Skip to main content Module core Copy item path Source a2a a2a_transport active_inference Active Inference Preload — predictive context loading. adaptive adaptive_chunking Adaptive chunk sizing from a rough “prefrontal” budget controller (tight → signatures, generous → full bodies). adaptive_mode_policy adaptive_thresholds addons Addon ecosystem: community extensions for lean-ctx (#858). agent_budget agent_identity agent_registry First-class agent identities: registry + lifecycle (GL #433, H3 Epic D). agent_runtime_env Bridge agent runtime/session environment variables across lean-ctx processes. agents aggressiveness Single 0.0–1.0 compression-intensity knob. anchor Hash-anchored line identifiers — the shared spine of anchored editing
(epic #1008 / GL#1009). ann_cache Process-wide cache for the HNSW AnnIndex anomaly Anomaly detection using Welford’s online algorithm for running
mean/variance and triggering alerts at >3x standard deviation. archive archive_fts artifact_index artifacts ast_walk Stack-safe traversal of tree-sitter ASTs. atomic_fs Shared, policy-free atomic-write mechanics. attention Convenience re-export: all attention-related modules. attention_context Attention-Weighted Context Assembly. attention_layout_driver attention_model Heuristic attention prediction model for LLM context optimization. attention_placement U-shaped placement (primacy/recency): alternate placing ranked items at context head vs tail. audit_trail auto_capture Opt-in automatic knowledge capture from tool outputs. auto_findings auto_mode_resolver autonomy_drivers bandit benchmark benchmark_compare billing Commercial-plane billing substrate (billing-plane-v1): plans, entitlements,
and usage metering derived from the signed savings ledger. Never gates local.
Commercial-plane billing substrate (billing-plane-v1, EPIC 13.6). binary_detect bm25_cache bm25_index bounce_tracker buddy budget_tracker Runtime budget tracking against role limits. budgets cache cache_telemetry Read-cache reliability telemetry. call_graph capabilities ccp_session_bundle chunks_ts Tree-sitter AST-aware code chunking for semantic search. claim_extractor Claim Extraction — decomposes context pipeline outputs into verifiable claims. cli_cache client_capabilities client_constraints cloud_files Detection of cloud-backed placeholder files (OneDrive “Files On-Demand” on
Windows, iCloud Drive “dataless” files on macOS). code_health Native code-health engine — clean code as a token-cost lever. codebook cognition_loop Hebbian-inspired Cognition Loop — periodic background reorganization of knowledge.
Runs 9 steps: seed promote, structural repair, fidelity check, lateral synthesis,
contradiction resolution, hebbian strengthen, decay, compact, observation synthesis. cognition_scheduler Opportunistic, time-gated trigger for the background Cognition Loop. cognitive_load Cognitive Load Theory–style scoring for code context (intrinsic / extraneous / germane). community Community detection on the code graph. compliance Framework compliance mappings + coverage reports (GL #424, H3 Epic A). compliance_report Signed CISO compliance report (GL #677) — “the deliverable a CISO shows
auditors”. compress_preview compress_preview — a read-only “what would compression do to this?” inspector
(#984, Headroom #1267). compression Convenience re-export: all compression-related modules. compression_safety compressor config config_heal Heal a config.toml that got stranded in the DATA dir (#594). conformance Conformance & reproducibility scorecard (conformance-v1, EPIC 12.17). consolidation Consolidation engine for provider data — hippocampal sleep replay. consolidation_engine Canonical session→knowledge consolidation (#995 Phase 4). content_cache Resident, bounded file-content cache shared across the search-index build and
ctx_search (issue #148). content_chunk Universal content chunk — the atomic unit of the Context Engine. context Convenience re-export: all context-related modules. context_artifacts context_column Context Column — the cortical column abstraction for data source pipelines. context_compiler Context Compiler – builds minimal context packages under budget constraints. context_deficit context_field Context Field Theory (CFT) – unified potential function for context items. context_handles Context Handles – sparse, lazy references to context items. context_ir context_ledger context_lint Injected-context linter (#960) — keeps lean-ctx’s OWN injected context
