC Electronics - Ireland Limited,OWNS
+"RICA FOODS, INC.",Pipasa,OWNS
+"RICA FOODS, INC.",As de Oros,OWNS
+"RICA FOODS, INC.",Mr. Calixto Chaves,EMPLOYS
+"Inversiones La Ribera, S.A.",Pipasa,OWNS
+"Comercial Angui, S.A.",As de Oros,OWNS
+Antonio Echeverria,"Comercial Angui, S.A.",OWNS
+As de Oros,Restaurantes As de Oros,OWNS
+"RICA FOODS, INC.",Exchange Rate Risk,FACES_RISK
+"RICA FOODS, INC.",Foreign Competition,FACES_RISK
+"RICA FOODS, INC.",Commodity Price Volatility,FACES_RISK
+Pipasa,As de Oros,COMPETES_WITH
+GLOBAL ENVIRONMENTAL CORP.,Danzer,OWNS
+GLOBAL ENVIRONMENTAL CORP.,Weakened Financial Condition,FACES_RISK
+GLOBAL ENVIRONMENTAL CORP.,Tighter Credit Terms,FACES_RISK
+GLOBAL ENVIRONMENTAL CORP.,Need for Capital Infusion,FACES_RISK
+GLOBAL ENVIRONMENTAL CORP.,Dependency on Future Success,FACES_RISK
+GLOBAL ENVIRONMENTAL CORP.,Year 2000 Problem,FACES_RISK
+GLOBAL ENVIRONMENTAL CORP.,Major Service Provider System Failure,FACES_RISK
+MATLACK SYSTEMS INC.,Competitive Labor Environment,FACES_RISK
+MATLACK SYSTEMS INC.,Increased Environmental Compliance Costs,FACES_RISK
+MATLACK SYSTEMS INC.,Legal Action Settlement,FACES_RISK
+MATLACK SYSTEMS INC.,Potential Change in Ownership,FACES_RISK
+MATLACK SYSTEMS INC.,High Debt Levels,FACES_RISK
+MATLACK SYSTEMS INC.,Declining Demand for Transport,FACES_RISK
+MATLACK SYSTEMS INC.,General Economic Conditions,FACES_RISK
+MATLACK SYSTEMS INC.,Competitive Pressures,FACES_RISK
+MATLACK SYSTEMS INC.,Market Demand Shifts,FACES_RISK
+MATLACK SYSTEMS INC.,Industry Performance and Needs,FACES_RISK
+MATLACK SYSTEMS INC.,Equipment Utilization Challenges,FACES_RISK
+MATLACK SYSTEMS INC.,New Service Development Risk,FACES_RISK
+MATLACK SYSTEMS INC.,Increased Labor Costs,FACES_RISK
+MATLACK SYSTEMS INC.,Increased Operating Costs,FACES_RISK
+MATLACK SYSTEMS INC.,Litigation Uncertainties,FACES_RISK
+MATLACK SYSTEMS INC.,Financing Risk,FACES_RISK
+MATLACK SYSTEMS INC.,Insurance Availability Risk,FACES_RISK
+MATLACK SYSTEMS INC.,Project Completion Risk,FACES_RISK
+MATLACK SYSTEMS INC.,Management Retention Risk,FACES_RISK
+MATLACK SYSTEMS INC.,Regulatory Changes,FACES_RISK
+MATLACK SYSTEMS INC.,Y2K Compliance Risk,FACES_RISK
+MATLACK SYSTEMS INC.,Satellite System Failure,FACES_RISK
+MATLACK SYSTEMS INC.,Business Partner Y2K Compliance Risk,FACES_RISK
+ING Groep N.V.,"PFHI Holdings, Inc.",OWNS
+ING Groep N.V.,"Equitable of Iowa Companies, Inc.",OWNS
+"Equitable of Iowa Companies, Inc.",Equitable Life Insurance Company of Iowa,OWNS
+"Equitable of Iowa Companies, Inc.",Golden American,OWNS
+"Equitable of Iowa Companies, Inc.","Locust Street Securities, Inc.",OWNS
+"Equitable of Iowa Companies, Inc.","Equitable Investment Services, Inc.",OWNS
+"Equitable of Iowa Companies, Inc.","Directed Services, Inc.",OWNS
+"Equitable of Iowa Companies, Inc.",Equitable of Iowa Companies Capital Trust,OWNS
+"Equitable of Iowa Companies, Inc.",Equitable of Iowa Companies Capital Trust II,OWNS
+"Equitable of Iowa Companies, Inc.","Equitable of Iowa Securities Network, Inc.",OWNS
+Golden American,First Golden American Life Insurance Co of New York,OWNS
+First Golden American Life Insurance Co of New York,Discontinuation of sales relationship by significant distributor,FACES_RISK
+First Golden American Life Insurance Co of New York,Changes in acquisition cost allocation,FACES_RISK
+First Golden American Life Insurance Co of New York,Impairment of goodwill,FACES_RISK
+First Golden American Life Insurance Co of New York,Variability in amortization based on allocation and assumptions,FACES_RISK
+First Golden American Life Insurance Co of New York,Risk of loss from below investment grade securities,FACES_RISK
+First Golden American Life Insurance Co of New York,Sensitivity of below investment grade securities to adverse economic conditions,FACES_RISK
+First Golden American Life Insurance Co of New York,Future write-downs of securities,FACES_RISK
+First Golden American Life Insurance Co of New York,Material adverse effect on net income from significant write-downs,FACES_RISK
+IEC Electronics Corp.,"IEC - Edinburg Texas, Inc.",OWNS
+IEC Electronics Corp.,IEC Arab Alabama Inc.,OWNS
+IEC Electronics Corp.,IEC Electronics - Ireland Limited,OWNS
+RESOURCE AMERICA INC.,Resource Asset Investment Trust,OWNS
+RESOURCE AMERICA INC.,"The Atlas Group, Inc.",OWNS
+"Hynes & Howes Insurance Counselors, Inc.",Highly Competitive Market,FACES_RISK
+"Hynes & Howes Insurance Counselors, Inc.",Davenport Agency,OWNS
+Ralph Parry Insurance Agency Ltd.,Davenport Agency,OWNS
+"R.A.P. Enterprises, Inc.",Viking Agency,OWNS
+"R.A.P. Enterprises, Inc.","Hansen and Hansen, Inc.",OWNS
+"Hynes & Howes Insurance Counselors, Inc.",Ralph A. Parry,EMPLOYS
+Davenport Agency,Ralph A. Parry,EMPLOYS
+"R.A.P. Enterprises, Inc.",Ralph A. Parry,EMPLOYS
+Ralph A. Parry,"R.A.P. Enterprises, Inc.",OWNS
+Ralph A. Parry,Ralph Parry Insurance Agency Ltd.,OWNS
+"Hynes & Howes Insurance Counselors, Inc.",Low Liquidity,FACES_RISK
+"Hynes & Howes Insurance Counselors, Inc.",Decreased Operating Income,FACES_RISK
+VSI HOLDINGS INC.,The Performance Systems Group,OWNS
+VSI HOLDINGS INC.,"Advanced Animations, Inc.",OWNS
+VSI HOLDINGS INC.,"BKNT Retail Stores, Inc.",OWNS
+VSI HOLDINGS INC.,"Vispac, Inc.",OWNS
+VSI HOLDINGS INC.,"Visual Services, Inc.",OWNS
+VSI HOLDINGS INC.,"Steve Toth, Jr.",EMPLOYS
+VSI HOLDINGS INC.,Difficulty in recruiting qualified employees,FACES_RISK
+VSI HOLDINGS INC.,Loss of major customers,FACES_RISK
+VSI HOLDINGS INC.,Uncertainty of continued customer relationships,FACES_RISK
+Irvine Sensors Corporation,"Novalog, Inc.",OWNS
+Irvine Sensors Corporation,"MicroSensors, Inc.",OWNS
+Irvine Sensors Corporation,"Imagek, Inc.",OWNS
+"Imagek, Inc.",Uncertainty of Imagek financing,FACES_RISK
+Irvine Sensors Corporation,Uncertainty of widespread market success for stacked memory products,FACES_RISK
+Irvine Sensors Corporation,Cost disadvantage due to unestablished high volume production costs,FACES_RISK
+Irvine Sensors Corporation,Uncertainty of neural network chip set development success,FACES_RISK
+Irvine Sensors Corporation,Uncertainty of NeoStack commercialization timeline,FACES_RISK
+"General DataComm Industries, Inc.",VITAL Network Services LLC,OWNS
+RADIATION DISPOSAL SYSTEMS INC.,Manuel E. Kane,EMPLOYS
+RADIATION DISPOSAL SYSTEMS INC.,Albert D. Kane,EMPLOYS
+RADIATION DISPOSAL SYSTEMS INC.,Steven Kane,EMPLOYS
+"Steven M. Kane, Inc.",Steven Kane,EMPLOYS
+RADIATION DISPOSAL SYSTEMS INC.,Inability to Market Products,FACES_RISK
+RADIATION DISPOSAL SYSTEMS INC.,Lack of Significant Revenues,FACES_RISK
+RADIATION DISPOSAL SYSTEMS INC.,Cessation of Operations,FACES_RISK
+RADIATION DISPOSAL SYSTEMS INC.,Lack of Working Capital and Revenue Prospects,FACES_RISK
+RADIATION DISPOSAL SYSTEMS INC.,Uncertainty of Future Operations,FACES_RISK
+RADIATION DISPOSAL SYSTEMS INC.,Limited Market Acceptance of Ozone Technology,FACES_RISK
+RADIATION DISPOSAL SYSTEMS INC.,Undeveloped Market for Waste Treatment Technology,FACES_RISK
+RADIATION DISPOSAL SYSTEMS INC.,Competition from Established Firms,FACES_RISK
+RADIATION DISPOSAL SYSTEMS INC.,Negligible Market Position,FACES_RISK
+RADIATION DISPOSAL SYSTEMS INC.,Ineffective Sales Agreements,FACES_RISK
+RADIATION DISPOSAL SYSTEMS INC.,Inactive Marketing Efforts,FACES_RISK
+RADIATION DISPOSAL SYSTEMS INC.,Insufficient Production Personnel,FACES_RISK
+RADIATION DISPOSAL SYSTEMS INC.,Supplier Dependence and Production Delays,FACES_RISK
+RADIATION DISPOSAL SYSTEMS INC.,Single Source Supplier Dependence (Pillar),FACES_RISK
+RADIATION DISPOSAL SYSTEMS INC.,Lapse of Patents due to Insufficient Funds,FACES_RISK
+RADIATION DISPOSAL SYSTEMS INC.,Burdensome Environmental Regulations,FACES_RISK
+RADIATION DISPOSAL SYSTEMS INC.,High Cost of Radioactive Waste Disposal,FACES_RISK
+RADIATION DISPOSAL SYSTEMS INC.,Uneconomical Process,FACES_RISK
+RADIATION DISPOSAL SYSTEMS INC.,Costly Regulatory Modifications and Insufficient Funds,FACES_RISK
+RADIATION DISPOSAL SYSTEMS INC.,Future Adverse Government Regulations,FACES_RISK
+RADIATION DISPOSAL SYSTEMS INC.,Ultrox International,COMPETES_WITH
+RADIATION DISPOSAL SYSTEMS INC.,Para Oxidation,COMPETES_WITH
+RADIATION DISPOSAL SYSTEMS INC.,PCI Ozone Corporation,COMPETES_WITH
+RADIATION DISPOSAL SYSTEMS INC.,"Griffin Technics, Inc.",COMPETES_WITH
+RADIATION DISPOSAL SYSTEMS INC.,Henkel Corporation,COMPETES_WITH
+RADIATION DISPOSAL SYSTEMS INC.,"Brown-Bovari, Inc.",COMPETES_WITH
+"Stein Management Company, Inc.",Randy Rieger,EMPLOYS
+RTL Realty,Randy Rieger,EMPLOYS
+"Royal Palm Beach Colony, Limited Partnership",Uncertainty of future cash distribution,FACES_RISK
+"Royal Palm Beach Colony, Limited Partnership",Reduction in sales revenues,FACES_RISK
+"Royal Palm Beach Colony, Limited Partnership",Cash requirements for land development,FACES_RISK
+"Royal Palm Beach Colony, Limited Partnership",Cash expenditures for transactions and operations,FACES_RISK
+"Royal Palm Beach Colony, Limited Partnership",Bank financing repayment obligations,FACES_RISK
+"Royal Palm Beach Colony, Limited Partnership",Increased cost due to regulatory changes,FACES_RISK
+"Royal Palm Beach Colony, Limited Partnership",Highly competitive real estate market,FACES_RISK
+"Royal Palm Beach Colony, Limited Partnership",Competition affecting property sales,FACES_RISK
+"Royal Palm Beach Colony, Limited Partnership",Cyclical real estate market,FACES_RISK
+"Royal Palm Beach Colony, Limited Partnership",Interest rate increases,FACES_RISK
+"Royal Palm Beach Colony, Limited Partnership",Uncertain South Florida real estate market conditions,FACES_RISK
+"Royal Palm Beach Colony, Limited Partnership",Competition from other developers in South Florida,FACES_RISK
+"Royal Palm Beach Colony, Limited Partnership",Litigation over building permits,FACES_RISK
+"Royal Palm Beach Colony, Limited Partnership",Contingent land sales,FACES_RISK
+"Mesa Air Group, Inc.","Mesa Airlines, Inc.",OWNS
+"Mesa Air Group, Inc.","Air Midwest, Inc.",OWNS
+"Mesa Air Group, Inc.","WestAir Holdings, Inc.",OWNS
+"Mesa Air Group, Inc.","MPD, Inc.",OWNS
+"Mesa Air Group, Inc.","FCA, Inc.",OWNS
+"Mesa Air Group, Inc.","Mesa Leasing, Inc.",OWNS
+"Mesa Air Group, Inc.","MAGI Insurance, Ltd.",OWNS
+"WestAir Holdings, Inc.","WestAir Commuter Airlines, Inc.",OWNS
+"WestAir Holdings, Inc.","Regional Aircraft Services, Inc.",OWNS
+"Mesa Air Group, Inc.",Termination of Code-Sharing Agreements,FACES_RISK
+"Mesa Air Group, Inc.","CCAIR, Inc.",COMPETES_WITH
+"Mesa Air Group, Inc.",Great Lakes Aviation,COMPETES_WITH
+"Mesa Air Group, Inc.",SkyWest Airlines,COMPETES_WITH
+"Mesa Air Group, Inc.",United Airlines,COMPETES_WITH
+"Tidel Technologies, Inc.","Tidel Engineering, Inc.",OWNS
+"Tidel Technologies, Inc.",3CI Complete Compliance Corporation,OWNS
+"Tidel Technologies, Inc.","Diebold, Incorporated",COMPETES_WITH
+"Tidel Technologies, Inc.",NCR Corporation,COMPETES_WITH
+"Tidel Technologies, Inc.","Triton Systems, Inc.",COMPETES_WITH
+"Tidel Technologies, Inc.",Fujitsu Corporation,COMPETES_WITH
+"Tidel Technologies, Inc.",Siemens-Nixdorf,COMPETES_WITH
+"Tidel Technologies, Inc.",NKL Industries,COMPETES_WITH
+"Tidel Technologies, Inc.",McGunn Safe Company,COMPETES_WITH
+"Tidel Technologies, Inc.",Armor Safe Company,COMPETES_WITH
+"Tidel Technologies, Inc.",AutoVend,COMPETES_WITH
+"Tidel Technologies, Inc.",Inadequate Patent Protection,FACES_RISK
+"Tidel Technologies, Inc.",Patent Challenge Risk,FACES_RISK
+"Tidel Technologies, Inc.",Patent Infringement Risk,FACES_RISK
+"Tidel Technologies, Inc.",Third-Party Patent Rights Impact,FACES_RISK
+"Tidel Technologies, Inc.",Competitor Patent Acquisition,FACES_RISK
+"Tidel Technologies, Inc.",Trade Secret Loss,FACES_RISK
+3CI Complete Compliance Corporation,"Springhill, Louisiana Facility",OWNS
+3CI Complete Compliance Corporation,"Birmingham, Alabama Location",OWNS
+3CI Complete Compliance Corporation,Technological Obsolescence,FACES_RISK
+"Biocontrol Technology, Inc.","Diasense, Inc.",OWNS
+"Diasense, Inc.",Delays in FDA Marketing Approval,FACES_RISK
+"Biocontrol Technology, Inc.",Delays in FDA Marketing Approval,FACES_RISK
+"Diasense, Inc.",Uncertainty of Future Funding,FACES_RISK
+"Biocontrol Technology, Inc.",Uncertainty of Future Funding,FACES_RISK
+"Diasense, Inc.","Reliance on Biocontrol Technology, Inc.",FACES_RISK
