nx_research_grow.nx

nx_research_grow.nx -- THE SELF-GROWING RESEARCHER INDEX (R4 self-grow loop) WITH CONTENT-DEDUP + CROSS-TOPIC FLAGGING. Each run indexes the DELTA (fetched docs not yet SEEN) into a NEW append-only SHARD, but FIRST fingerprints each doc's CONTENT (FNV-1a 64-bit). If that content fingerprint is already registered, the doc is NOT re-indexed (no duplicate as coverage grows) -- instead the registry records the new (topic,path) for that fingerprint, and if the new topic differs from the canonical topic we EMIT A CROSS-TOPIC FLAG: "this information serves two research topics, found in the same place." So the corpus compounds on every use, never duplicates content, never loses data, and the structure intelligently links topics that share sources. Bounded batch per run = OOM-safe (no sys_munmap). Reuses the FIXED nx_inv build + nx_dirent walk. Topic = the filename prefix before the first '_' (perma_/sem_/food_/ng_/...). expect_exit: 0 license_tier: ORIGINAL