(the rules block + advertised tool descriptions) high-signal. context_os context_overhead Honest accounting of the fixed per-turn context lean-ctx injects (GitHub #361). context_overlay Reversible context overlays — user/policy manipulations that modify
context items without changing source files (“synaptic modulation”). context_package context_packing Submodular context packing — choosing which items to include under a
budget so the selected set maximises coverage of what the task cares about. context_policies Context Policy Engine – declarative rules for context governance. context_proof context_proof_v2 ContextProofV2 — Claim-based verification schema. context_radar context_snapshot Context Snapshot (CONTEXT_SNAPSHOT_V1) — the git-anchored, signed, temporal
artifact behind the Context Time Machine (GL epic #1022, Phase 0 #1023). contextops contracts conversation Conversation identity for read-cache scoping. cooccurrence Hebbian file co-access graph — “files that fire together, wire together”. crash_log Persistent crash log for panic diagnosability. cross_source_edges Cross-source graph edges — connects external data to code via the graph index. cross_source_hints Cross-source hints — lateral connections between cortical columns. cyclomatic Cyclomatic complexity via tree-sitter (decision-point counting). data_consolidate Consolidate a split data layout into one canonical directory (GH #414). data_dir datadog_push Agentless Datadog export (GL #401, setup path B). debug_log Opt-in human-readable debug log of intercepted tool activity (#520). deep_queries Tree-sitter deep queries for extracting imports, call sites, and type definitions. degradation_policy dense_backend deps diagnostics_store Active compiler/linter diagnostics as a context-priority signal (#499). edit_metering Edit-efficiency metering (#1008, honest-metering philosophy #361). edit_quality Quality loop v1 (#494): compression-caused edit failures feed back into
mode selection. editor_registry editor_signal Editor focus signal (#500). efficacy Learning-efficacy evidence (#549, VIS-3). egress Egress / output DLP for agent writes & actions (GL #676) — the output
side of the Great Filter. embedding_index Persistent, incremental embedding index. embedding_quant int8 scalar quantization + SIMD-friendly scoring for embedding vectors. embeddings Embedding engine for semantic code search. energy Electricity-footprint estimate for saved tokens. entropy episodic_memory Episodic Memory — persistent cross-session experiences with outcomes. error eval_ab Deterministic with/without output-quality eval (#232). eval_harness Retrieval evaluation harness for lean-ctx hybrid search. events eviction_orchestrator Connects the RSS-based memory_guard to real cache eviction via HomeostasisController. evidence Evidence schema — an attributable claim with a confidence score and source. evidence_bundle Evidence bundle generator (evidence-bundle-v1, GL #425, H3 Epic A). evidence_ledger extension_registry Pluggable read-modes, compressors, and chunkers (extension-registry-v1). extractive Extractive prose compression: keep the most informative sentences within a
char budget instead of truncating to the prefix. extractors Format extractors & format-aware chunkers (extractors-v1, EPIC 12.13). feedback fep_prefetch FEP prefetch (#9): active-inference-style warmup. filters finops_export FinOps cost/savings export (GL #402): CloudZero CBF, Vantage, FOCUS CSV. firewall Context firewall: replace large tool outputs with a compact digest + retrieval ref. free_energy_budget Free-Energy Budget — optimal context allocation under token constraints. gain gamma_cover Minimal TF-IDF chunk subset covering peers within γ bits of residual information. git Native git support: remote-repo reading (cached shallow clone) and a shadow
history of agent changes. git_cache TTL-based cache for git command results. git_signals Git working-set signals for relevance ranking (#497). git_util Minimal, dependency-free git shell-outs shared by the context-artifact and
analysis tools (ctx_impact, ctx_architecture, context_artifacts). godot Godot-specific parsers and helpers. gotcha_tracker grammar_usage Persistent per-extension signature-backend telemetry (GH #690 Phase 2). graph Convenience re-export: all graph-related modules. graph_analysis Static dependency-graph analyses (God-Nodes, import cycles). graph_cache Resident graph-index cache (Phase 5 of the efficiency epic). graph_context Graph-driven context loading — automatically includes related files