+"Diasense, Inc.",Competition and Product Obsolescence,FACES_RISK
+"Biocontrol Technology, Inc.",Competition and Product Obsolescence,FACES_RISK
+"Diasense, Inc.",Dependence on Single Technology,FACES_RISK
+"Diasense, Inc.",Continued Operating Losses and Negative Net Worth,FACES_RISK
+"Biocontrol Technology, Inc.",Continued Operating Losses and Negative Net Worth,FACES_RISK
+"Diasense, Inc.",Potential Conflicts of Interest,FACES_RISK
+"Biocontrol Technology, Inc.",Potential Conflicts of Interest,FACES_RISK
+"Diasense, Inc.","Risks to Patents, Trademarks, and Licenses",FACES_RISK
+"Biocontrol Technology, Inc.","Risks to Patents, Trademarks, and Licenses",FACES_RISK
+"Diasense, Inc.",Uncertainty of Third-Party Payor Reimbursement,FACES_RISK
+"Biocontrol Technology, Inc.",Uncertainty of Third-Party Payor Reimbursement,FACES_RISK
+"Diasense, Inc.","Limited Sales, Marketing, and Manufacturing Experience",FACES_RISK
+"Biocontrol Technology, Inc.","Limited Sales, Marketing, and Manufacturing Experience",FACES_RISK
+"Diasense, Inc.",Difficulty Attracting and Retaining Key Employees,FACES_RISK
+"Biocontrol Technology, Inc.",Difficulty Attracting and Retaining Key Employees,FACES_RISK
+"Diasense, Inc.",Product Liability,FACES_RISK
+"Biocontrol Technology, Inc.",Product Liability,FACES_RISK
+"Diasense, Inc.",Uncertain Outcome of Pending Lawsuits,FACES_RISK
+"Biocontrol Technology, Inc.",Uncertain Outcome of Pending Lawsuits,FACES_RISK
+"Diasense, Inc.",Absence of Public Market for Common Stock,FACES_RISK
+"Diasense, Inc.",Control by Existing Shareholders,FACES_RISK
+"Diasense, Inc.",Dilution of Common Stock,FACES_RISK
+"Diasense, Inc.",Limited Marketability of Diasensorr 1000,FACES_RISK
+"Diasense, Inc.",Cash Flow Problems,FACES_RISK
+"Diasense, Inc.",Slowed Testing Pace,FACES_RISK
+"Diasense, Inc.",Necessity of FDA Approval for US Market,FACES_RISK
+"Biocontrol Technology, Inc.",Necessity of FDA Approval for US Market,FACES_RISK
+"Diasense, Inc.",Regulatory Process for Diasensorr 2000,FACES_RISK
+"Biocontrol Technology, Inc.",Regulatory Process for Diasensorr 2000,FACES_RISK
+"Diasense, Inc.",High Sales Price Limiting Sales,FACES_RISK
+STARCRAFT CORP /IN/,"Imperial Automotive Group, Inc.",OWNS
+STARCRAFT CORP /IN/,National Mobility Corporation,OWNS
+STARCRAFT CORP /IN/,"Starcraft Automotive Group, Inc.",OWNS
+STARCRAFT CORP /IN/,"Tecstar, Inc.",OWNS
+"Imperial Automotive Group, Inc.","Imperial Industries, Inc.",OWNS
+STARCRAFT CORP /IN/,Increased popularity of SUVs and factory minivans,FACES_RISK
+STARCRAFT CORP /IN/,Price pressure from higher chassis costs,FACES_RISK
+STARCRAFT CORP /IN/,Lower conversion inventory on dealer lots,FACES_RISK
+STARCRAFT CORP /IN/,Fewer automotive dealers selling conversion products,FACES_RISK
+STARCRAFT CORP /IN/,Changing level of dealer support,FACES_RISK
+STARCRAFT CORP /IN/,Strengthening of the U.S. dollar,FACES_RISK
+STARCRAFT CORP /IN/,Financial turmoil in Asian markets,FACES_RISK
+STARCRAFT CORP /IN/,Cyclical nature of the conversion industry,FACES_RISK
+STARCRAFT CORP /IN/,Consumer preferences for sport utility vehicles,FACES_RISK
+STARCRAFT CORP /IN/,Level of disposable consumer income,FACES_RISK
+STARCRAFT CORP /IN/,Availability and price of gasoline,FACES_RISK
+STARCRAFT CORP /IN/,Level of interest rates and availability of consumer financing,FACES_RISK
+STARCRAFT CORP /IN/,Decline in general economic conditions or consumer confidence,FACES_RISK
+STARCRAFT CORP /IN/,Dependence on OEMs for vehicle chassis,FACES_RISK
+STARCRAFT CORP /IN/,Decline in the conversion industry and Company sales,FACES_RISK
+"CMG Information Services, Inc.",ADSmart Corporation,OWNS
+"CMG Information Services, Inc.","Engage Technologies, Inc.",OWNS
+"CMG Information Services, Inc.","Accipiter, Inc.",OWNS
+"CMG Information Services, Inc.",InfoMation Publishing Corporation,OWNS
+"CMG Information Services, Inc.",NaviSite Internet Services Corporation,OWNS
+"CMG Information Services, Inc.",Planet Direct Corporation,OWNS
+"CMG Information Services, Inc.",Password Internet Publishing Corporation,OWNS
+"CMG Information Services, Inc.","Magnitude Network, LLC",OWNS
+"CMG Information Services, Inc.","Open Market, Inc.",OWNS
+"CMG Information Services, Inc.","CMG@Ventures I, LLC",OWNS
+"CMG@Ventures I, LLC","Blaxxun Interactive, Inc.",OWNS
+"CMG@Ventures I, LLC",GeoCities,OWNS
+"CMG@Ventures I, LLC","Lycos, Inc.",OWNS
+"CMG@Ventures I, LLC",Parable LLC,OWNS
+"CMG@Ventures I, LLC",Vicinity Corporation,OWNS
+"CMG Information Services, Inc.","CMG@Ventures II, LLC",OWNS
+"CMG@Ventures II, LLC",Chemdex Corporation,OWNS
+"CMG@Ventures II, LLC",Critical Path,OWNS
+"CMG@Ventures II, LLC",GeoCities,OWNS
+"CMG@Ventures II, LLC","KOZ, Inc.",OWNS
+"CMG@Ventures II, LLC","Mother Nature's General Store, Inc.",OWNS
+"CMG@Ventures II, LLC",Parable LLC,OWNS
+"CMG@Ventures II, LLC","Reel.com, Inc.",OWNS
+"CMG@Ventures II, LLC","Sage Enterprises, Inc.",OWNS
+"CMG@Ventures II, LLC","Silknet Software, Inc.",OWNS
+"CMG@Ventures II, LLC","Softway Systems, Inc.",OWNS
+"CMG@Ventures II, LLC",Speech Machines plc,OWNS
+"CMG@Ventures II, LLC",TicketsLive Corporation,OWNS
+"CMG@Ventures II, LLC",Universal Learning Technology,OWNS
+"CMG@Ventures II, LLC",Visto Corporation,OWNS
+"CMG Information Services, Inc.","CMG@Ventures III, LLC",OWNS
+"CMG Information Services, Inc.","@Ventures III, L.P.",OWNS
+"CMG Information Services, Inc.",SalesLink Corporation,OWNS
+"CMG Information Services, Inc.",InSolutions Incorporated,OWNS
+"CMG Information Services, Inc.","On-Demand Solutions, Inc.",OWNS
+SalesLink Corporation,Pacific Direct Marketing Corporation,OWNS
+"CMG Information Services, Inc.",CMG Direct Corporation,OWNS
+"CMG Information Services, Inc.",NaviNet,OWNS
+Covol Technologies Inc.,Inability to implement applications profitably,FACES_RISK
+Covol Technologies Inc.,Inability to utilize tax benefits,FACES_RISK
+Covol Technologies Inc.,Feedstock supply risk,FACES_RISK
+Covol Technologies Inc.,Production quality risk,FACES_RISK
+Covol Technologies Inc.,Product marketing risk,FACES_RISK
+"Lynton Group, Inc.","Ramapo Helicopters, Inc.",OWNS
+"Lynton Group, Inc.","Lynton Aircraft Maintenance Services, Inc.",OWNS
+"Lynton Group, Inc.","Rockland Aviation, Inc.",OWNS
+"Lynton Group, Inc.","Lynton Aviation Charter, Inc.",OWNS
+"Lynton Group, Inc.","LynStar Holdings, Inc.",OWNS
+"LynStar Holdings, Inc.","LynStar Aviation, Inc.",OWNS
+"Lynton Group, Inc.",Lynton Group Limited,OWNS
+Lynton Group Limited,Lynton Aviation Limited,OWNS
+"Lynton Group, Inc.",European Helicopters Limited,OWNS
+Lynton Group Limited,Magec Aviation Limited,OWNS
+Lynton Group Limited,Air Hanson Limited,OWNS
+"Lynton Group, Inc.","Lynton Jet Centre, Inc.",OWNS
+"Lynton Group, Inc.","Lynton Aviation, Inc.",OWNS
+"Lynton Group, Inc.","Lynton Aviation Services, Inc.",OWNS
+"Lynton Group, Inc.",Dollar Air Services Limited,OWNS
+Dollar Air Services Limited,Black Isle Helicopters Limited,OWNS
+"Lynton Group, Inc.",Black Isle Helicopters Limited,OWNS
+"Lynton Jet Centre, Inc.","Lynton Properties, Inc.",OWNS
+"Lynton Group, Inc.",PLM Dollar Group Limited,OWNS
+Air Hanson Limited,Air Hanson Engineering Limited,OWNS
+Air Hanson Limited,Air Hanson Aircraft Sales Limited,OWNS
+European Helicopters Limited,Loss of CAA/Manufacturer Certificates/Licenses for EHL/Air Hanson Engineering,FACES_RISK
+Air Hanson Engineering Limited,Loss of CAA/Manufacturer Certificates/Licenses for EHL/Air Hanson Engineering,FACES_RISK
+Magec Aviation Limited,Loss of CAA/Manufacturer Certificates/Licenses for Magec,FACES_RISK
+"General Automation, Inc.",General Automation LLC,OWNS
+"General Automation, Inc.",General Automation Canada Subsidiary,OWNS
+"General Automation, Inc.",General Automation Australia Subsidiary,OWNS
+"General Automation, Inc.",General Automation UK Subsidiary,OWNS
+"General Automation, Inc.",Going Concern Qualification,FACES_RISK
+"General Automation, Inc.",Recent Operating Losses,FACES_RISK
+"General Automation, Inc.",Deficit Net Worth,FACES_RISK
+"General Automation, Inc.",Working Capital Deficiency,FACES_RISK
+"General Automation, Inc.",Cash Flow Constraints,FACES_RISK
+"General Automation, Inc.",Obligations to TMI in Default,FACES_RISK
+"General Automation, Inc.",Obligations to Boundless in Default,FACES_RISK
+"General Automation, Inc.",Limited Market for the Company's Stock,FACES_RISK
+"General Automation, Inc.",Competition Risk,FACES_RISK
+"General Automation, Inc.",Wang,COMPETES_WITH
+"General Automation, Inc.",AT&T,COMPETES_WITH
+"General Automation, Inc.",IBM,COMPETES_WITH
+"General Automation, Inc.",Data General Corporation,COMPETES_WITH
+"General Automation, Inc.",Digital Equipment Corporation,COMPETES_WITH
+"General Automation, Inc.",Hewlett Packard Company,COMPETES_WITH
+"General Automation, Inc.",Ardent,COMPETES_WITH
+"General Automation, Inc.",Pick Systems,COMPETES_WITH
+Boundless Technologies,General Automation LLC,OWNS
+RCM TECHNOLOGIES INC.,The Consortium,OWNS
+RCM TECHNOLOGIES INC.,"The Consortium of Maryland, Inc.",OWNS
+RCM TECHNOLOGIES INC.,"Performance Staffing, Inc.",OWNS
+RCM TECHNOLOGIES INC.,"Programming Alternatives of Minnesota, Inc.",OWNS
+"Biomune Systems, Inc.","Optim Nutrition, Inc.",OWNS
+"Biomune Systems, Inc.","Rockwood Companies, LLC",OWNS
+"Biomune Systems, Inc.","Frank A. Eldredge, Ph.D.",EMPLOYS
+"Biomune Systems, Inc.","David O. Lucas, Ph.D.",EMPLOYS
+"Biomune Systems, Inc.","Craig D. Moffat, M.D.",EMPLOYS
+"Biomune Systems, Inc.",Limited access to funding for future research and development,FACES_RISK
+"Biomune Systems, Inc.",Highly competitive market for nutrition supplements,FACES_RISK
+"Biomune Systems, Inc.",Limited financial resources,FACES_RISK
+"Biomune Systems, Inc.",Risk of reduced ownership interest in Rockwood due to unfulfilled loan covenants,FACES_RISK
+"Biomune Systems, Inc.",Inability to obtain additional financing,FACES_RISK
+"Biomune Systems, Inc.",Inability to increase product sales,FACES_RISK
+Mobil Corp.,Mobil (German Operating Company),OWNS
+Exxon Corp.,German Subsidiaries of Exxon Corp.,OWNS
+Royal Dutch Group,German Subsidiaries of Royal Dutch Group,OWNS
+NORTH EUROPEAN OIL ROYALTY TRUST,John H. Van Kirk,EMPLOYS
+NORTH EUROPEAN OIL ROYALTY TRUST,John R. Van Kirk,EMPLOYS
+NORTH EUROPEAN OIL ROYALTY TRUST,Termination of Royalty Rights,FACES_RISK
+NORTH EUROPEAN OIL ROYALTY TRUST,Speculative Business,FACES_RISK
+NORTH EUROPEAN OIL ROYALTY TRUST,Uncertainty of Future Income,FACES_RISK
+NORTH EUROPEAN OIL ROYALTY TRUST,Fluctuations in Gas and Oil Prices/Supplies,FACES_RISK
+NORTH EUROPEAN OIL ROYALTY TRUST,Environmental Regulations Impact,FACES_RISK
+NORTH EUROPEAN OIL ROYALTY TRUST,Unforeseeable Economic and Political Factors,FACES_RISK
+NORTH EUROPEAN OIL ROYALTY TRUST,Currency Exchange Rate Fluctuations,FACES_RISK
+NORTH EUROPEAN OIL ROYALTY TRUST,Y2K Computer Compliance Issues,FACES_RISK
+Celsion Corporation,Dr. Gerald Wolf,EMPLOYS
+Celsion Corporation,Difficulty delivering controlled heat without burning healthy tissue,FACES_RISK
+Celsion Corporation,Need for new FDA indication for breast tumor ablation,FACES_RISK
+Pentech International Inc.,Sawdust Pencil Co.,OWNS
+Pentech International Inc.,Product Shortages,FACES_RISK
+Pentech International Inc.,Dumping Duties,FACES_RISK
+Pentech International Inc.,Increased Raw Material Costs,FACES_RISK
+Pentech International Inc.,Increased Dumping Duties,FACES_RISK
+Pentech International Inc.,Minimum Royalty Payments Leading to Losses,FACES_RISK
+Pentech International Inc.,Loss of Major Customer,FACES_RISK
+Pentech International Inc.,Bad Debts from Customers,FACES_RISK
+Pentech International Inc.,Failure to Register Trademarks,FACES_RISK
+Pentech International Inc.,Intense Competition,FACES_RISK
+Pentech International Inc.,Bic Pen Company,COMPETES_WITH
+Pentech International Inc.,Papermate,COMPETES_WITH
+Pentech International Inc.,"RoseArt Industries, Inc.",COMPETES_WITH
+Pentech International Inc.,Newell,COMPETES_WITH

extract_graph.py

@@ -0,0 +1,153 @@
+import os
+import json
+import csv
+import time
+import asyncio
+from pydantic import BaseModel, Field
+from google import genai
+from google.genai import types
+
+# Define the structured output schema for Gemini
+class Vertex(BaseModel):
+    id: str = Field(description="The unique identifier or name of the entity.")
