Source Analysis — Vespa Code-map.md vs Full 150-Module Workspace Inventory

vespa/Code-map.md provides the canonical entry-point narrative for the Vespa engine, explicitly scoping itself to “the modules you are most likely to encounter” across four subsystems (container-jdisc, content-nodes, config-admin, and utility-libs).

Quantitative Coverage Breakdown

1. Mapped Core Modules (~25 modules):
   • Account for the vast majority of active engine LOC (searchlib ~305k LOC, container-search ~171k LOC, config-model ~132k LOC, searchcore ~131k LOC, vespalib ~123k LOC, storage ~108k LOC).
2. Unmapped Engine & Infrastructure Modules (~125 top-level directories in vespa/):
   • Networking & RPC: fnet (C++ async framing/network library), jrt (Java RPC transport), vespa-http-client, http-utils.
   • Search & Streaming Variants: vsm (Vespa Streaming Matcher for streaming search mode), searchsummary, searchcommon, searchvisitor, linguistics, opennlp-linguistics.
   • ML & Inference Runtimes: container-onnxruntime, container-llama, model-integration, model-evaluation.
   • Operations, Security & Telemetry: security-utils, vespa-athenz, flags, logd, logserver, metrics, metrics-proxy, zookeeper-server.

Actionable Output

• tools/garden/analyze_garden.py scans vespa/ on every run, computes exact module citation coverage in sources:, and emits non-overlapping parallel lease batches in garden-health and project-unmapped-vespa-modules-harvest.