Helix 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.

Proof (30 seconds)

WIP benchmark numbers — compressor disabled (default LLM-free config), N=15 query shapes, May 2026:

With the optional compressor enabled (Claude Haiku splice), median improves to ~5×. In multi-turn sessions, the session delivery register elides already-seen documents — observed 37× reduction on repeated retrievals within a conversation.

Reproducer: python benchmarks/bench_rag_vs_sike_tokens.py against your own genome.

Agent contract: every /context response carries know { found, confidence } (grounded — you may answer) or miss { reason, escalate_to } (not found — don't answer from genome). Stale results downgrade to miss(reason="stale"|"cold"|"superseded") via the freshness gate.

Get started (60 seconds)

# 1. Install
pip install helix-context
python -m spacy download en_core_web_sm

# 2. Ingest your codebase
helix ingest path/to/your/project/ --recursive

# 3. Query it
helix query "how does the splice step work?"

# 4. Or start the proxy for IDE integration
helix-server   # binds to 127.0.0.1:11437

For extras matrix, BGE-M3 backfill, and tray setup: docs/SETUP.md.

Agent surfaces

Three ways to drive Helix — same retrieval primitives, same JSON shapes:

# CLI — no server, no daemon, subprocess-drivable
helix query    "what does the splice step do?" --json
helix packet   "edit the splice step" --task-type edit --json
helix gene get abc123 --json
helix neighbors "splice step" --k 10 --json
helix refresh-targets "edit the splice step" --json
helix status
helix diag corpus

Full CLI reference: docs/clients/cli.md. MCP tool schemas: docs/api/mcp-tools.md.

Pipeline (2 minutes)

Seven stages per turn, all LLM-free except optional splice:

  query
    │
    ▼
┌──────────────┐
│ 0. Classify  │  rule-based: decoder mode + assembly cap
└──────┬───────┘
       ▼
┌──────────────┐
│ 1. Extract   │  heuristic keyword + entity extraction
└──────┬───────┘
       ▼
┌──────────────┐  FTS5 BM25 + BGE-M3 dense (1024-dim) + tags
│ 2. Retrieve  │  + synonym expansion + co-activation + SR
│              │  ranked via RRF or additive fusion
└──────┬───────┘
       ▼
┌──────────────┐
│ 3. Re-rank   │  CPU classifier scores (optional)
└──────┬───────┘
       ▼
┌──────────────┐
│ 4. Splice    │  Headroom Kompress (CPU) or LLM compressor
└──────┬───────┘
       ▼
┌──────────────┐  token budget + legibility headers (fired tiers,
│ 5. Assemble  │  confidence ◆/◇/⬦, compression ratio) +
│   + Stage 7  │  freshness gate (stale/cold/superseded → miss)
└──────┬───────┘  + session delivery (elide already-seen docs)
       ▼
┌──────────────┐
│ 6. Persist   │  query+response → knowledge store (background)
└──────┘───────┘
       ▼
   know { } or miss { }

• know/miss contract: know means the context is grounded, agent may answer. miss means don't answer from genome — escalate via escalate_to tools or refetch from refresh_targets.
• Caller model class: /context accepts caller_model_class: "generic" | "small_moe" | "frontier" to select render branch (ordering, assembly cap, decoder mode). See docs/api/context-endpoint.md §7.

IDE + MCP integration

{
  "mcpServers": {
    "helix-context": {
      "command": "python",
      "args": ["-m", "helix_context.mcp_server"],
      "cwd": "/absolute/path/to/your/project",
      "env": { "HELIX_MCP_URL": "http://127.0.0.1:11437" }
    }
  }
}

models:
  - name: Helix (Local)
    provider: openai
    model: gemma3:e4b
    apiBase: http://127.0.0.1:11437/v1
    apiKey: EMPTY
    roles: [chat]
    defaultCompletionOptions:
      contextLength: 128000
      maxTokens: 4096

OPENAI_BASE_URL=http://localhost:11437/v1 your-app

Knowledge store management

[genome]
path = "genomes/main/genome.db"   # relative to helix run directory

Backup (safe while running — WAL mode):

cp genomes/main/genome.db backups/genome-$(date +%Y%m%d).db

BGE-M3 backfill (one-time, after install):

python scripts/backfill_bgem3_v2.py genomes/main/genome.db

Observability

scripts\setup-grafana-telem.ps1     # Windows
scripts/setup-grafana-telem.sh      # Linux / macOS

Dashboard: http://localhost:3000/d/helix-overview. Full surface: docs/architecture/OBSERVABILITY.md.

Gotchas

• Knowledge store path is genomes/main/genome.db (not project root). Delete to start fresh.
• BGE-M3 backfill is one-time post-install — embedding_dense_v2 IS NULL until you run scripts/backfill_bgem3_v2.py. Low retrieval rate without it.
• Fusion mode defaults to "additive" (back-compat). Flip to "rrf" in [retrieval] after running scripts/calibrate_thresholds.py.
• Session delivery (session_delivery_enabled = true) tracks delivered docs per session, elides repeats. ~40% token savings on multi-turn. Pass ignore_delivered: true in /context body for benchmarks.
• know/miss contract requires the agent prompt fragment to be honored — without it, frontier models confabulate. Import helix_context.agent_prompt.full_fragment().
• Naming lexicon: biology terms (gene, genome, ribosome) have canonical software equivalents (document, knowledge store, compressor). Both work in code; new code uses software terms. See docs/ROSETTA.md.

Testing

python -m pytest tests/ -m "not live" -v   # ~1950 tests, no external services

Documentation

Acknowledgments

Built on: spaCy NER · Howard 2005 TCM · Stachenfeld 2017 SR · SQLite FTS5 BM25 · BGE-M3 · Kompress · Headroom

License

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