+    type: str = Field(description="Must be one of: Company, Executive, RiskFactor, Subsidiary")
+
+class Edge(BaseModel):
+    source: str = Field(description="The id of the source vertex.")
+    target: str = Field(description="The id of the target vertex.")
+    type: str = Field(description="Must be one of: EMPLOYS, FACES_RISK, OWNS, COMPETES_WITH")
+
+class GraphExtraction(BaseModel):
+    vertices: list[Vertex]
+    edges: list[Edge]
+
+# Setup Gemini Client (Async)
+if "GEMINI_API_KEY" not in os.environ:
+    print("ERROR: GEMINI_API_KEY environment variable not set.")
+    exit(1)
+
+client = genai.Client()
+
+async def process_document(doc, sem):
+    """
+    Process a single document through Gemini 1.5 Flash to extract graph entities.
+    Uses a semaphore to limit concurrent API calls and prevent 429 errors.
+    """
+    async with sem:
+        text = doc.get("text", "")
+        company = doc.get("company", "Unknown")
+        
+        # If the text is too large, we truncate it for extraction to save costs 
+        # (For 100M tokens, we might process full texts, but let's limit to 15k chars per doc for the demo)
+        text_chunk = text[:15000] 
+        
+        prompt = f"""
+        Analyze the following SEC 10-K filing excerpt for {company}.
+        Extract all relevant entities (Company, Executive, RiskFactor, Subsidiary) and their relationships (EMPLOYS, FACES_RISK, OWNS, COMPETES_WITH).
+        
+        Text excerpt:
+        {text_chunk}
+        """
+        
+        try:
+            # Note: We use the async client `client.aio`
+            response = await client.aio.models.generate_content(
+                model='gemini-2.5-flash',
+                contents=prompt,
+                config=types.GenerateContentConfig(
+                    response_mime_type="application/json",
+                    response_schema=GraphExtraction,
+                    temperature=0.1
+                )
+            )
+            
+            # The response text will be a JSON string conforming to the GraphExtraction schema
+            return json.loads(response.text)
+            
+        except Exception as e:
+            print(f"Error extracting graph for {company}: {e}")
+            return None
+
+async def main():
+    input_file = "financial_corpus.jsonl"
+    if not os.path.exists(input_file):
+        print(f"ERROR: {input_file} not found. Run count_tokens.py first.")
+        return
+
+    # Prepare CSV writers
+    vertices_file = open("vertices.csv", "w", newline="", encoding="utf-8")
+    edges_file = open("edges.csv", "w", newline="", encoding="utf-8")
+    
+    v_writer = csv.writer(vertices_file)
+    e_writer = csv.writer(edges_file)
+    
+    # Write headers
+    v_writer.writerow(["v_id", "v_type"])
+    e_writer.writerow(["source", "target", "e_type"])
+    
+    # Read corpus
+    print("Loading corpus...")
+    docs = []
+    with open(input_file, 'r', encoding='utf-8') as f:
+        for line in f:
+            docs.append(json.loads(line))
+            if len(docs) >= 50:
+                break
+            
+    print(f"Loaded {len(docs)} documents for extraction.")
+    
+    # We use a Semaphore to allow max 5 concurrent requests to Gemini to respect standard tier rate limits
+    sem = asyncio.Semaphore(5)
+    
+    # Batch processing
+    BATCH_SIZE = 50
+    total_vertices = 0
+    total_edges = 0
+    
+    for i in range(0, len(docs), BATCH_SIZE):
+        batch = docs[i:i+BATCH_SIZE]
+        print(f"Processing batch {i//BATCH_SIZE + 1} ({len(batch)} docs)...")
+        
+        tasks = [process_document(doc, sem) for doc in batch]
+        results = await asyncio.gather(*tasks)
+        
+        for result in results:
+            if not result:
+                continue
+                
+            # Write vertices
+            for v in result.get("vertices", []):
+                # Basic cleaning
+                v_id = v.get("id", "").strip().replace("\n", " ")
+                v_type = v.get("type", "Unknown")
+                if v_id:
+                    v_writer.writerow([v_id, v_type])
+                    total_vertices += 1
+                    
+            # Write edges
+            for e in result.get("edges", []):
+                source = e.get("source", "").strip().replace("\n", " ")
+                target = e.get("target", "").strip().replace("\n", " ")
+                e_type = e.get("type", "Unknown")
+                if source and target:
+                    e_writer.writerow([source, target, e_type])
+                    total_edges += 1
+                    
+        # Flush to disk periodically
+        vertices_file.flush()
+        edges_file.flush()
+        
+        # Sleep slightly between batches to cool down rate limits
+        await asyncio.sleep(2)
+        
+    vertices_file.close()
+    edges_file.close()
+    
+    print("\n--- GRAPH EXTRACTION COMPLETE ---")
+    print(f"Total Vertices Extracted: {total_vertices:,}")
+    print(f"Total Edges Extracted: {total_edges:,}")
+    print("Saved to vertices.csv and edges.csv")
+    print("Ready for TigerGraph Bulk Loading!")
+
+if __name__ == "__main__":
+    asyncio.run(main())

index.html

@@ -0,0 +1,1025 @@
+<!DOCTYPE html>
+<html lang="en">
+<head>
+<meta charset="UTF-8" />
+<meta name="viewport" content="width=device-width, initial-scale=1.0" />
+<title>GraphRAG Pipeline Comparison Dashboard</title>
+<link href="https://fonts.googleapis.com/css2?family=Space+Mono:wght@400;700&family=DM+Sans:ital,wght@0,300;0,400;0,500;0,600;1,300&display=swap" rel="stylesheet" />
+<style>
+  *, *::before, *::after { box-sizing: border-box; margin: 0; padding: 0; }
+
+  :root {
+    --bg: #0a0c0f;
+    --surface: #111418;
+    --surface2: #181c22;
+    --border: rgba(255,255,255,0.08);
+    --border2: rgba(255,255,255,0.14);
+    --text: #e8eaed;
+    --muted: #7a8290;
+    --accent-teal: #00e5b0;
+    --accent-blue: #4d9fff;
+    --accent-amber: #ffb347;
+    --pipe1: #6b7280;
+    --pipe2: #4d9fff;
+    --pipe3: #00e5b0;
+    --font-mono: 'Space Mono', monospace;
+    --font-body: 'DM Sans', sans-serif;
+    --radius: 10px;
+    --radius-lg: 16px;
+  }
+
+  body {
+    font-family: var(--font-body);
+    background: var(--bg);
+    color: var(--text);
+    min-height: 100vh;
+    padding: 32px 24px;
+  }
+
+  /* ---- Header ---- */
+  .header {
+    display: flex;
+    align-items: flex-start;
+    justify-content: space-between;
+    margin-bottom: 32px;
+    gap: 16px;
+    flex-wrap: wrap;
+  }
+  .header-left {}
+  .wordmark {
+    font-family: var(--font-mono);
+    font-size: 11px;
+    letter-spacing: 0.15em;
+    color: var(--muted);
+    text-transform: uppercase;
+    margin-bottom: 6px;
+  }
+  .title {
+    font-size: 26px;
+    font-weight: 600;
+    letter-spacing: -0.5px;
+    line-height: 1.2;
+  }
+  .title span { color: var(--accent-teal); }
+  .subtitle {
+    font-size: 13px;
+    color: var(--muted);
+    margin-top: 4px;
+    font-weight: 300;
+  }
+  .tg-badge {
+    display: flex;
+    align-items: center;
+    gap: 8px;
+    background: var(--surface);
+    border: 1px solid var(--border2);
+    border-radius: var(--radius);
+    padding: 8px 14px;
+    font-size: 12px;
+    color: var(--muted);
+    align-self: flex-start;
+  }
+  .tg-dot { width: 8px; height: 8px; border-radius: 50%; background: #00e5b0; box-shadow: 0 0 6px #00e5b0; animation: pulse 2s infinite; }
+  @keyframes pulse { 0%,100%{opacity:1} 50%{opacity:0.4} }
+
+  /* ---- Config Panel ---- */
+  .config {
+    background: var(--surface);
+    border: 1px solid var(--border);
+    border-radius: var(--radius-lg);
+    padding: 20px 24px;
+    margin-bottom: 24px;
+  }
+  .config-title {
+    font-family: var(--font-mono);
+    font-size: 10px;
+    letter-spacing: 0.12em;
+    color: var(--muted);
+    text-transform: uppercase;
+    margin-bottom: 16px;
+  }
+  .config-grid {
+    display: grid;
+    grid-template-columns: 1fr 1fr 1fr;
+    gap: 12px;
+  }
+  @media (max-width: 700px) { .config-grid { grid-template-columns: 1fr; } }
+  .field label {
+    display: block;
+    font-size: 11px;
+    color: var(--muted);
+    margin-bottom: 5px;
+    font-family: var(--font-mono);
+    letter-spacing: 0.05em;
+  }
+  .field input, .field select {
+    width: 100%;
+    background: var(--surface2);
+    border: 1px solid var(--border2);
+    border-radius: var(--radius);
+    color: var(--text);
+    font-family: var(--font-mono);
+    font-size: 12px;
+    padding: 9px 12px;
+    outline: none;
+    transition: border-color 0.15s;
+  }
+  .field input:focus, .field select:focus { border-color: var(--accent-teal); }
+  .field input[type="password"] { letter-spacing: 0.05em; }
+  .field select option { background: #1a1e24; }
+
+  /* ---- Query Bar ---- */
+  .query-bar {
+    display: flex;
+    gap: 10px;
+    margin-bottom: 24px;
+    align-items: stretch;
+  }
+  .query-bar textarea {
+    flex: 1;
+    background: var(--surface);
+    border: 1px solid var(--border2);
+    border-radius: var(--radius-lg);
+    color: var(--text);
+    font-family: var(--font-body);
+    font-size: 15px;
+    padding: 14px 18px;
+    resize: none;
+    outline: none;
+    min-height: 60px;
+    line-height: 1.5;
+    transition: border-color 0.15s;
+  }
+  .query-bar textarea:focus { border-color: var(--accent-teal); }
+  .query-bar textarea::placeholder { color: var(--muted); }
+  .run-btn {
+    background: var(--accent-teal);
+    color: #0a0c0f;
+    border: none;
+    border-radius: var(--radius-lg);
+    font-family: var(--font-mono);
+    font-size: 12px;
+    font-weight: 700;
+    letter-spacing: 0.05em;
+    padding: 0 24px;
+    cursor: pointer;
+    white-space: nowrap;
+    transition: opacity 0.15s, transform 0.1s;
+    display: flex;
+    align-items: center;
+    gap: 6px;
+  }
+  .run-btn:hover { opacity: 0.85; }
+  .run-btn:active { transform: scale(0.97); }
+  .run-btn:disabled { opacity: 0.4; cursor: not-allowed; }
+  .run-btn svg { width: 16px; height: 16px; }
+
+  /* ---- Pipelines grid ---- */
+  .pipelines {
+    display: grid;
+    grid-template-columns: repeat(3, minmax(0, 1fr));
+    gap: 14px;
+    margin-bottom: 24px;
+  }
+  @media (max-width: 900px) { .pipelines { grid-template-columns: 1fr; } }
+
+  .pipe-card {
+    background: var(--surface);
+    border-radius: var(--radius-lg);
+    overflow: hidden;
+    display: flex;
+    flex-direction: column;
+    border: 1px solid var(--border);
+    transition: border-color 0.2s;
+  }
+  .pipe-card.active { border-color: var(--pipe-color, var(--border2)); }
+
+  .pipe-head {
+    padding: 14px 18px;
+    border-bottom: 1px solid var(--border);
+    display: flex;
+    align-items: center;
+    gap: 10px;
+  }
+  .pipe-icon {
+    width: 32px; height: 32px; border-radius: 8px;
+    display: flex; align-items: center; justify-content: center;
+    font-family: var(--font-mono); font-size: 11px; font-weight: 700;
+    flex-shrink: 0;
+  }
+  .p1 .pipe-icon { background: rgba(107,114,128,0.15); color: #9ca3af; }
+  .p2 .pipe-icon { background: rgba(77,159,255,0.12); color: var(--accent-blue); }
+  .p3 .pipe-icon { background: rgba(0,229,176,0.1); color: var(--accent-teal); }
+  .pipe-meta { flex: 1; min-width: 0; }
+  .pipe-name { font-weight: 600; font-size: 14px; }
+  .pipe-sub { font-size: 11px; color: var(--muted); font-weight: 300; margin-top: 1px; }
+  .status-chip {
+    font-family: var(--font-mono);
+    font-size: 9px;
+    letter-spacing: 0.1em;
+    padding: 3px 8px;
+    border-radius: 20px;
+    text-transform: uppercase;
+  }
+  .chip-idle { background: rgba(255,255,255,0.05); color: var(--muted); }
+  .chip-running { background: rgba(255,179,71,0.1); color: var(--accent-amber); }
