Cymatix Context

Coordinate-index engine for LLM agents. Retrieves, weighs, and compresses your codebase into a context window — without a single LLM call on the retrieval path.

One SQLite knowledge store, a seven-stage pipeline, and an explicit know / miss contract on every response. The engine's namesake cymatics stage — an MD5-binned 256-dimensional term spectrum — is a candidate-reordering signal that has not yet been isolated against hashed bag-of-words or random-bin controls; treat it as an experimental cheap feature, not a proven one.

Proof (30 seconds)

Token economics — compressor disabled (the default LLM-free config), N=15 query shapes, May 2026:

That denominator is a configurable modeled baseline, not a measured competitor run: top-5 × 1,500 + 500 overhead = 8,000 tokens. Reproducer: benchmarks/bench_rag_vs_sike_tokens.py, against your own store. The multi-turn session-delivery figures (~40% savings, 37× on repeated retrievals) are unverified design estimates pending the cymatix_session_tokens_saved_total counter.

Historical v0.9.1 defaults (released 2026-08-30) — 829K-fragment EnterpriseRAG-Bench bed, n=470, delivered basis, on a fully algorithmic retrieval path (dense, SPLADE and PKI default-off since 2026-08-15 / -16 / -17, each flip receipt-gated). Two defaults graduated in v0.9.1, each measured as a paired row: the wave-1 ranking flip (#407, rrf_k 60 → 20 plus all-classes eps_band combinators) moved gold-document delivery 0.555 → 0.630 and recall@12 0.651 → 0.681 with zero question-type regressions; the delivered-seat floor (#409, [budget] min_delivered_docs = 12) then moved delivery 0.630 → 0.668 (+18/−0) with the ranking bases byte-identical, confirmed on two more corpora with zero paired losses. Release gate: benchmarks/dogfood/receipts/sweep_v091_gate_2026-08-30.json (ALL PASS). These are retrieval-layer measurements, not end-to-end grades — the ERB judge protocol has not been re-run on these defaults.

Shipped defaults (v0.9.2, released 2026-09-08) — one ingestion default changes: [ingestion] entity_autolink_hub_cutoff = 200 excludes entities with more than 200 existing postings from auto-link probes during ingestion. Set it to 0 to restore legacy linking. The paired receipts and scope are documented in Configuration and the CHANGELOG.

0.9.0 shipped defaults, for reference — 829k bed, n=469: 56.5% gold-document delivery (265/469), recall@12 0.659 (benchmarks/dogfood/erb/receipts/sema_readgate_829k_n469.json). That is a different ledger row from the 0.9.1 lines above — a different bed build and ingest concurrency — so the two are not a before/after pair.

Methodology, the ERB correctness/delivery pair-quote rule, the sharded gap (#275), and the dense-off latency disclosure (×2.5–2.6 at 100k, shrinking at the 829k operating point; receipts in the CHANGELOG, #374) all live on Benchmarks and Receipts.

Get started

Python 3.11+. The core install is dependency-light (FastAPI + SQLite, no torch); extras add what you turn on — cpu (spaCy ingest tagging), embeddings (opt-in dense recall), mcp, ast, otel, launcher-tray, all. Full extras matrix, GPU detection, and three worked workflows: Getting Started · docs/SETUP.md.

pip install "cymatix-context[embeddings,cpu,mcp]"    # recommended working set
python -m spacy download en_core_web_sm              # ingest tagger model

cymatix ingest path/to/your/project/ --recursive     # 1. build the store
cymatix query "how does the splice step work?"       # 2. ask it — no server
cymatix packet "edit the splice step" --task-type edit --json   # 3. agent bundle
cymatix-server                                       # 4. proxy on 127.0.0.1:11437

Pipeline

Seven stages per turn, all LLM-free except the optional splice. Stage by stage, and where the model boundary actually sits: Pipeline.

  query
    ▼
  0. Classify   rule-based: decoder mode + assembly cap
    ▼
  1. Extract    heuristic keyword + entity extraction
    ▼
  2. Retrieve   FTS5 BM25 + tags (+ opt-in BGE-M3 dense) + synonym expansion
    │           + co-activation + SR + cymatics 256-bin spectrum scoring,
    ▼           ranked via RRF (default) or additive fusion
  3. Re-rank    CPU classifier scores (optional)
    ▼
  4. Splice     Headroom Kompress (CPU) or LLM compressor (optional)
    ▼
  5. Assemble   token budget + legibility headers (fired tiers, confidence
  + Stage 7     ◆/◇/⬦, compression ratio) + freshness gate (stale/cold/
    ▼           superseded → miss) + session delivery (elide seen docs)
  6. Persist    query+response → knowledge store (background)
    ▼
  know { } or miss { }

Surfaces

Three ways in, same retrieval primitives, same JSON shapes. Direct MCP needs neither the model proxy nor the tray — a healthy headless server is sufficient. Configuration lives in cymatix.toml; env vars use the CYMATIX_* prefix. Reference: CLI · HTTP API · MCP and IDE Integration · Configuration.

The know/miss contract

• know { found, confidence } — the context is grounded; the agent may answer.
• miss { reason, escalate_to } — don't answer from the knowledge store; escalate, or refetch from refresh_targets. The freshness gate downgrades stale / cold / superseded results into a miss.
• The shape is stable; the confidence is provisional. The contract shape is load-bearing, but the confidence scalar is under active recalibration and is not yet a reliable trust signal on current internal beds (#287, #239) — rely on found / reason. Full semantics: Agent Contract.

Gotchas

• Knowledge store path is genomes/main/genome.db, not the project root. Delete it to start fresh; it auto-creates on first use.
• The synonym map is critical. "No relevant context" usually means the query keywords don't map to the tags assigned at ingest — add them under [synonyms].
• Session delivery (session_delivery_enabled = true) elides already-delivered documents per session; the ~40% multi-turn saving is an unverified design estimate. Pass ignore_delivered: true in the /context body for benchmarks.
• Sharded scale gap. The sharded path trails the unsharded engine by ~31pp recall@10 / ~30pp MRR on the xl bed (#275) — prefer unsharded for accuracy-sensitive corpora.
• The agent prompt fragment is load-bearing. Without it, frontier models confabulate past a miss. Import cymatix_context.agent_prompt.full_fragment().

Observability

Optional Grafana/Tempo/Loki sidecar: scripts\setup-grafana-telem.ps1 (Windows) or scripts/setup-grafana-telem.sh (Linux/macOS), dashboards at localhost:3000. Full surface: docs/architecture/OBSERVABILITY.md · wiki: Observability.

Documentation

The wiki is the narrative documentation — 15 pages, Troubleshooting included, also rendered at https://cymatixcontext.com/wiki/. Questions, and arguments about the receipts: Discord.

Built on spaCy NER, SQLite FTS5 BM25, BGE-M3, Kompress, Headroom, and the Howard 2005 TCM / Stachenfeld 2017 SR literature — full attributions in NOTICE.

How this was built

Cymatix Context is architected and QA-directed by Michael Bachaud. Implementation, refactoring, draft documentation, and test generation are produced by AI coding agents under spec- and benchmark-gated review. The human owns the product thesis, architecture selection, acceptance criteria, experiment design, and falsification authority; the models own the code production.

License

Apache-2.0. See NOTICE for third-party attributions.