based on Property Graph proximity and token budgeting. graph_coordinator Non-blocking, PropertyGraph-first build coordinator for the dashboard graph
routes (#696 phase C). graph_enricher Unified Graph Enricher — indexes Git history, tests, and knowledge into the PropertyGraph. graph_export graph_features Multi-layer graph descriptors inspired by GNN message passing:
PageRank (global importance), local clustering, HITS hubs/authorities,
and weakly-connected community ids. graph_index graph_parity Shadow-mode parity harness (#682.3): prove the PropertyGraph reproduces
everything graph_index exposes through the GraphProviderbefore any backend flip
(#682.4) relies on PG. graph_provider handle Stable cross-turn symbol handles: path#name@Lline. handoff_ledger handoff_transfer_bundle hasher Central hashing utility using BLAKE3. heatmap hebbian_cache Hebbian Co-Access Cache with Boltzmann-Temperature Eviction. hnsw Lightweight HNSW (Hierarchical Navigable Small World) index for approximate nearest neighbors. home homeostasis Homeostatic Memory Guard — Proactive multi-level resource management. http_client hybrid_search Hybrid search combining BM25 (lexical) with dense vector search. ide_permissions Read the active IDE’s tool-permission configuration and resolve an effective
action for a lean-ctx tool, so lean-ctx can mirror (“inherit”) the user’s
IDE permission rules instead of forming a second, ungoverned execution path. immune_detector Immune detector (#8): artificial-immune-system screening for context
poisoning. import_resolver Import-to-file resolution (AST-driven import strings → project paths). index_admission Admission control for heavy index builds (#685). index_bundle Encrypted index bundles for the hosted Personal Index (GL #392). index_filter Index-time file filters (#735): one shared corpus-membership decision for
every index builder. index_namespace index_orchestrator index_paths Index path utilities — project-root normalization and graph match keys. information_bottleneck QUITO-X–style trade-off: compress by dropping low token-entropy lines while targeting an output/input token ratio. ingestion Ingestion front-door (ingestion-spec-v1). input_filters Inbound content filters (GL #675) — the input side of the Great Filter. instruction_compiler integrity Build-origin integrity verification and rebrand resistance. intent_engine intent_protocol intent_router interrupt User-initiated cooperative cancellation (Ctrl-C / SIGINT). introspect Cognition introspection — proves that each scientific subsystem is wired
and actually active at runtime. io_boundary io_health journal json_crush Deterministic JSON crusher — single source of truth for structural JSON
compaction (#934, Headroom “Smart Crusher” port, GitLab #935). json_sample Statistical JSON array row-sampling (#1147, SmartCrusher-equivalent). jsonc knowledge knowledge_bootstrap knowledge_bridge Cross-Agent Knowledge Bridge — controlled sharing of high-confidence facts between agents. knowledge_domain Convenience re-export: all knowledge-related modules. knowledge_embedding Embedding-based Knowledge Retrieval for ctx_knowledge. knowledge_provider_extract Knowledge auto-extraction from provider data (issues, PRs, DB schemas). knowledge_relations knowledge_vault Zero-knowledge Personal-Cloud knowledge vault (GL #467) — the E2E
envelope for /api/sync/knowledge. language_capabilities layout_pin XDG layout commitment pin (GL #623 / #624). learning_sync Team sync of the learning layers (#550, VIS-4). levenshtein Shared edit-distance “did you mean” helper (#712). limits litm litm_calibration Empirical LITM placement calibration (#539, EFF-2). llm_enhance Optional LLM enhancement layer. llm_feedback locomo LoCoMo memory benchmark (#291). logging loop_detection markdown_compact Deterministic Markdown/documentation compaction for LLM-agent reads. mcp_catalog MCP Tool-Catalog Federation (#210) — Engine pillar . mcp_manifest mdl_selector Minimum Description Length–style selection among abstract read modes (proxy compressed lengths). memory Convenience re-export: all memory-related modules. memory_archive Multi-store, lossless memory archive (#995 Phase 1). memory_boundary memory_capacity Single memory capacity manager (#995 Phase 2). memory_consolidation Sleep-inspired consolidation for in-memory knowledge entries (NREM merge, REM prune, replay boost). memory_guard Process-level RAM guardian with adaptive eviction and hard OOM protection. memory_lifecycle Memory Lifecycle Management — consolidation, decay, compaction, archival. memory_policy memory_salience Content salience for memory: a cheap, deterministic keyword score over a