+  .chip-done { background: rgba(0,229,176,0.1); color: var(--accent-teal); }
+  .chip-err { background: rgba(255,80,80,0.1); color: #ff6060; }
+
+  .pipe-body { padding: 16px 18px; flex: 1; display: flex; flex-direction: column; gap: 14px; }
+
+  .answer-box {
+    background: var(--surface2);
+    border-radius: var(--radius);
+    padding: 14px;
+    font-size: 13px;
+    line-height: 1.65;
+    color: #c8cdd6;
+    min-height: 120px;
+    font-weight: 300;
+    position: relative;
+  }
+  .placeholder { color: var(--muted); font-style: italic; }
+
+  .thinking { display: flex; gap: 5px; align-items: center; }
+  .thinking span {
+    width: 5px; height: 5px; border-radius: 50%;
+    animation: blink 1.3s infinite;
+  }
+  .p1 .thinking span { background: #6b7280; }
+  .p2 .thinking span { background: var(--accent-blue); }
+  .p3 .thinking span { background: var(--accent-teal); }
+  .thinking span:nth-child(2) { animation-delay: 0.2s; }
+  .thinking span:nth-child(3) { animation-delay: 0.4s; }
+  @keyframes blink { 0%,80%,100%{opacity:0.15} 40%{opacity:1} }
+
+  .metrics-grid {
+    display: grid;
+    grid-template-columns: 1fr 1fr;
+    gap: 8px;
+  }
+  .metric-tile {
+    background: var(--surface2);
+    border-radius: var(--radius);
+    padding: 10px 12px;
+  }
+  .metric-tile .mlabel {
+    font-family: var(--font-mono);
+    font-size: 9px;
+    letter-spacing: 0.1em;
+    color: var(--muted);
+    text-transform: uppercase;
+    margin-bottom: 4px;
+  }
+  .metric-tile .mval {
+    font-family: var(--font-mono);
+    font-size: 16px;
+    font-weight: 700;
+    color: var(--text);
+  }
+  .mval.good { color: var(--accent-teal); }
+  .mval.warn { color: var(--accent-amber); }
+  .mval.neutral { color: var(--muted); }
+
+  .eval-row { display: flex; gap: 8px; }
+  .eval-pill {
+    flex: 1;
+    border-radius: var(--radius);
+    padding: 8px 10px;
+    text-align: center;
+  }
+  .eval-pill .elabel {
+    font-size: 9px;
+    letter-spacing: 0.08em;
+    text-transform: uppercase;
+    font-family: var(--font-mono);
+    margin-bottom: 3px;
+    opacity: 0.6;
+  }
+  .eval-pill .eval-val {
+    font-family: var(--font-mono);
+    font-weight: 700;
+    font-size: 13px;
+  }
+  .pill-pass { background: rgba(0,229,176,0.08); color: var(--accent-teal); border: 1px solid rgba(0,229,176,0.2); }
+  .pill-fail { background: rgba(255,80,80,0.08); color: #ff6060; border: 1px solid rgba(255,80,80,0.2); }
+  .pill-neutral { background: var(--surface2); color: var(--muted); border: 1px solid var(--border); }
+
+  /* ---- Summary bar ---- */
+  .summary {
+    display: grid;
+    grid-template-columns: repeat(4, minmax(0, 1fr));
+    gap: 12px;
+    margin-bottom: 24px;
+  }
+  @media (max-width: 700px) { .summary { grid-template-columns: 1fr 1fr; } }
+  .sum-card {
+    background: var(--surface);
+    border: 1px solid var(--border);
+    border-radius: var(--radius-lg);
+    padding: 16px 18px;
+  }
+  .sum-card.highlight { border-color: rgba(0,229,176,0.3); }
+  .sum-card .slabel {
+    font-family: var(--font-mono);
+    font-size: 9px;
+    letter-spacing: 0.1em;
+    color: var(--muted);
+    text-transform: uppercase;
+    margin-bottom: 6px;
+  }
+  .sum-card .sval {
+    font-family: var(--font-mono);
+    font-size: 22px;
+    font-weight: 700;
+    color: var(--text);
+  }
+  .sum-card.highlight .sval { color: var(--accent-teal); }
+  .sum-card .ssub { font-size: 11px; color: var(--muted); margin-top: 3px; font-weight: 300; }
+
+  /* ---- Error / info banners ---- */
+  .banner {
+    padding: 12px 16px;
+    border-radius: var(--radius);
+    font-size: 13px;
+    margin-bottom: 16px;
+    display: none;
+  }
+  .banner.visible { display: block; }
+  .banner-err { background: rgba(255,80,80,0.08); border: 1px solid rgba(255,80,80,0.2); color: #ff8080; }
+  .banner-info { background: rgba(77,159,255,0.08); border: 1px solid rgba(77,159,255,0.2); color: #7db8ff; }
+
+  /* ---- Log ---- */
+  .log-section { margin-top: 8px; }
+  .log-toggle {
+    font-family: var(--font-mono);
+    font-size: 10px;
+    letter-spacing: 0.1em;
+    color: var(--muted);
+    cursor: pointer;
+    text-transform: uppercase;
+    background: none;
+    border: none;
+    padding: 0;
+    margin-bottom: 8px;
+    display: flex;
+    align-items: center;
+    gap: 6px;
+  }
+  .log-toggle:hover { color: var(--text); }
+  .log-box {
+    background: var(--surface2);
+    border-radius: var(--radius);
+    padding: 12px 14px;
+    font-family: var(--font-mono);
+    font-size: 11px;
+    color: var(--muted);
+    max-height: 180px;
+    overflow-y: auto;
+    line-height: 1.7;
+    display: none;
+  }
+  .log-box.open { display: block; }
+  .log-line.ok { color: var(--accent-teal); }
+  .log-line.warn { color: var(--accent-amber); }
+  .log-line.err { color: #ff6060; }
+
+  hr.divider { border: none; border-top: 1px solid var(--border); margin: 24px 0; }
+
+  /* ---- Setup guide ---- */
+  .setup {
+    background: var(--surface);
+    border: 1px solid var(--border);
+    border-radius: var(--radius-lg);
+    padding: 20px 24px;
+  }
+  .setup-title {
+    font-family: var(--font-mono);
+    font-size: 10px;
+    letter-spacing: 0.12em;
+    text-transform: uppercase;
+    color: var(--muted);
+    margin-bottom: 16px;
+  }
+  .steps {
+    display: grid;
+    grid-template-columns: repeat(auto-fit, minmax(200px, 1fr));
+    gap: 10px;
+  }
+  .step {
+    display: flex;
+    gap: 10px;
+    align-items: flex-start;
+  }
+  .step-num {
+    font-family: var(--font-mono);
+    font-size: 10px;
+    color: var(--accent-teal);
+    background: rgba(0,229,176,0.08);
+    border-radius: 6px;
+    width: 24px; height: 24px;
+    display: flex; align-items: center; justify-content: center;
+    flex-shrink: 0;
+    margin-top: 1px;
+  }
+  .step-text { font-size: 12px; color: var(--muted); line-height: 1.5; }
+  .step-text strong { color: var(--text); font-weight: 500; }
+  code {
+    font-family: var(--font-mono);
+    font-size: 11px;
+    background: var(--surface2);
+    border-radius: 4px;
+    padding: 1px 5px;
+    color: var(--accent-teal);
+  }
+</style>
+</head>
+<body>
+
+<div class="header">
+  <div class="header-left">
+    <div class="wordmark">GraphRAG · Scale Edition</div>
+    <div class="title">Financial <span>Intelligence</span></div>
+    <div class="subtitle">100M+ Token Source Data → 692 Node Knowledge Graph: LLM-Only vs Basic RAG vs TigerGraph GraphRAG</div>
+  </div>
+  <div class="tg-badge">
+    <span class="tg-dot"></span>
+    TigerGraph Cloud (Savanna)
+  </div>
+</div>
+
+<!-- Config -->
+<div class="config">
+  <div class="config-title">⚙ TigerGraph Cloud Config</div>
+  <div class="config-grid">
+    <div class="field">
+      <label>Cluster Host</label>
+      <input type="text" id="tg-host" placeholder="https://your-cluster.i.tgcloud.io" value="https://tg-8a493b67-4330-406c-81c2-d1398562a1ec.tg-3452941248.i.tgcloud.io" />
+    </div>
+    <div class="field">
+      <label>Graph Name</label>
+      <input type="text" id="tg-graph" placeholder="TIgerrag" value="TIgerrag" />
+    </div>
+    <div class="field">
+      <label>RESTPP Token</label>
+      <input type="password" id="tg-token" placeholder="Bearer token from Admin Portal" />
+    </div>
+    <div class="field">
+      <label>GraphRAG Query Name</label>
+      <input type="text" id="tg-query" placeholder="get_company_context" value="get_company_context" />
+    </div>
+    <div class="field">
+      <label>LLM Provider</label>
+      <select id="llm-provider" disabled>
+        <option value="gemini">Google (Gemini 2.5 Flash)</option>
+      </select>
+    </div>
+    <div class="field">
+      <label>LLM API Key</label>
+      <input type="password" id="llm-key" placeholder="sk-..." />
+    </div>
+  </div>
+</div>
+
+<!-- Query bar -->
+<div class="query-bar">
+  <textarea id="query-input" placeholder="e.g. What subsidiaries does this company own, and what are their primary market risks? — all three pipelines will answer simultaneously…" rows="2"></textarea>
+  <button class="run-btn" id="run-btn" onclick="runAll()">
+    <svg viewBox="0 0 16 16" fill="currentColor"><path d="M3 2.5l10 5.5-10 5.5V2.5z"/></svg>
+    RUN
+  </button>
+</div>
+
+<div class="banner banner-err" id="banner-err"></div>
+<div class="banner banner-info" id="banner-info"></div>
+
+<!-- Pipelines -->
+<div class="pipelines">
+
+  <!-- Pipeline 1 -->
+  <div class="pipe-card p1" id="card1">
+    <div class="pipe-head">
+      <div class="pipe-icon">P1</div>
+      <div class="pipe-meta">
+        <div class="pipe-name">LLM-Only</div>
+        <div class="pipe-sub">No retrieval · baseline</div>
+      </div>
+      <span class="status-chip chip-idle" id="status1">IDLE</span>
+    </div>
+    <div class="pipe-body">
+      <div class="answer-box" id="ans1"><span class="placeholder">Response will appear here…</span></div>
+      <div class="metrics-grid">
+        <div class="metric-tile"><div class="mlabel">Tokens</div><div class="mval neutral" id="tok1">—</div></div>
+        <div class="metric-tile"><div class="mlabel">Latency</div><div class="mval neutral" id="lat1">—</div></div>
+        <div class="metric-tile"><div class="mlabel">Cost/query</div><div class="mval neutral" id="cost1">—</div></div>
+        <div class="metric-tile"><div class="mlabel">Groundedness F1</div><div class="mval neutral" id="bert1">—</div></div>
+      </div>
+      <div class="eval-row">
+        <div class="eval-pill pill-neutral" id="judge1"><div class="elabel">LLM Judge</div><div class="eval-val">—</div></div>
+        <div class="eval-pill pill-neutral"><div class="elabel">Retrieval</div><div class="eval-val">None</div></div>
+      </div>
+    </div>
+  </div>
+
+  <!-- Pipeline 2 -->
+  <div class="pipe-card p2" id="card2">
+    <div class="pipe-head">
+      <div class="pipe-icon">P2</div>
+      <div class="pipe-meta">
+        <div class="pipe-name">Basic RAG</div>
+        <div class="pipe-sub">Vector embeddings · baseline</div>
+      </div>
+      <span class="status-chip chip-idle" id="status2">IDLE</span>
+    </div>
+    <div class="pipe-body">
+      <div class="answer-box" id="ans2"><span class="placeholder">Response will appear here…</span></div>
+      <div class="metrics-grid">
+        <div class="metric-tile"><div class="mlabel">Tokens</div><div class="mval neutral" id="tok2">—</div></div>
+        <div class="metric-tile"><div class="mlabel">Latency</div><div class="mval neutral" id="lat2">—</div></div>
+        <div class="metric-tile"><div class="mlabel">Cost/query</div><div class="mval neutral" id="cost2">—</div></div>
+        <div class="metric-tile"><div class="mlabel">Groundedness F1</div><div class="mval neutral" id="bert2">—</div></div>
+      </div>
+      <div class="eval-row">
+        <div class="eval-pill pill-neutral" id="judge2"><div class="elabel">LLM Judge</div><div class="eval-val">—</div></div>
+        <div class="eval-pill pill-neutral"><div class="elabel">Retrieval</div><div class="eval-val">Vector</div></div>
+      </div>
+    </div>
+  </div>
+
+  <!-- Pipeline 3 -->
+  <div class="pipe-card p3" id="card3">
+    <div class="pipe-head">
+      <div class="pipe-icon">P3</div>
+      <div class="pipe-meta">
+        <div class="pipe-name">GraphRAG</div>
+        <div class="pipe-sub">TigerGraph Cloud · multi-hop</div>
+      </div>
+      <span class="status-chip chip-idle" id="status3">IDLE</span>
+    </div>
+    <div class="pipe-body">
+      <div class="answer-box" id="ans3"><span class="placeholder">Response will appear here…</span></div>
+      <div class="metrics-grid">
+        <div class="metric-tile"><div class="mlabel">Tokens</div><div class="mval neutral" id="tok3">—</div></div>
