Agent Claude Code
Run a live smoke test of the /ask endpoint (SSE-streamed RAG). Boots fireseqsearchserver via tests/runlogseq.sh, runs tests/testask.py (protocol/invariant assertions) and tests/testendpoints.py --ask against a user-supplied question, and reports on answer grounding, citation validity, source quality, streaming…
Agent Claude Code
Run a live smoke test of fireseqsearchserver. Boots the server via tests/runlogseq.sh, runs tests/testendpoints.py against a user-supplied query, and reports on snippet quality, score distribution, summary status, and any errors in the log. Use when the user wants to validate query behavior end-to-end.
Agent Claude Code
A/B-compare two chat models on fireseqsearchserver's two LLM-driven endpoints — /ask (multi-source synthesis, streamed) and /highlight (single-source extraction, one-shot). Boots the server twice (once per model) reusing the same index and summary cache so retrieval is identical; for each test question runs /ask, then…
Agent Claude Code
Run a live smoke test of fireseqsearchserver against an Obsidian vault. Drives tests/runsmoke.sh in one of two modes — lite (committed astro-wiki-lite fixture; fast, proves the plumbing) or full (the real 366-note AstroWiki2.0 vault; the only mode that can grade whether score priority and /ask answers are correct).…
Agent Claude Code
Audit the chunker's stripping quality on the live Logseq corpus. Boots fireseqsearchserver with FIRESEQDUMPPROCESSEDDIR set so every indexed note is mirrored to a dump tree containing RAW / PREPROCESSED / CHUNKS sections. Samples N notes and reports two things per file — lingering noise (Logseq syntax that should have…
Instructions file
Instructions for Endle/fireSeqSearch, covering fireseqsearch, surface area, locked decisions — don't relitigate without strong evidence, retrieval and chunking (flavour-aware via --notebook obsidian).