fact/finding’s raw text. mode_predictor model_registry multi_repo multiscale_index Renormalization-Inspired Multi-Scale Indexing. nc_compress Non-code compression tuning (nc-compress-v1, EPIC 12.14). neural Neural context compression — trained models replacing heuristic filters. ocp Open Context Protocol (OCP) v0.1 export adapter. openapi OpenAPI 3.0 document for the public /v1 surface, generated from a single
in-code endpoint inventory (endpointshttp_server; the SSOT lives here so it stays compiled and
drift-tested without the http-server feature. ort_environment ONNX Runtime global environment: single init per process, runtime dylib loading. ort_execution_providers ONNX Runtime execution provider selection: CPU default, opt-in GPU providers. output_echo Output-echo detection (#501). output_sanitizer Output sanitizer: detects and cleans degenerate model artifacts from compressed output. output_verification owasp_alignment Maps lean-ctx security features to the OWASP Top 10 for Agentic Applications (2025).
Used by lean-ctx audit CLI and /v1/audit/events endpoint. pagerank PageRank computation on the Property Graph. path_locks Process-wide per-path advisory locks. path_mode_memory Persistent per-path bounce memory (#496). path_resolve Shared path-resolution for tool handlers. pathjail paths Typed XDG base-directory resolvers for lean-ctx (GH #408 / GL #602). pathutil patterns persona Context personas (persona-spec-v1). pgvector_store Optional pgvector (PostgreSQL) backend for dense (embedding) search. pipeline Context Pipeline plugins policy Context Policy Packs v1 — Policies-as-Code (GL #489). pop_pruning portable_binary predictive_coding Predictive Coding for mode outputs — send only prediction errors (deltas). predictive_prefetch Predictive Prefetch via Free Energy Minimization. preservation procedural_memory Procedural Memory — recurring workflow detection and template storage. process_guard Global concurrency limiter for lean-ctx processes. profile_suggest Repo-stack-aware profile recommendation (#851). profiles Context Profiles progressive_compression Progressive compression: newest segments stay verbose; older tiers lose detail under exponential budget shrink. project_hash property_graph Property Graph Engine — SQLite-backed code knowledge graph. prospective_memory protect Explicit, deterministic preservation of user-marked spans across every
compressor (read, shell, proxy, prose). protocol provider_bandit Provider Bandit — Thompson Sampling for provider selection. provider_cache Provider caching awareness — helps LLM providers cache repeated context. providers qdrant_store Optional Qdrant backend for dense (embedding) search. quality qubo_select QUBO context-selection spike (#10) — research only, never the default. rabin_karp read_stub_index Persistent, conversation-scoped index of fully-delivered file reads (#955). recovery Recovery-hint policy + grammar. redaction reference_docs Generated reference documentation. repomap Repo map: PageRank-based symbol importance ranking across a codebase. roles Role-based access control for agent governance. route_extractor rule_artifacts The committed LEAN-CTX.md rule artifacts as (relative_path, content)
pairs — the single source shared by the regenerator (gen_rules example)
and the drift gate (tests/rules_drift.rs) so the project copy and the
rust/ copy can never disagree. rules_canonical Canonical rules source — single source of truth for all lean-ctx guidance. rules_channel Cross-channel rule deduplication policy (#684). rules_overhead Shared accounting for the rules files an agent auto-loads each session. rules_sections Profile-aware rules section builders (#756). safety_needles saliency Saliency Map — information-theoretic chunk ranking and deduplication. sandbox sandbox_landlock sandbox_seatbelt sanitize savings_autopush Opt-in background auto-push of the local savings batch to the team server. savings_footer savings_ledger Verified Savings Ledger (G1) — the per-event, auditable counterfactual store. scent_field Stigmergic scent field for zero-token multi-agent coordination (#540, EFF-3). scorecard Reproducible scorecard (#211): one command → compression savings + retrieval