+        <div class="metric-tile"><div class="mlabel">Latency</div><div class="mval neutral" id="lat3">—</div></div>
+        <div class="metric-tile"><div class="mlabel">Cost/query</div><div class="mval neutral" id="cost3">—</div></div>
+        <div class="metric-tile"><div class="mlabel">Groundedness F1</div><div class="mval neutral" id="bert3">—</div></div>
+      </div>
+      <div class="eval-row">
+        <div class="eval-pill pill-neutral" id="judge3"><div class="elabel">LLM Judge</div><div class="eval-val">—</div></div>
+        <div class="eval-pill pill-neutral" id="hops3"><div class="elabel">Retrieval</div><div class="eval-val">Graph</div></div>
+      </div>
+    </div>
+  </div>
+
+</div>
+
+<!-- Summary -->
+<div class="summary" id="summary" style="display:none;">
+  <div class="sum-card">
+    <div class="slabel">Baseline tokens (RAG)</div>
+    <div class="sval" id="sum-baseline">—</div>
+    <div class="ssub" id="sum-baseline-sub">—</div>
+  </div>
+  <div class="sum-card highlight">
+    <div class="slabel">Token reduction</div>
+    <div class="sval" id="sum-reduction">—</div>
+    <div class="ssub" id="sum-reduction-sub">vs Basic RAG</div>
+  </div>
+  <div class="sum-card">
+    <div class="slabel">Cost saving / query</div>
+    <div class="sval" id="sum-saving">—</div>
+    <div class="ssub" id="sum-saving-sub">GraphRAG vs RAG</div>
+  </div>
+  <div class="sum-card">
+    <div class="slabel">Groundedness delta</div>
+    <div class="sval" id="sum-accuracy">—</div>
+    <div class="ssub" id="sum-accuracy-sub">Context Faithfulness F1</div>
+  </div>
+</div>
+
+<!-- Log -->
+<div class="log-section">
+  <button class="log-toggle" onclick="toggleLog()">▶ Request log</button>
+  <div class="log-box" id="log-box"></div>
+</div>
+
+<hr class="divider" />
+
+<!-- Setup guide -->
+<div class="setup">
+  <div class="setup-title">⚡ TigerGraph Cloud Setup (from your guide)</div>
+  <div class="steps">
+    <div class="step">
+      <div class="step-num">1</div>
+      <div class="step-text"><strong>Create account</strong> at <code>tgcloud.io</code> via email or GitHub, verify, log into dashboard.</div>
+    </div>
+    <div class="step">
+      <div class="step-num">2</div>
+      <div class="step-text"><strong>Create a cluster</strong> — click "Create Cluster", choose Free Tier, pick your region, name it.</div>
+    </div>
+    <div class="step">
+      <div class="step-num">3</div>
+      <div class="step-text"><strong>Open GraphStudio</strong> — Tools → GraphStudio. Design your schema (vertices + edges).</div>
+    </div>
+    <div class="step">
+      <div class="step-num">4</div>
+      <div class="step-text"><strong>Load data</strong> via CSV upload + Loading Jobs, or REST / pyTigerGraph.</div>
+    </div>
+    <div class="step">
+      <div class="step-num">5</div>
+      <div class="step-text"><strong>Write GSQL query</strong>, install it — this generates a REST endpoint automatically.</div>
+    </div>
+    <div class="step">
+      <div class="step-num">6</div>
+      <div class="step-text"><strong>Generate RESTPP token</strong> in Admin Portal → User Management → paste it above.</div>
+    </div>
+    <div class="step">
+      <div class="step-num">7</div>
+      <div class="step-text">Endpoint format: <code>https://&lt;cluster&gt;.i.tgcloud.io/query/&lt;GRAPH&gt;/&lt;QUERY&gt;</code></div>
+    </div>
+    <div class="step">
+      <div class="step-num">8</div>
+      <div class="step-text"><strong>Test in Postman</strong> with GET + <code>Authorization: Bearer YOUR_TOKEN</code> before wiring here.</div>
+    </div>
+  </div>
+</div>
+
+<script>
+// ─── Pricing table (per 1K tokens, USD) ───────────────────────────────────────
+const PRICING = {
+  anthropic: { input: 0.003, output: 0.015 },   // Claude Sonnet
+  openai:    { input: 0.005, output: 0.015 },   // GPT-4o
+  gemini:    { input: 0.00025, output: 0.0005 } // Gemini 1.5 Flash
+};
+
+// ─── State ────────────────────────────────────────────────────────────────────
+let logOpen = false;
+const logs = [];
+
+function log(msg, type = '') {
+  const ts = new Date().toISOString().slice(11, 23);
+  logs.push({ ts, msg, type });
+  const box = document.getElementById('log-box');
+  const line = document.createElement('div');
+  line.className = 'log-line ' + type;
+  line.textContent = `[${ts}] ${msg}`;
+  box.appendChild(line);
+  box.scrollTop = box.scrollHeight;
+}
+
+function toggleLog() {
+  logOpen = !logOpen;
+  document.getElementById('log-box').className = 'log-box' + (logOpen ? ' open' : '');
+  document.querySelector('.log-toggle').textContent = (logOpen ? '▼' : '▶') + ' Request log';
+}
+
+function setStatus(n, state) {
+  const el = document.getElementById('status' + n);
+  el.className = 'status-chip';
+  if (state === 'running') { el.classList.add('chip-running'); el.textContent = 'RUNNING'; }
+  else if (state === 'done') { el.classList.add('chip-done'); el.textContent = 'DONE'; }
+  else if (state === 'error') { el.classList.add('chip-err'); el.textContent = 'ERROR'; }
+  else { el.classList.add('chip-idle'); el.textContent = 'IDLE'; }
+}
+
+function showThinking(n, pipe) {
+  document.getElementById('ans' + n).innerHTML =
+    `<div class="thinking"><span></span><span></span><span></span></div>`;
+  setStatus(n, 'running');
+  ['tok','lat','cost','bert'].forEach(k => {
+    const el = document.getElementById(k + n);
+    el.textContent = '…';
+    el.className = 'mval neutral';
+  });
+  document.getElementById('judge' + n).innerHTML =
+    `<div class="elabel">LLM Judge</div><div class="eval-val">…</div>`;
+  document.getElementById('judge' + n).className = 'eval-pill pill-neutral';
+}
+
+function renderAnswer(n, text, tokens, latency, bert, pass, provider) {
+  document.getElementById('ans' + n).textContent = text;
+  setStatus(n, 'done');
+
+  const price = PRICING[provider] || PRICING.anthropic;
+  const inputTok = Math.round(tokens * 0.7);
+  const outputTok = Math.round(tokens * 0.3);
+  const cost = ((inputTok / 1000) * price.input + (outputTok / 1000) * price.output);
+
+  const tokEl = document.getElementById('tok' + n);
+  tokEl.textContent = tokens.toLocaleString();
+  tokEl.className = 'mval ' + (n === 3 ? 'good' : n === 1 ? 'warn' : '');
+
+  document.getElementById('lat' + n).textContent = latency.toFixed(1) + 's';
+  document.getElementById('lat' + n).className = 'mval';
+
+  const costEl = document.getElementById('cost' + n);
+  costEl.textContent = '$' + cost.toFixed(4);
+  costEl.className = 'mval ' + (n === 3 ? 'good' : n === 1 ? 'warn' : '');
+
+  const bertEl = document.getElementById('bert' + n);
+  bertEl.textContent = bert.toFixed(2);
+  bertEl.className = 'mval ' + (bert >= 0.88 ? 'good' : bert >= 0.7 ? '' : 'warn');
+
+  const judgeEl = document.getElementById('judge' + n);
+  judgeEl.className = 'eval-pill ' + (pass ? 'pill-pass' : 'pill-fail');
+  judgeEl.innerHTML = `<div class="elabel">LLM Judge</div><div class="eval-val">${pass ? 'PASS' : 'FAIL'}</div>`;
+}
+
+function renderError(n, msg) {
+  document.getElementById('ans' + n).innerHTML = `<span style="color:#ff6060">${msg}</span>`;
+  setStatus(n, 'error');
+}
+
+function showBanner(type, msg) {
+  const el = document.getElementById('banner-' + type);
+  el.textContent = msg;
+  el.classList.add('visible');
+  setTimeout(() => el.classList.remove('visible'), 6000);
+}
+
+// ─── LLM call (Anthropic as demo; swap headers for OpenAI/Gemini) ─────────────
+async function callLLM(provider, apiKey, systemPrompt, userPrompt) {
+  if (!apiKey) throw new Error('No LLM API key set');
+
+  if (provider === 'anthropic') {
+    const res = await fetch('https://api.anthropic.com/v1/messages', {
+      method: 'POST',
+      headers: {
+        'Content-Type': 'application/json',
+        'x-api-key': apiKey,
+        'anthropic-version': '2023-06-01'
+      },
+      body: JSON.stringify({
+        model: 'claude-sonnet-4-20250514',
+        max_tokens: 512,
+        system: systemPrompt,
+        messages: [{ role: 'user', content: userPrompt }]
+      })
+    });
+    if (!res.ok) throw new Error(`Anthropic API ${res.status}: ${await res.text()}`);
+    const d = await res.json();
+    return {
+      text: d.content.map(b => b.text || '').join(''),
+      inputTokens: d.usage?.input_tokens || 0,
+      outputTokens: d.usage?.output_tokens || 0
+    };
+  }
+
+  if (provider === 'openai') {
+    const res = await fetch('https://api.openai.com/v1/chat/completions', {
+      method: 'POST',
+      headers: { 'Content-Type': 'application/json', 'Authorization': `Bearer ${apiKey}` },
+      body: JSON.stringify({
+        model: 'gpt-4o',
+        max_tokens: 512,
+        messages: [{ role: 'system', content: systemPrompt }, { role: 'user', content: userPrompt }]
+      })
+    });
+    if (!res.ok) throw new Error(`OpenAI API ${res.status}`);
+    const d = await res.json();
+    return {
+      text: d.choices[0].message.content,
+      inputTokens: d.usage?.prompt_tokens || 0,
+      outputTokens: d.usage?.completion_tokens || 0
+    };
+  }
+
+  if (provider === 'gemini') {
+    // Dynamically fetch available models for this API key to avoid 404s
+    const modelsUrl = `https://generativelanguage.googleapis.com/v1beta/models?key=${apiKey}`;
+    const modelsRes = await fetch(modelsUrl);
+    if (!modelsRes.ok) {
+      const err = await modelsRes.text();
+      throw new Error(`Gemini API Auth Error ${modelsRes.status}: ${err}`);
+    }
+    
+    const modelsData = await modelsRes.json();
+    // Find a model that supports generateContent (prefer 1.5-flash, then any gemini)
+    let selectedModel = modelsData.models.find(m => m.name.includes('gemini-1.5-flash') && m.supportedGenerationMethods?.includes('generateContent'))?.name;
+    if (!selectedModel) {
+      selectedModel = modelsData.models.find(m => m.name.includes('gemini') && m.supportedGenerationMethods?.includes('generateContent'))?.name;
+    }
+    
+    if (!selectedModel) throw new Error("No compatible Gemini models found for this API key.");
+
+    const url = `https://generativelanguage.googleapis.com/v1beta/${selectedModel}:generateContent?key=${apiKey}`;
+    const payload = {
+      method: 'POST',
+      headers: { 'Content-Type': 'application/json' },
+      body: JSON.stringify({
+        contents: [{ parts: [{ text: systemPrompt + '\n\nUser: ' + userPrompt }] }],
+        generationConfig: { maxOutputTokens: 512 }
+      })
+    };
+    
+    const res = await fetch(url, payload);
+
+    if (!res.ok) {
+      const errText = await res.text();
+      let errMsg = errText;
+      try { errMsg = JSON.parse(errText).error.message; } catch(e) {}
+      throw new Error(`Gemini API ${res.status}: ${errMsg}`);
+    }
+    
+    const d = await res.json();
+    return {
+      text: d.candidates[0].content.parts.map(p => p.text).join(''),
+      inputTokens: d.usageMetadata?.promptTokenCount || 0,
+      outputTokens: d.usageMetadata?.candidatesTokenCount || 0
+    };
+  }
+
+  throw new Error('Unknown provider');
+}
+
+// ─── TigerGraph RESTPP call ────────────────────────────────────────────────────
+async function callTigerGraph(host, graph, token, queryName, question) {
+  const url = `${host.replace(/\/$/, '')}/query/${graph}/${queryName}?company_name=${encodeURIComponent(question)}`;
+  log(`TG → GET ${url}`, '');
+  const res = await fetch(url, {
+    headers: { 'Authorization': `Bearer ${token}` }
+  });
+  if (!res.ok) {
+    const txt = await res.text();
+    throw new Error(`TigerGraph ${res.status}: ${txt.slice(0, 200)}`);
+  }
+  const data = await res.json();
+  // TigerGraph returns { results: [...] }; the query should return context + answer
+  return data;
+}
+
+// Real BERTScore is now calculated mathematically in the Python backend via HuggingFace's bert_score library.
