P33: LLM Foundation — Ollama, pgvector, embedding index

New Issue

Scaffold the llm module and stand up local inference + vector storage: Ollama compose service (nvidia runtime), pgvector on the existing Postgres, markdown-aware chunker, and an incremental embedding indexer over the 164k regulations.gov comments and the Zotero/bib corpus, with multi-host Ollama dispatch across the 3060/4090/5080 pool. Spec: docs/superpowers/specs/2026-07-16-llm-module-design.md

No due date

100% Completed

llm: rebuild postgres pgvector without AVX-512 — HNSW build SIGILLs the shared server on Zen2 infra llm

#580 by kert was closed 2026-07-17 20:13:45 +00:00 0 / 3

llm: multi-host Ollama dispatch (3060/4090/5080) infra llm

#568 by kert was closed 2026-07-17 19:37:21 +00:00 0 / 4

llm: incremental embedding indexer data llm

#567 by kert was closed 2026-07-17 19:37:21 +00:00 0 / 4

llm: chunker for comment extractions + corpus docs llm

#566 by kert was closed 2026-07-17 19:37:21 +00:00 0 / 4

llm: pgvector schema + migrations infra llm

#565 by kert was closed 2026-07-17 19:37:20 +00:00 0 / 4

llm: ollama compose service + model bake-off infra llm

#564 by kert was closed 2026-07-17 19:37:20 +00:00 0 / 4

llm: module scaffold + stack[llm] extra llm

#563 by kert was closed 2026-07-17 19:37:20 +00:00 0 / 4