FramerD is a high-performance distributed object database optimized for storing and operating over large pointer-intensive data structures, especially those used to represent the semantic knowledge used in intelligent document analysis and search. Conventional databases based on relational algebra (any SQL-oriented product) are not optimized for pointer-intensive data, but many knowledge-based algorithms make heavy use of pointer-following and expansion over such data. FramerD uses an alternate model supporting billions of *unique objects* with lightweight pointers and a range of caching strategies for resolving object pointers to complex structured objects. This model is incorporated with the ability to manage very large dynamically modifiable *indices* that map arbitrary structured objects to collections of object references. The underlying content model is a dynamically typed data model based on associational structures for mapping property identifiers (which may themselves be object pointers) to values. These associational structures are called "frames" after the term classically used for such entities in artificial intelligence.