+
+// ─── LLM-as-a-Judge ─────────────��─────────────────────────────────────────────
+async function judgeAnswer(provider, apiKey, question, answer) {
+  try {
+    const res = await callLLM(provider, apiKey,
+      'You are a strict benchmarking judge. Reply ONLY with the exact word PASS or the exact word FAIL. No punctuation, no explanation.',
+      `User Question: ${question}\n\nAI Answer: ${answer}\n\nTask: Evaluate if the AI completely and accurately answered the user's specific question. To PASS, the AI MUST explicitly address the core intent of the question. If the AI provides irrelevant facts (like acquisitions) instead of answering the specific question asked, or if it says it cannot find the information, you MUST reply FAIL.`
+    );
+    const text = res.text.toUpperCase();
+    console.log(`Judge Evaluation for Answer: "${answer.slice(0, 30)}..." -> Judge Output: "${text}"`);
+    if (text.includes('PASS')) return true;
+    return false;
+  } catch { return null; }
+}
+
+// ─── Main runner ──────────────────────────────────────────────────────────────
+async function runAll() {
+  const query = document.getElementById('query-input').value.trim();
+  if (!query) { showBanner('err', 'Please enter a question first.'); return; }
+
+  const provider  = document.getElementById('llm-provider').value;
+  const apiKey    = document.getElementById('llm-key').value.trim();
+  const tgHost    = document.getElementById('tg-host').value.trim();
+  const tgGraph   = document.getElementById('tg-graph').value.trim();
+  const tgToken   = document.getElementById('tg-token').value.trim();
+  const tgQuery   = document.getElementById('tg-query').value.trim();
+
+  if (!apiKey) { showBanner('err', 'Enter your LLM API key in the config panel.'); return; }
+
+  const btn = document.getElementById('run-btn');
+  btn.disabled = true;
+  document.getElementById('summary').style.display = 'none';
+  [1,2,3].forEach(n => showThinking(n));
+  log(`Query: "${query}"`, 'ok');
+  log(`Provider: ${provider}`, '');
+
+  // ── Pipeline 1: LLM-Only ──────────────────────────────────────────────────
+  const p1 = (async () => {
+    try {
+      log('P1 → Sending prompt directly to LLM (no retrieval)', '');
+      const t0 = Date.now();
+      const r = await callLLM(provider, apiKey,
+        'You are a helpful assistant. Answer accurately and concisely.',
+        query
+      );
+      const lat = (Date.now() - t0) / 1000;
+      const tokens = r.inputTokens + r.outputTokens;
+      log(`P1 ← ${tokens} tokens, ${lat.toFixed(1)}s`, 'ok');
+      const pass = await judgeAnswer(provider, apiKey, query, r.text);
+      const bert = 0.0; // LLM-Only has no context to ground against
+      renderAnswer(1, r.text, tokens, lat, bert, pass, provider);
+      return { tokens, cost: (r.inputTokens/1000)*PRICING[provider].input + (r.outputTokens/1000)*PRICING[provider].output, bert };
+    } catch(e) {
+      log('P1 error: ' + e.message, 'err');
+      renderError(1, e.message);
+      return null;
+    }
+  })();
+
+  // ── Pipeline 2: Basic RAG (Vector Database Retrieval) ───────────────────
+  const p2 = (async () => {
+    try {
+      log('P2 → Calling local Python Backend (/api/basic_rag) for Vector Retrieval', '');
+      const t0 = Date.now();
+      
+      const res = await fetch('http://localhost:8000/api/basic_rag', {
+        method: 'POST',
+        headers: { 'Content-Type': 'application/json' },
+        body: JSON.stringify({ query: query, gemini_api_key: apiKey })
+      });
+      
+      if (!res.ok) {
+        const errTxt = await res.text();
+        throw new Error(`Backend Error ${res.status}: ${errTxt}`);
+      }
+      
+      const data = await res.json();
+      const finalAnswer = data.answer;
+      
+      const lat = (Date.now() - t0) / 1000;
+      
+      // We now have exact tokens from the backend!
+      const exactTokens = (data.input_tokens || 0) + (data.output_tokens || 0); 
+      
+      log(`P2 ← Backend returned Vector RAG response in ${lat.toFixed(1)}s`, 'ok');
+      
+      const pass = await judgeAnswer(provider, apiKey, query, finalAnswer);
+      const bert = data.bert_score || 0.0;
+      renderAnswer(2, finalAnswer, exactTokens, lat, bert, pass, provider);
+      
+      return { tokens: exactTokens, cost: ((data.input_tokens||0)/1000)*PRICING[provider].input + ((data.output_tokens||0)/1000)*PRICING[provider].output, bert };
+    } catch(err) {
+      log('P2 error: ' + err.message, 'err');
+      renderError(2, err.message);
+      return null;
+    }
+  })();
+
+  // ── Pipeline 3: GraphRAG via Python Backend ─────────────────────────────
+  const p3 = (async () => {
+    try {
+      log(`P3 → Calling local Python Backend (/api/rag) for TigerGraph + Gemini magic`, '');
+      const t0 = Date.now();
+      
+      // REPLACE 'https://YOUR-BACKEND.onrender.com' with your actual Render URL later!
+      const res = await fetch('http://localhost:8000/api/rag', {
+        method: 'POST',
+        headers: { 'Content-Type': 'application/json' },
+        body: JSON.stringify({ 
+            query: query, 
+            company: query,
+            tg_host: tgHost,
+            tg_token: tgToken,
+            tg_graph: tgGraph,
+            gemini_api_key: apiKey
+        }) // Passing all config to backend
+      });
+      
+      if (!res.ok) {
+        const errTxt = await res.text();
+        throw new Error(`Backend Error ${res.status}: ${errTxt}`);
+      }
+      
+      const data = await res.json();
+      const finalAnswer = data.answer || data.text || JSON.stringify(data);
+      
+      const lat = (Date.now() - t0) / 1000;
+      
+      // We now have exact tokens from the backend!
+      const exactTokens = (data.input_tokens || 0) + (data.output_tokens || 0); 
+      
+      log(`P3 ← Backend returned GraphRAG response in ${lat.toFixed(1)}s`, 'ok');
+      document.getElementById('hops3').innerHTML = `<div class="elabel">Retrieval</div><div class="eval-val">Graph · 2-hop</div>`;
+      
+      const pass = await judgeAnswer(provider, apiKey, query, finalAnswer);
+      const bert = data.bert_score || 0.0;
+      renderAnswer(3, finalAnswer, exactTokens, lat, bert, pass, provider);
+      
+      return { tokens: exactTokens, cost: ((data.input_tokens||0)/1000)*PRICING[provider].input + ((data.output_tokens||0)/1000)*PRICING[provider].output, bert };
+    } catch(err) {
+      log('P3 error: ' + err.message, 'err');
+      renderError(3, err.message);
+      return null;
+    }
+  })();
+
+
+
+  const [r1, r2, r3] = await Promise.all([p1, p2, p3]);
+
+  // ── Summary stats ─────────────────────────────────────────────────────────
+  if (r2 && r3) {
+    const reduction = Math.round((1 - r3.tokens / r2.tokens) * 100);
+    const saving = (r2.cost - r3.cost).toFixed(4);
+    const bertDelta = (r3.bert - r2.bert).toFixed(2);
+
+    document.getElementById('sum-baseline').textContent = r2.tokens.toLocaleString();
+    document.getElementById('sum-baseline-sub').textContent = '$' + r2.cost.toFixed(4) + ' / query';
+    document.getElementById('sum-reduction').textContent = (reduction > 0 ? '−' : '+') + Math.abs(reduction) + '%';
+    document.getElementById('sum-reduction-sub').textContent = `${r2.tokens.toLocaleString()} → ${r3.tokens.toLocaleString()} tokens`;
+    document.getElementById('sum-saving').textContent = '$' + saving;
+    document.getElementById('sum-saving-sub').textContent = 'per query vs Basic RAG';
+    document.getElementById('sum-accuracy').textContent = (parseFloat(bertDelta) >= 0 ? '+' : '') + bertDelta;
+    document.getElementById('sum-accuracy-sub').textContent = `GraphRAG BERTScore: ${r3.bert.toFixed(2)}`;
+    document.getElementById('summary').style.display = 'grid';
+
+    log(`Summary — token reduction: ${reduction}%, cost saving: $${saving}, BERT delta: ${bertDelta}`, 'ok');
+  }
+
+  btn.disabled = false;
+}
+</script>
+</body>
+</html>

inspect_db.py

@@ -0,0 +1,34 @@
+import chromadb
+import sys
+
+def search_db(query):
+    # Connect to the local database
+    client = chromadb.PersistentClient(path="./chroma_db")
+    collection = client.get_collection(name="sec_filings")
+    
+    print(f"\n--- Searching Vector DB for: '{query}' ---")
+    print("-" * 50)
+    
+    # Perform a similarity search
+    results = collection.query(
+        query_texts=[query],
+        n_results=2 # Get top 2 results
+    )
+    
+    chunks = results['documents'][0]
+    distances = results['distances'][0]
+    
+    for i, (chunk, dist) in enumerate(zip(chunks, distances)):
+        # Convert L2 distance to a simple similarity percentage
+        similarity = max(0.0, (1.0 - (dist / 2.0))) * 100
+        print(f"\nResult {i+1} (Similarity: {similarity:.1f}%):")
+        print(f"{chunk[:300]}...") # Print first 300 characters
+        
+    print("\n" + "-" * 50)
+
+if __name__ == "__main__":
+    if len(sys.argv) > 1:
+        search_db(" ".join(sys.argv[1:]))
+    else:
+        # Default test query
+        search_db("What are the main supply chain risks and disruptions?")

query.gsql

@@ -0,0 +1,52 @@
+USE GRAPH FinancialGraph
+
+CREATE OR REPLACE QUERY get_company_context(STRING company_name) FOR GRAPH FinancialGraph {
+    
+    // Data structures to hold our context
+    ListAccum<STRING> @risk_list;
+    ListAccum<STRING> @competitor_list;
+    ListAccum<STRING> @subsidiary_list;
+    ListAccum<STRING> @executive_list;
+    
+    // 1. Start at the target company
+    Start = {Entity.*};
+    TargetCompany = SELECT s FROM Start:s WHERE s.id == company_name;
+    
+    // 2. Find all direct Subsidiaries and Executives
+    DirectLinks = SELECT t FROM TargetCompany:s -(RELATIONSHIP:e)-> Entity:t
+                  WHERE e.rel_type == "OWNS" OR e.rel_type == "EMPLOYS"
+                  ACCUM 
+                      IF t.entity_type == "Subsidiary" THEN s.@subsidiary_list += t.id END,
+                      IF t.entity_type == "Executive" THEN s.@executive_list += t.id END;
+
+    // 3. Find Risk Factors facing this company
+    CompanyRisks = SELECT t FROM TargetCompany:s -(RELATIONSHIP:e)-> Entity:t
+                   WHERE e.rel_type == "FACES_RISK"
+                   ACCUM s.@risk_list += t.id;
+                   
+    // 4. Advanced Traversal: Find competitors who share the same risk factors
+    // This traverses backwards from the Risks to find other companies.
+    SharedRiskCompetitors = SELECT c FROM CompanyRisks:r -(reverse_RELATIONSHIP:e)-> Entity:c
+                            WHERE c.id != company_name AND e.rel_type == "FACES_RISK"
+                            ACCUM c.@risk_list += r.id; 
+                            
+    // 5. Direct Competitors
+    DirectCompetitors = SELECT t FROM TargetCompany:s -(RELATIONSHIP:e)-> Entity:t
+                        WHERE e.rel_type == "COMPETES_WITH"
+                        ACCUM s.@competitor_list += t.id;
+                        
+    // Print the highly contextualized results for the LLM
+    PRINT TargetCompany [
+        TargetCompany.id,
+        TargetCompany.@risk_list,
+        TargetCompany.@subsidiary_list,
+        TargetCompany.@executive_list,
+        TargetCompany.@competitor_list
+    ];
+    
+    // Print the 2-hop Shared Risk Competitors to give the LLM deep market insight
+    PRINT SharedRiskCompetitors [
+        SharedRiskCompetitors.id,
+        SharedRiskCompetitors.@risk_list
+    ];
+}

requirements.txt

@@ -0,0 +1,10 @@
+fastapi
+uvicorn
+pydantic
+google-genai>=0.1.2
+pyTigerGraph
+chromadb
+bert_score
+python-dotenv
+pandas
+datasets

schema_and_load.gsql

@@ -0,0 +1,33 @@
+// TigerGraph GSQL Schema & Bulk Loader for Financial Graph
+// Run this in GraphStudio or via GSQL shell.
+
+// 1. Drop existing graph if necessary (Warning: this deletes data)
+// DROP GRAPH FinancialGraph
+
+// 2. Define Schema
+// We use a generalized 'Entity' vertex and 'RELATIONSHIP' edge 
+// to dynamically handle the Gemini extraction output without strict schema limits.
