WIP

Author: habedi

python/fraud/README.md

@@ -4,9 +4,11 @@ This example detects suspicious patterns and fraudulent behaviors in a real-time
 
 ### How It Works
 
-1. Generates a continuous stream of financial transactions, accounts, and registration events.
-2. Ingests the stream into IssunDB; map accounts, transactions, and devices as nodes and edges.
-3. Executes Cypher queries concurrently over the database to identify circular transfers, shared devices, and stolen credentials.
+1. Generates a deterministic, (seeded) stream of financial events (that includes account, device, and merchant registrations, transfers, payments, and
+   device logins information).
+2. Ingests the stream into IssunDB; maps accounts, devices, and merchants as nodes and transfers, payments, and logins as edges.
+3. Runs four Cypher-based detectors after every batch insert to flag things like circular transfer rings, shared devices, money-mule fan-in, and
+   velocity bursts.
 
 More detailed workflow is shown below:
 

python/recommendation/README.md

@@ -4,10 +4,12 @@ This example builds a social network graph and recommends content and friends to
 
 ### How It Works
 
-1. Creates a graph of users, interests, followed tags, and published posts in IssunDB.
-2. Computes text embeddings for posts and interests so we can run vector-based similarity checks on them.
-3. Combines graph traversal (friend-of-a-friend collaborative filtering) and vector search (semantic similarity) to recommend relevant articles and
-   new connections.
+1. Creates a graph of users, topics, and posts in IssunDB, with `FOLLOWS`, `POSTED`, `ABOUT`, and `LIKES` edges; each user has topic affinities
+   that drive their interest vector.
+2. Computes interest-vector embeddings for users and posts and builds a full-text index over post text, so the graph supports both semantic and
+   keyword search.
+3. Provides four recommendation features, including friend-of-friend suggestions through Cypher, kindred users and posts through vector search,
+   trending topics through Cypher aggregation over recent likes, and a hybrid discover feed that fuses vector, text, and one-hop graph expansion.
 
 More detailed workflow is shown below:
 

rust/code-explorer/README.md

@@ -5,8 +5,10 @@ This example indexes Rust source code, builds a syntax dependency graph, and ran
 ### How It Works
 
 1. Parses Rust source files into abstract syntax trees (using the `syn` library).
-2. Stores syntax elements (like modules, functions, and structs) as nodes, and dependencies (like calls, imports, and definitions) as edges in IssunDB.
-3. Computes the structural importance of functions using the PageRank algorithm on the constructed graph.
+2. Stores files, functions, structs, enums, and traits as nodes, with `CONTAINS`, `CALLS`, `METHOD_OF`, and `IMPLEMENTS` edges between them.
+3. Answers structural questions over the graph, including callers and callees through Cypher, dead-code candidates, transitive impact through native
+   adjacency traversal, the shortest call path between two functions, weakly connected components, and cycle detection.
+4. Ranks functions by structural importance with PageRank over the code graph.
 
 More detailed workflow is shown below:
 

rust/graphrag-agent/README.md

@@ -5,7 +5,7 @@ This example implements a knowledge graph extraction and question-answering pipe
 ### How It Works
 
 1. Reads text documents, splits them into text chunks, extracts entities, and records entity co-occurrences.
-2. Indexes chunks using a vector index (for semantic retrieval) and an BM25 text index (for keyword retrieval).
+2. Indexes chunks using a vector index (for semantic retrieval) and a BM25 text index (for keyword retrieval).
 3. Performs a hybrid search on both indexes and resolves the reciprocal-rank fusion of the results.
 4. Traverses the graph structure to extract adjacent entities and document contexts.
 5. Combines the retrieved context and queries an LLM to generate the final answer.