recall/MRR + latency over a fixed, deterministic scenario matrix. search Convenience re-export: all search-related modules. search_delta search_index Resident line-search index for ctx_search (Phase 1 of the efficiency epic). search_reranking Post-RRF Reranking Pipeline for code-aware search. secret_detection security_posture Unified, human-facing view of lean-ctx’s two independent security planes : semantic_cache semantic_chunks Semantic Chunking with Attention Bridges. sensitivity Per-item sensitivity model with a uniform policy floor (#212). server_capabilities GET /v1/capabilities — runtime discovery of what this lean-ctx instance
supports, so any client (any language) can branch on real features instead
of trial calls. The HTTP route lives in http_server; the payload builder
lives here so it stays compiled (and drift-tested) without the
http-server feature.session session_diff Session diffing — structured comparison of two session states. session_domain Convenience re-export: all session-related modules. session_summary AI session summaries (#292): periodically distil the working session into a
compact, semantically-recallable digest. session_token setup_report share Frictionless share helpers: copy text to the OS clipboard and open files/URLs. shell_allowlist Shell allowlist with AST-based command parsing. signatures Signature extraction — regex fallback layer. signatures_ts skillify Skillify (#290): distill a project’s recurring session patterns into versioned,
git-committable .cursor/rules/skillify-*.mdc rule files. slo Context SLOs — configurable service level objectives for context metrics. slow_log smells Code smell detection engine. splade_retrieval SPLADE-style sparse expansion layered on BM25 candidate retrieval. spreading_activation Spreading activation — associative retrieval over a weighted graph. startup_guard stats storage_maintenance Daemon-safe storage maintenance. structural_diff Structural diff using tree-sitter chunk identities (named declarations). structural_tokenizer Structural tokenizer treating idiomatic multi-token spans as single motifs. structured_compact Lossless compaction of structured data (JSON / JSON Lines). structured_read subagent_contract Sub-agent context contracts (GL#450). surprise Predictive Surprise Scoring — conditional entropy relative to LLM knowledge. symbol_map syntax_validate Post-edit syntax gate (#1008): a tree-sitter parse check used to reject an
edit that turns a cleanly parsing file into a broken one. tabular_crush Deterministic tabular (CSV/TSV) crusher — columnar redundancy factoring for
delimited data (#982, Headroom tabular-compressor port, GitLab Epic #973). task_briefing task_relevance tdd_schema team_slo Team-server SLO instrumentation — the measurement half of the hosted-index
reliability gate (GL #391). telemetry Telemetry and metrics collection following OpenTelemetry GenAI conventions. terse Terse Compression Engine theme threshold_learning Online-learned per-extension compression-threshold deltas (#538, EFF-1). tokenizer_translation_driver tokens tool_health lean-ctx tools health — does every advertised MCP tool and injected rule
earn its always-on token cost? (#848)tool_lifecycle Shared tool lifecycle — ensures CLI and MCP paths have identical side effects. tool_profiles transcript_compact Transcript/conversation compaction for agent session JSONL files. type_ref_edges Cross-file type-reference edges (GH #398). update_scheduler OS-specific auto-update scheduler management.
Supports macOS LaunchAgent, Linux systemd/cron, Windows Task Scheduler. updater verification_observability version_check visualizer Data collection and HTML rendering for the interactive visualizer. walk_filter Central filter_entry predicate shared by every directory walker
(graph/index builders, ctx_search, ctx_tree, ctx_glob, CLI scans). web Web & research context layer. workflow workspace_config workspace_trust Workspace trust for project-local .lean-ctx.toml overrides (GH security
audit, finding 4). wrapped wrapped_share Opt-in, self-hostable Wrapped share page. wrapped_svg Shareable SVG “Wrapped” card. xdg_migrate On-demand XDG split migration (GH #408 / GL #606). yaml_crush Deterministic YAML crusher — structural compaction for YAML documents (#985,
Headroom YAML-compressor port, GitLab Epic #973).