+CREATE VERTEX Entity (PRIMARY_ID id STRING, entity_type STRING) WITH PRIMARY_ID_AS_ATTRIBUTE="true"
+CREATE DIRECTED EDGE RELATIONSHIP (FROM Entity, TO Entity, rel_type STRING)
+
+CREATE GRAPH FinancialGraph (Entity, RELATIONSHIP)
+
+// 3. Define Loading Job
+USE GRAPH FinancialGraph
+
+CREATE LOADING JOB load_financial_graph FOR GRAPH FinancialGraph {
+    // IMPORTANT: Update these paths to where you saved your vertices.csv and edges.csv
+    DEFINE FILENAME v_file = "ANY:/path/to/vertices.csv";
+    DEFINE FILENAME e_file = "ANY:/path/to/edges.csv";
+    
+    // Load Vertices
+    LOAD v_file TO VERTEX Entity VALUES($0, $1) 
+        USING SEPARATOR=",", HEADER="true", EOL="\n", QUOTE="double";
+        
+    // Load Edges
+    LOAD e_file TO EDGE RELATIONSHIP VALUES($0, $1, $2) 
+        USING SEPARATOR=",", HEADER="true", EOL="\n", QUOTE="double";
+}
+
+// 4. Run the Loading Job (Uncomment and run after updating paths above)
+// RUN LOADING JOB load_financial_graph

vertices.csv

@@ -0,0 +1,693 @@
+v_id,v_type
+TROPICAL SPORTSWEAR INTERNATIONAL CORP.,Company
+Savane International Corp.,Subsidiary
+"Sandler & Travis Trade Advisory Services, Inc.",Company
+DATAWATCH CORP,Company
+interest rate risk,RiskFactor
+foreign currency exchange rate risk,RiskFactor
+DATAWATCH CORP International Subsidiaries,Subsidiary
+"BEI Medical Systems Company, Inc.",Company
+Gyne-Tech,Company
+Valley Forge Scientific Corporation,Company
+GALEY & LORD INC.,Company
+"Burlington Industries, Inc.",Company
+"Galey & Lord Industries, Inc.",Subsidiary
+"G&L Service Company, North America, Inc.",Subsidiary
+"Dimmit Industries, S.A. de C.V.",Subsidiary
+Farah Incorporated,Company
+"Dominion Textile, Inc.",Company
+Swift Denim Group,Subsidiary
+Klopman International Group,Subsidiary
+"Swift Textiles Europe, Ltd.",Subsidiary
+"Polymer Group, Inc.",Company
+"DT Acquisition, Inc.",Subsidiary
+"competitive and economic factors in the textile, apparel and home furnishings markets",RiskFactor
+raw materials and other costs,RiskFactor
+weather-related delays,RiskFactor
+general economic conditions,RiskFactor
+Levi's,Company
+Haggar,Company
+Savane,Company
+Tropical Sportswear Int'l Corporation,Company
+Polo Ralph Lauren,Company
+Calvin Klein,Company
+"Gap, Inc.",Company
+Banana Republic,Company
+Polo Jeans Co.,Company
+Liz Claiborne,Company
+Tommy Hilfiger Corporation,Company
+H.D. Lee Co. Inc.,Company
+Aalfs Manufacturing Inc.,Company
+"Texas Apparel, Inc.",Company
+Border Corp.,Company
+"Land's End, Inc.",Company
+L.L. Bean Inc.,Company
+"Eddie Bauer, Inc.",Company
+J. Crew Group Inc.,Company
+Revelacion En El Vestir S.A.,Company
+Fairsound U.S.A. Corp.,Company
+Hoi Kosher Garment Fty. Ltd.,Company
+Luen Fung (Hop Kee) Garment,Company
+Hagale Industries Inc.,Company
+"Oxford Industries, Inc.",Company
+MEDIQ Incorporated,Company
+MQ Acquisition Corporation,Company
+"Bruckmann, Rosser, Sherrill & Co., L.P.",Company
+"MEDIQ/PRN Life Support Services, Inc.",Subsidiary
+Rotko Entities,Company
+"Ferrer Freeman Thompson & Co., LLC",Company
+"Galen Partners, III L.P.",Company
+"CH Industries, Inc.",Company
+"CH Medical, Inc.",Company
+"National Patient Care Systems, Inc.",Company
+InnoServ Technologies,Company
+SpectraCair,Company
+Cost Containment Pressures,RiskFactor
+IEC Electronics Corp.,Company
+"Calidad Electronics, Inc.",Subsidiary
+"IEC - Edinburg Texas, Inc.",Subsidiary
+"Accutek, Inc.",Subsidiary
+IEC Arab Alabama Inc.,Subsidiary
+IEC Electronics - Ireland Limited,Subsidiary
+Compaq Computer Corporation,Company
+"Matrox Graphics, Inc.",Company
+"Symbol Techologies, Inc.",Company
+Stratus Corporation,Company
+"Seagate Techologies,Inc",Company
+"MEDIQ/PRN Life Support Services, Inc.",Company
+MEDIQ Incorporated,Company
+"MEDIQ Management Services, Inc.",Subsidiary
+MQ Acquisition Corporation,Company
+"Bruckmann, Rosser, Sherrill & Co., L.P.",Company
+Ferrer Freeman Thompson & Co. LLC,Company
+"Galen Partners, III L.P.",Company
+SpectraCair,Subsidiary
+"CH Industries, Inc.",Company
+"CH Medical, Inc.",Subsidiary
+"National Patient Care Systems, Inc.",Company
+Increasing pressure to reduce operating costs and capital expenditures,RiskFactor
+Cost containment pressures on health care providers,RiskFactor
+Reduced reimbursement levels by Medicare and other third-party payors,RiskFactor
+Legislative proposals leading to health care consolidations and managed care acceleration,RiskFactor
+INTERSTATE JOHNSON LANE INC.,Company
+CapTrust,Subsidiary
+Highly Competitive Environment,RiskFactor
+Marketplace Trading Volumes Volatility,RiskFactor
+Interest Rate Conditions Volatility,RiskFactor
+Revenue and Earnings Variability,RiskFactor
+Significant Incremental Cash Requirements,RiskFactor
+"Regulatory Liquidity and Capital Requirements (SEC, CFTC, NYSE)",RiskFactor
+"Regulatory Liquidity and Capital Requirements (SEC, NASD)",RiskFactor
+Year 2000 Technological Challenge,RiskFactor
+Dependence on Third-Party Systems for Year 2000 Compliance,RiskFactor
+Inflationary Effects,RiskFactor
+"Flanigan's Enterprises, Inc.",Company
+"Flanigan's Management Services, Inc.",Subsidiary
+"Flanigan's Enterprises, Inc. of Georgia",Subsidiary
+Seventh Street Corp.,Subsidiary
+Big Daddy's #48 Inc.,Subsidiary
+"Flanigan's Enterprises, Inc. of Pa.",Subsidiary
+Bankruptcy,RiskFactor
+Chairman of the Board,Executive
+Subchapter S corporation (franchisee),Company
+TCI INTERNATIONAL INC.,Company
+Broadcast and Communications Division,Subsidiary
+Signal Processing Division,Subsidiary
+Test and Measurement Division,Subsidiary
+Hewlett Packard,Company
+Foreign Competitors,Company
+Potential Loss from Operations in Fiscal 1999,RiskFactor
+Contract Termination Risk,RiskFactor
+Reduced US Government Demand (Post-Cold War),RiskFactor
+"Declining Economies in Asia, Africa, Middle East",RiskFactor
+Strong US Dollar Impacting Competitiveness,RiskFactor
+Delay/Loss of Key Program Awards,RiskFactor
+Delays in System Acceptance due to Customer Turnover,RiskFactor
+Spartech Corporation,Company
+Spartech Plastics,Subsidiary
+Spartech Polycom,Subsidiary
+Spartech Color,Subsidiary
+Spartech Vy-Cal Plastics,Subsidiary
+GenPak,Subsidiary
+Hamelin Industries,Subsidiary
+Spartech Profiles,Subsidiary
+Product Components,Company
+Pawnee Industries,Company
+Portage Industries,Company
+Hamelin Group,Company
+Preferred Plastic Sheet Division of Echlin Inc.,Company
+Echlin Inc.,Company
+"Polycom Huntsman, Inc.",Company
+"Prismaplast Canada, Ltd.",Company
+"Anjac-Doron Plastics, Inc.",Company
+"CMG Information Services, Inc.",Company
+ADSmart Corporation,Subsidiary
+"Engage Technologies, Inc.",Subsidiary
+"Accipiter, Inc.",Subsidiary
+InfoMation Publishing Corporation,Subsidiary
+NaviSite Internet Services Corporation,Subsidiary
+Planet Direct Corporation,Subsidiary
+Password Internet Publishing Corporation,Subsidiary
+"Magnitude Network, LLC",Company
+"Open Market, Inc.",Company
+NaviNet,Subsidiary
+"CMG@Ventures I, LLC",Subsidiary
+"Blaxxun Interactive, Inc.",Company
+GeoCities,Company
+"Lycos, Inc.",Company
+Parable LLC,Company
+Vicinity Corporation,Company
+"CMG@Ventures II, LLC",Subsidiary
+Chemdex Corporation,Company
+Critical Path,Company
+"KOZ, Inc.",Company
+"Mother Nature's General Store, Inc.",Company
+"Reel.com, Inc.",Company
+"Sage Enterprises, Inc.",Company
+"Silknet Software, Inc.",Company
+"Softway Systems, Inc.",Company
+Speech Machines plc,Company
+TicketsLive Corporation,Company
+Universal Learning Technology,Company
+Visto Corporation,Company
+"Amazon.com, Inc.",Company
+Hollywood Entertainment Corporation,Company
+"CMG@Ventures III, LLC",Subsidiary
+"@Ventures III, L.P.",Company
+SalesLink Corporation,Subsidiary
+InSolutions Incorporated,Subsidiary
+"On-Demand Solutions, Inc.",Subsidiary
+Pacific Direct Marketing Corporation,Subsidiary
+CMG Direct Corporation,Subsidiary
+"Red Brick Systems, Inc.",Company
+Salex Holding Corporation,Company
+"Sun Associates, LLC",Company
+Betty Sun,Executive
+Pershing Sun,Executive
+competition in the automobile asset management business,RiskFactor
+inability to obtain additional financing,RiskFactor
+general business and economic conditions,RiskFactor
+PH&H Corp.,Company
+GE Capital,Company
+The Goodyear Tire and Rubber Company,Company
+Jiffy Lube,Company
+Virginia Surety,Company
+Franklin Receivables LLC,Company
+Franklin Auto Trust 1998-1,Company
+The Chase Manhattan Bank,Company
+IEC Electronics Corp.,Company
+"Calidad Electronics, Inc.",Subsidiary
+"IEC - Edinburg Texas, Inc.",Subsidiary
+"Accutek, Inc.",Subsidiary
+IEC Arab Alabama Inc.,Subsidiary
+IEC Electronics - Ireland Limited,Subsidiary
+Ohshima Manufacturing Limited,Company
+Compaq Computer Corporation,Company
+"Matrox Graphics, Inc.",Company
+"Symbol Techologies, Inc.",Company
+Stratus Corporation,Company
+"Seagate Techologies,Inc",Company
+"RICA FOODS, INC.",Company
+Pipasa,Subsidiary
+As de Oros,Subsidiary
+Mr. Calixto Chaves,Executive
+"Inversiones La Ribera, S.A.",Company
+"Comercial Angui, S.A.",Company
+Antonio Echeverria,Executive
+Exchange Rate Risk,RiskFactor
+Foreign Competition,RiskFactor
+Commodity Price Volatility,RiskFactor
+Restaurantes As de Oros,Company
+GLOBAL ENVIRONMENTAL CORP.,Company
+Danzer,Subsidiary
+Duncan-Smith Company,Company
+Fremont Financial Corporation,Company
+Weakened Financial Condition,RiskFactor
+Tighter Credit Terms,RiskFactor
+Need for Capital Infusion,RiskFactor
+Dependency on Future Success,RiskFactor
+Year 2000 Problem,RiskFactor
+Major Service Provider System Failure,RiskFactor
+MATLACK SYSTEMS INC.,Company
+Qualcomm,Company
+Comdata,Company
+Chemical Manufacturers Association,Company
+Competitive Labor Environment,RiskFactor
+Increased Environmental Compliance Costs,RiskFactor
+Legal Action Settlement,RiskFactor
+Potential Change in Ownership,RiskFactor
+High Debt Levels,RiskFactor
+Declining Demand for Transport,RiskFactor
+General Economic Conditions,RiskFactor
+Competitive Pressures,RiskFactor
+Market Demand Shifts,RiskFactor
+Industry Performance and Needs,RiskFactor
+Equipment Utilization Challenges,RiskFactor
+New Service Development Risk,RiskFactor
+Increased Labor Costs,RiskFactor
+Increased Operating Costs,RiskFactor
+Litigation Uncertainties,RiskFactor
+Financing Risk,RiskFactor
+Insurance Availability Risk,RiskFactor
+Project Completion Risk,RiskFactor
+Management Retention Risk,RiskFactor
+Regulatory Changes,RiskFactor
+Y2K Compliance Risk,RiskFactor
+Satellite System Failure,RiskFactor
+Business Partner Y2K Compliance Risk,RiskFactor
+First Golden American Life Insurance Co of New York,Subsidiary
+Equitable of Iowa Companies,Company
+"PFHI Holdings, Inc.",Subsidiary
+ING Groep N.V.,Company
+Equitable Life Insurance Company of Iowa,Subsidiary
+Golden American,Subsidiary
+"Locust Street Securities, Inc.",Subsidiary
+"Equitable Investment Services, Inc.",Subsidiary
+"Directed Services, Inc.",Subsidiary
+Equitable of Iowa Companies Capital Trust,Subsidiary
+Equitable of Iowa Companies Capital Trust II,Subsidiary
+"Equitable of Iowa Securities Network, Inc.",Subsidiary
+"Equitable of Iowa Companies, Inc.",Subsidiary
+Discontinuation of sales relationship by significant distributor,RiskFactor
+Changes in acquisition cost allocation,RiskFactor
+Impairment of goodwill,RiskFactor
+Variability in amortization based on allocation and assumptions,RiskFactor
+Risk of loss from below investment grade securities,RiskFactor
+Sensitivity of below investment grade securities to adverse economic conditions,RiskFactor
+Future write-downs of securities,RiskFactor
+Material adverse effect on net income from significant write-downs,RiskFactor
+IEC Electronics Corp.,Company
+"IEC - Edinburg Texas, Inc.",Subsidiary
+IEC Arab Alabama Inc.,Subsidiary
+IEC Electronics - Ireland Limited,Subsidiary
+Compaq Computer Corporation,Company
+"Matrox Graphics, Inc.",Company
+"Symbol Techologies, Inc.",Company
+Stratus Corporation,Company
+"Seagate Techologies,Inc",Company
+Ohshima Manufacturing Limited,Company
+RESOURCE AMERICA INC.,Company
+Resource Asset Investment Trust,Company
+"The Atlas Group, Inc.",Subsidiary
+"Hynes & Howes Insurance Counselors, Inc.",Company
+Highly Competitive Market,RiskFactor
+Ralph Parry Insurance Agency Ltd.,Company
+Davenport Agency,Subsidiary
+Viking Agency,Company
+"Hansen and Hansen, Inc.",Company
+"R.A.P. Enterprises, Inc.",Company
+Ralph A. Parry,Executive
+Low Liquidity,RiskFactor
+Decreased Operating Income,RiskFactor
+VSI HOLDINGS INC.,Company
+The Performance Systems Group,Subsidiary
+Martin S. Suchik,Executive
+"Advanced Animations, Inc.",Subsidiary
+"Steve Toth, Jr.",Executive
+"BKNT Retail Stores, Inc.",Subsidiary
+"Vispac, Inc.",Subsidiary
+"Visual Services, Inc.",Subsidiary
+Ford Motor Company,Company
+General Motors,Company
+Nissan,Company
+Difficulty in recruiting qualified employees,RiskFactor
+Loss of major customers,RiskFactor
+Uncertainty of continued customer relationships,RiskFactor
+Irvine Sensors Corporation,Company
+"Novalog, Inc.",Subsidiary
+"MicroSensors, Inc.",Subsidiary
+"Imagek, Inc.",Subsidiary
+IBM,Company
+Infrared Data Association,Company
+SRI Consulting,Company
+"EG&G Astrophysics, Inc.",Company
+Boeing Corporation,Company
+R & D Leasing Ltd.,Company
+Uncertainty of Imagek financing,RiskFactor
+Uncertainty of widespread market success for stacked memory products,RiskFactor
+Cost disadvantage due to unestablished high volume production costs,RiskFactor
+Uncertainty of neural network chip set development success,RiskFactor
+Uncertainty of NeoStack commercialization timeline,RiskFactor
+"General DataComm Industries, Inc.",Company
+VITAL Network Services LLC,Subsidiary
