retoorandClaude Sonnet 5
9dbd1fe667
- Plexer: always-active terminal multiplexing. NativePlexerController/ TmuxPlexerController share one interface, chosen by which env var is already set (TAI_PLEXER_SOCKET vs TMUX_PANE); NativePlexerBackend is the pty.fork()-based wrapper (raw mode, SIGWINCH resize forwarding, a Unix-socket control protocol, continuous session logging, a Ctrl+A prefix key distinct from tmux's own Ctrl+B). Auto-wraps new interactive shells via a .bashrc hook mirroring the existing command-not-found install pattern, gated on $TMUX/$TAI_PLEXER_ACTIVE/ a $HOME/.tai-no-plexer bypass. New plexer_send/read/type/status tools, env-gated on a detected controller. - tests/container.py + docker-compose.yml + tests/docker/Dockerfile: a disposable podman/docker image for the tests that must never touch a real machine (installing the plexer hook into a real ~/.bashrc, a live tmux session). Required suite (test_seal.py, test_tai.py, tests/plexer.py, tests/bashrc.py) gates the result; tests/rag.py runs as best-effort since its real-LLM dependency can't be satisfied inside a disposable container with no nested engine. - tests/bashrc.py: real end-to-end hook install + bash -i auto-wrap, anti-recursion, tmux-active skip, non-interactive skip, a live TmuxPlexerController test, and an install/uninstall round-trip. - tests/plexer.py: NativePlexerBackend under a real pty, plus a dedicated robustness pass (file permissions, >64KB payload framing, a malformed control request, a genuine self-deadlock fix in the write-to-child path). - tai_test.py: the shared async process-driving framework (pipe/pty transport, group-wide kill-and-reap, a StdinBridge for --interactive) both test files build on. - ChatClient last-resort backend: when there's no local opencode CLI but the sandbox box is already running, install opencode into it (official Node tarball, not apt -- apt's nodejs/npm pull in webpack/babel/jest as transitive deps) and route through it instead of giving up. Only reachable after the first retry round already failed everything else; never creates the box itself. TmuxPlexerController.status() fixed to use `list-panes` instead of `display-message -p`, which silently fell back to the default pane instead of erroring against a dead target. - Fix 4 pre-existing test_tai.py bugs surfaced by running it in a clean environment for the first time: two stale mock signatures (subprocess.run's stdout= kwarg, Chat.complete's deadline= kwarg), one $TMUX-dependent catalog assertion, one tty-dependent color assertion. - research/: a sourced research archive (Source Dossier/Citation Card/Conclusion tiers) backing the above design decisions, with its own audit.py consistency checker. - .gitignore: fix a blanket *.md rule that was about to silently drop the entire research/ archive from version control; add __pycache__/ and the one-off debug transcript this session started from. Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
3.1 KiB
3.1 KiB
Reciprocal Rank Fusion: How It Works and When to Use It
Metadata
| Field | Value |
|---|---|
| Institute | Molodetz Institute — Research & Retrieval Division |
| Subject | sqlite-rag-hybrid-search |
| Document type | Source Dossier |
| Tier | 1 |
| Source title | Reciprocal Rank Fusion: how it works and when to use it |
| Source URL | https://bigdataboutique.com/blog/reciprocal-rank-fusion-how-it-works-and-when-to-use-it |
| Source publisher / author | BigData Boutique (consultancy blog) |
| Source type | blog post |
| Source date (published) | Unknown — not stated |
| Date researched | 2026-10-09 |
| Researched by | tai (agent), model: Claude Sonnet 5, on behalf of retoor |
| Research method | web search synthesis |
| Confidence | Medium |
| Status | Active |
Summary
Explains why RRF, specifically, is the standard fusion mechanism for combining a BM25 ranking with a vector-similarity ranking: the two score types (an unbounded positive BM25 score vs. a bounded cosine-similarity score) are not directly comparable, so fusing by rank position rather than raw score avoids an apples-to-oranges comparison. Cites the original Cormack, Clarke & Büttcher 2009 SIGIR paper, and states RRF is now the default hybrid-search fusion method in OpenSearch, Elasticsearch, Azure AI Search, MongoDB Atlas, and Weaviate.
Key findings
- The core justification for RRF over naive score-averaging: BM25 and cosine similarity live on fundamentally different scales, so only rank order is safely comparable between them.
- Names five major production search systems (OpenSearch, Elasticsearch, Azure AI Search, MongoDB Atlas, Weaviate) that ship RRF as their default hybrid-fusion method — this is a strong, checkable claim about industry adoption.
- Traces RRF's academic origin to Cormack, Clarke & Büttcher (SIGIR 2009), giving a verifiable citation trail rather than treating RRF as a folk technique.
Pros / strengths
- Strongest adoption evidence of any source in this subject — naming specific, well-known production systems makes this a falsifiable, checkable claim rather than a vague "widely used" assertion.
- Correctly identifies the mechanism of the problem RRF solves (incomparable score scales), not just that it's popular.
Cons / limitations / caveats
- A consultancy blog post, not the original academic paper — the Cormack/Clarke/Büttcher citation should be treated as secondhand attribution unless the original 2009 paper is independently pulled.
- Does not quantify how much better RRF performs versus alternatives in any benchmark; the adoption claim is about popularity, not about measured superiority.
Relevance to tai
Was the strongest single piece of evidence, during the now-superseded hybrid-search design phase, that RRF (not score-weighted fusion) would have been the correct choice had tai kept a vector layer. Retained for the record and in case a future, non-vector second ranked list ever needs fusing with BM25. See conclusions/no-vectors-bm25-chunked-rag.md.
Changelog
| Date | Change |
|---|---|
| 2026-10-09 | Initial dossier created. |