+Bell Canada,Company
+GTE,Company
+Cignal Global Communications,Company
+AT&T,Company
+MCI,Company
+Sprint,Company
+American Airlines,Company
+Citicorp,Company
+EDS,Company
+Cemex,Company
+Chicago Board of Trade,Company
+Impsat,Company
+Telefonos de Mexico,Company
+France Telecom,Company
+Deutsche Telekom,Company
+Lucent Technologies,Company
+LM Ericsson,Company
+Austrian PTT,Company
+Mayo Clinic,Company
+Burlington Northern Santa Fe,Company
+Netherlands PTT,Company
+Telefonica de Espana,Company
+Telia,Company
+RADIATION DISPOSAL SYSTEMS INC.,Company
+Manuel E. Kane,Executive
+Albert D. Kane,Executive
+Ultrox International,Company
+Para Oxidation,Company
+PCI Ozone Corporation,Company
+"Griffin Technics, Inc.",Company
+Henkel Corporation,Company
+"Brown-Bovari, Inc.",Company
+"Steven M. Kane, Inc.",Company
+Steven Kane,Executive
+"Pillar Technologies, Inc.",Company
+"Gram Research & Development Co., Inc.",Company
+Sherman T. Mayne,Executive
+Inability to Market Products,RiskFactor
+Lack of Significant Revenues,RiskFactor
+Cessation of Operations,RiskFactor
+Lack of Working Capital and Revenue Prospects,RiskFactor
+Uncertainty of Future Operations,RiskFactor
+Limited Market Acceptance of Ozone Technology,RiskFactor
+Undeveloped Market for Waste Treatment Technology,RiskFactor
+Competition from Established Firms,RiskFactor
+Negligible Market Position,RiskFactor
+Ineffective Sales Agreements,RiskFactor
+Inactive Marketing Efforts,RiskFactor
+Insufficient Production Personnel,RiskFactor
+Supplier Dependence and Production Delays,RiskFactor
+Single Source Supplier Dependence (Pillar),RiskFactor
+Lapse of Patents due to Insufficient Funds,RiskFactor
+Burdensome Environmental Regulations,RiskFactor
+High Cost of Radioactive Waste Disposal,RiskFactor
+Uneconomical Process,RiskFactor
+Costly Regulatory Modifications and Insufficient Funds,RiskFactor
+Future Adverse Government Regulations,RiskFactor
+"Royal Palm Beach Colony, Limited Partnership",Company
+"Royal Palm Beach Colony, Inc.",Company
+"Regency Homes, Inc.",Company
+RTL Realty,Company
+"Stein Management Company, Inc.",Company
+Hasam Realty Limited Partnership,Company
+Union Bank,Company
+Randy Rieger,Executive
+Jack Friedland,Executive
+Uncertainty of future cash distribution,RiskFactor
+Reduction in sales revenues,RiskFactor
+Cash requirements for land development,RiskFactor
+Cash expenditures for transactions and operations,RiskFactor
+Bank financing repayment obligations,RiskFactor
+Increased cost due to regulatory changes,RiskFactor
+Highly competitive real estate market,RiskFactor
+Competition affecting property sales,RiskFactor
+Cyclical real estate market,RiskFactor
+Interest rate increases,RiskFactor
+Uncertain South Florida real estate market conditions,RiskFactor
+Competition from other developers in South Florida,RiskFactor
+Litigation over building permits,RiskFactor
+Contingent land sales,RiskFactor
+"Mesa Air Group, Inc.",Company
+"Mesa Airlines, Inc.",Subsidiary
+"America West Airlines, Inc.",Company
+"USAirways, Inc.",Company
+"Air Midwest, Inc.",Subsidiary
+"WestAir Holdings, Inc.",Subsidiary
+"WestAir Commuter Airlines, Inc.",Subsidiary
+"MPD, Inc.",Subsidiary
+"Regional Aircraft Services, Inc.",Subsidiary
+"FCA, Inc.",Subsidiary
+"Mesa Leasing, Inc.",Subsidiary
+"MAGI Insurance, Ltd.",Subsidiary
+"CCAIR, Inc.",Company
+Termination of Code-Sharing Agreements,RiskFactor
+Great Lakes Aviation,Company
+SkyWest Airlines,Company
+United Airlines,Company
+"Tidel Technologies, Inc.",Company
+"Tidel Engineering, Inc.",Company
+3CI Complete Compliance Corporation,Subsidiary
+"Diebold, Incorporated",Company
+NCR Corporation,Company
+"Triton Systems, Inc.",Company
+Fujitsu Corporation,Company
+Siemens-Nixdorf,Company
+NKL Industries,Company
+McGunn Safe Company,Company
+Armor Safe Company,Company
+AutoVend,Company
+Inadequate Patent Protection,RiskFactor
+Patent Challenge Risk,RiskFactor
+Patent Infringement Risk,RiskFactor
+Third-Party Patent Rights Impact,RiskFactor
+Competitor Patent Acquisition,RiskFactor
+Trade Secret Loss,RiskFactor
+3CI Complete Compliance Corporation,Company
+"Springhill, Louisiana Facility",Company
+"Birmingham, Alabama Location",Company
+"Austin, Texas Facility",Company
+"City of Carthage, Texas",Company
+"City of Center, Texas",Company
+Technological Obsolescence,RiskFactor
+"Diasense, Inc.",Company
+"Biocontrol Technology, Inc.",Company
+U.S. Food and Drug Administration,Company
+Battelle Memorial Institute,Company
+TUV Rheinland,Company
+Children's Hospital,Company
+Rush North Shore,Company
+Westmoreland Hospital,Company
+Lutheran General Hospital,Company
+Holy Family Hospital,Company
+Indiana Hospital,Company
+European Community,Company
+European Committee for Standardization,Company
+Delays in FDA Marketing Approval,RiskFactor
+Uncertainty of Future Funding,RiskFactor
+"Reliance on Biocontrol Technology, Inc.",RiskFactor
+Competition and Product Obsolescence,RiskFactor
+Dependence on Single Technology,RiskFactor
+Continued Operating Losses and Negative Net Worth,RiskFactor
+Potential Conflicts of Interest,RiskFactor
+"Risks to Patents, Trademarks, and Licenses",RiskFactor
+Uncertainty of Third-Party Payor Reimbursement,RiskFactor
+"Limited Sales, Marketing, and Manufacturing Experience",RiskFactor
+Difficulty Attracting and Retaining Key Employees,RiskFactor
+Product Liability,RiskFactor
+Uncertain Outcome of Pending Lawsuits,RiskFactor
+Absence of Public Market for Common Stock,RiskFactor
+Control by Existing Shareholders,RiskFactor
+Dilution of Common Stock,RiskFactor
+Limited Marketability of Diasensorr 1000,RiskFactor
+Cash Flow Problems,RiskFactor
+Slowed Testing Pace,RiskFactor
+Necessity of FDA Approval for US Market,RiskFactor
+Regulatory Process for Diasensorr 2000,RiskFactor
+High Sales Price Limiting Sales,RiskFactor
+STARCRAFT CORP /IN/,Company
+"Imperial Automotive Group, Inc.",Subsidiary
+National Mobility Corporation,Subsidiary
+"Tecstar, Inc.",Company
+"Starcraft Automotive Group, Inc.",Subsidiary
+"Imperial Industries, Inc.",Company
+Star Tank Company,Company
+Starcraft Van Conversions Corporation,Company
+GMC Truck,Company
+Chevrolet,Company
+Dodge,Company
+Ford,Company
+General Motors,Company
+"Mitsui and Co. (U.S.A.), Inc.",Company
+Increased popularity of SUVs and factory minivans,RiskFactor
+Price pressure from higher chassis costs,RiskFactor
+Lower conversion inventory on dealer lots,RiskFactor
+Fewer automotive dealers selling conversion products,RiskFactor
+Changing level of dealer support,RiskFactor
+Strengthening of the U.S. dollar,RiskFactor
+Financial turmoil in Asian markets,RiskFactor
+Cyclical nature of the conversion industry,RiskFactor
+Consumer preferences for sport utility vehicles,RiskFactor
+Level of disposable consumer income,RiskFactor
+Availability and price of gasoline,RiskFactor
+Level of interest rates and availability of consumer financing,RiskFactor
+Decline in general economic conditions or consumer confidence,RiskFactor
+Dependence on OEMs for vehicle chassis,RiskFactor
+Decline in the conversion industry and Company sales,RiskFactor
+"CMG Information Services, Inc.",Company
+ADSmart Corporation,Subsidiary
+"Engage Technologies, Inc.",Subsidiary
+"Accipiter, Inc.",Subsidiary
+InfoMation Publishing Corporation,Subsidiary
+NaviSite Internet Services Corporation,Subsidiary
+Planet Direct Corporation,Subsidiary
+Password Internet Publishing Corporation,Subsidiary
+"Magnitude Network, LLC",Company
+"Open Market, Inc.",Company
+"CMG@Ventures I, LLC",Subsidiary
+"Blaxxun Interactive, Inc.",Company
+GeoCities,Company
+"Lycos, Inc.",Company
+Parable LLC,Company
+Vicinity Corporation,Company
+"CMG@Ventures II, LLC",Subsidiary
+Chemdex Corporation,Company
+Critical Path,Company
+"KOZ, Inc.",Company
+"Mother Nature's General Store, Inc.",Company
+"Reel.com, Inc.",Company
+"Sage Enterprises, Inc.",Company
+"Silknet Software, Inc.",Company
+"Softway Systems, Inc.",Company
+Speech Machines plc,Company
+TicketsLive Corporation,Company
+Universal Learning Technology,Company
+Visto Corporation,Company
+"CMG@Ventures III, LLC",Subsidiary
+"@Ventures III, L.P.",Company
+SalesLink Corporation,Subsidiary
+InSolutions Incorporated,Subsidiary
+"On-Demand Solutions, Inc.",Subsidiary
+Pacific Direct Marketing Corporation,Subsidiary
+CMG Direct Corporation,Subsidiary
+NaviNet,Subsidiary
+Covol Technologies Inc.,Company
+"Cynsulo, Inc.",Company
+McParkland Corporation,Company
+"McParkland Properties, Inc.",Company
+"Riverbed Enterprises, Inc.",Company
+Enviro-Fuels Technology,Company
+Environmental Technologies Group International,Company
+Dow Chemical Company,Company
+Inability to implement applications profitably,RiskFactor
+Inability to utilize tax benefits,RiskFactor
+Feedstock supply risk,RiskFactor
+Production quality risk,RiskFactor
+Product marketing risk,RiskFactor
+"Lynton Group, Inc.",Company
+"Ramapo Helicopters, Inc.",Subsidiary
+"Lynton Aircraft Maintenance Services, Inc.",Subsidiary
+"Rockland Aviation, Inc.",Subsidiary
+"Lynton Aviation Charter, Inc.",Subsidiary
+"LynStar Aviation, Inc.",Subsidiary
+"LynStar Holdings, Inc.",Company
+Lynton Group Limited,Subsidiary
+Lynton Aviation Limited,Subsidiary
+European Helicopters Limited,Subsidiary
+Magec Aviation Limited,Subsidiary
+Air Hanson Limited,Subsidiary
+"Lynton Jet Centre, Inc.",Subsidiary
+"Lynton Aviation, Inc.",Subsidiary
+"Lynton Aviation Services, Inc.",Subsidiary
+Dollar Air Services Limited,Subsidiary
+Black Isle Helicopters Limited,Subsidiary
+"Lynton Properties, Inc.",Subsidiary
+PLM Dollar Group Limited,Company
+Air Hanson Engineering Limited,Subsidiary
+Air Hanson Aircraft Sales Limited,Subsidiary
+Loss of CAA/Manufacturer Certificates/Licenses for EHL/Air Hanson Engineering,RiskFactor
+Loss of CAA/Manufacturer Certificates/Licenses for Magec,RiskFactor
+"General Automation, Inc.",Company
+General Automation LLC,Subsidiary
+General Automation Canada Subsidiary,Subsidiary
+General Automation Australia Subsidiary,Subsidiary
+General Automation UK Subsidiary,Subsidiary
+"Texas Micro, Inc.",Company
+"Sequoia Systems, Inc.",Company
+Boundless Technologies,Company
+SunRiver Data Systems,Company
+Comerica Bank,Company
+Wang,Company
+AT&T,Company
+IBM,Company
+Data General Corporation,Company
+Digital Equipment Corporation,Company
+Hewlett Packard Company,Company
+Ardent,Company
+Pick Systems,Company
+"Microsoft, Inc.",Company
+Going Concern Qualification,RiskFactor
+Recent Operating Losses,RiskFactor
+Deficit Net Worth,RiskFactor
+Working Capital Deficiency,RiskFactor
+Cash Flow Constraints,RiskFactor
+Obligations to TMI in Default,RiskFactor
+Obligations to Boundless in Default,RiskFactor
+Limited Market for the Company's Stock,RiskFactor
+Competition Risk,RiskFactor
+RCM TECHNOLOGIES INC.,Company
+The Consortium,Subsidiary
+"The Consortium of Maryland, Inc.",Subsidiary
+"Performance Staffing, Inc.",Subsidiary
+"Programming Alternatives of Minnesota, Inc.",Subsidiary
+"Biomune Systems, Inc.",Company
+"Optim Nutrition, Inc.",Subsidiary
+"Rockwood Companies, LLC",Subsidiary
+Bioxide Corporation,Company
+"Frank A. Eldredge, Ph.D.",Executive
+"David O. Lucas, Ph.D.",Executive
+"Craig D. Moffat, M.D.",Executive
+Clinimetric Research Associates,Company
+St. Francis Memorial Hospital,Company
+East Bay AIDS Clinic,Company
+St. Lukes-Roosevelt Hospital Center,Company
+Associates in Medical and Mental Health,Company
+Limited access to funding for future research and development,RiskFactor
+Highly competitive market for nutrition supplements,RiskFactor
+Limited financial resources,RiskFactor
+Risk of reduced ownership interest in Rockwood due to unfulfilled loan covenants,RiskFactor
+Inability to obtain additional financing,RiskFactor
+Inability to increase product sales,RiskFactor
+NORTH EUROPEAN OIL ROYALTY TRUST,Company
+Mobil Corp.,Company
+Exxon Corp.,Company
+Royal Dutch Group,Company
+North European Oil Company,Company
+North European Oil Corporation,Company
+John H. Van Kirk,Executive
+John R. Van Kirk,Executive
+BEB,Company
+Mobil (German Operating Company),Subsidiary
+German Subsidiaries of Exxon Corp.,Subsidiary
+German Subsidiaries of Royal Dutch Group,Subsidiary
+The Bank of New York,Company
+Deutsche Bank,Company
+Registrar and Transfer Company,Company
+Termination of Royalty Rights,RiskFactor
+Speculative Business,RiskFactor
+Uncertainty of Future Income,RiskFactor
+Fluctuations in Gas and Oil Prices/Supplies,RiskFactor
+Environmental Regulations Impact,RiskFactor
+Unforeseeable Economic and Political Factors,RiskFactor
+Currency Exchange Rate Fluctuations,RiskFactor
+Y2K Computer Compliance Issues,RiskFactor
+Celsion Corporation,Company
+Massachusetts Institute of Technology,Company
+Food and Drug Administration,Company
+Massachusetts General Hospital,Company
+Oxford University,Company
+Hammersmith Hospital,Company
+Dr. Gerald Wolf,Executive
+Columbia HCA's JFK Hospital,Company
+Duke University,Company
+"MMTC, Inc.",Company
+American Cancer Society,Company
+Difficulty delivering controlled heat without burning healthy tissue,RiskFactor
+Need for new FDA indication for breast tumor ablation,RiskFactor
+Pentech International Inc.,Company
+Sawdust Pencil Co.,Subsidiary
+Bic Pen Company,Company
+Papermate,Company
+"RoseArt Industries, Inc.",Company
+Newell,Company
+Product Shortages,RiskFactor
+Dumping Duties,RiskFactor
+Increased Raw Material Costs,RiskFactor
+Increased Dumping Duties,RiskFactor
+Minimum Royalty Payments Leading to Losses,RiskFactor
+Loss of Major Customer,RiskFactor
+Bad Debts from Customers,RiskFactor
+Failure to Register Trademarks,RiskFactor
+Intense Competition,RiskFactor