Modular Runtime + RAG Batching + Performance Optimizations#264
Merged
Modular Runtime + RAG Batching + Performance Optimizations#264
Conversation
added 30 commits
February 1, 2026 17:05
…type gen Rust Foundation (continuum-core/src/code/): - FileEngine: read/write/edit/delete with per-persona workspace scoping - ChangeGraph: DAG of ChangeNodes with undo via reverse diff - DiffEngine: unified diff computation (similar crate) - PathSecurity: workspace isolation, path traversal guard, extension allowlist - CodeSearch: regex + glob search with .gitignore support (ignore crate) - Tree: recursive directory tree generation - GitBridge: git status and diff operations - IPC handlers for all 12 code/* endpoints (359 tests passing) TypeScript Commands (8 generated via CommandGenerator): - code/read, code/write, code/edit, code/diff - code/search, code/tree, code/undo, code/history - Each with Types.ts, ServerCommand.ts, BrowserCommand.ts, README, tests Type Safety (ts-rs single source of truth): - 14 Rust types exported via #[derive(TS)] → shared/generated/code/ - Zero hand-written wire type duplicates - All object/any casts eliminated from code/* commands - CommandParams.userId used as canonical identity field RAG Integration: - CodeToolSource: dynamic coding workflow guidance in persona system prompts - Only shows tools persona has permission to use - Budget-aware with minimal fallback - 15 unit tests passing Infrastructure fixes: - PersonaToolExecutor now injects userId (standard CommandParams field) - CLAUDE.md documents ts-rs pattern and regeneration workflow
Delete old pre-Rust development/code/read and development/code/pattern-search commands that caused TS2300 duplicate identifier collision with new code/* commands. Remove legacy CodeDaemon methods (readFile, searchCode, getGitLog, clearCache, getCacheStats, getRepositoryRoot), their types, and the PathValidator/FileReader modules — all superseded by Rust IPC workspace ops. - Delete commands/development/code/ (7 files) - Delete daemons/code-daemon/server/modules/ (PathValidator, FileReader) - Clean CodeDaemonTypes.ts: remove 222 lines of legacy types - Clean CodeDaemon.ts: remove 7 legacy static methods - Clean CodeDaemonServer.ts: remove old CodeDaemonImpl class - Fix cli.ts: replace CODE_COMMANDS import with string literals - Fix PersonaToolDefinitions.ts: update essentialTools to code/* - Regenerate server/generated.ts and command constants
…strator CodingModelSelector routes coding tasks to frontier models with provider fallback. PlanFormulator decomposes tasks into executable step DAGs via LLM. CodeAgentOrchestrator executes plans with budget enforcement, retry logic, and dependency-ordered step execution. 51 unit tests.
CodingPlanEntity is a first-class persistent entity for coding plans. Supports hierarchical delegation (parentPlanId), team assignment (assignees + leadId), governance integration (proposalId), and real-time execution tracking. CodeAgentOrchestrator now persists plans via DataDaemon with best-effort semantics (works without DB in unit tests). 80 unit tests passing.
…ordination Phase 4A — Sandbox & Security Tiers: - SecurityTier: 4-tier access control (discovery/read/write/system) - ToolAllowlistEnforcer: per-tier command filtering with glob matching - ExecutionSandbox: process-isolated code execution with timeout/output limits - Risk assessment integrated into PlanFormulator output Phase 4B — Self-Modifying Skills: - SkillEntity: persistent skill registry with full lifecycle - skill/propose: AI creates command specifications - skill/generate: programmatic CommandGenerator invocation - skill/validate: sandbox compilation + test execution - skill/activate: dynamic tool registration - skill/list: query skill registry Phase 4C — Multi-Agent Coordination & Delegation: - CodeCoordinationStream: file-level MUTEX via BaseCoordinationStream - PlanGovernance: risk-based approval routing (auto-approve low risk, require approval for multi-agent/high-risk/system-tier) - CodeTaskDelegator: union-find plan decomposition into parallel file clusters, load-balanced agent assignment, sub-plan creation, result consolidation - DryRun mode: execute plans read-only, mock write operations 342 tests across 12 test files, all passing.
Two mechanical throttle layers were overriding AI cognition: 1. Temperature system: each AI "servicing" a room subtracted -0.2, so 14 personas crashed rooms to 0.00 in 35 seconds. Fixed by flipping to +0.05 warmth (active conversation stays alive). Removed hard -0.1 priority penalty for cold rooms. 2. InferenceCoordinator: gating calls consumed per-message "cards" in messageResponders, so when actual response generation tried to acquire a slot with the same messageId, every persona was denied. Rewrote from 489→197 lines, removing 6 mechanical rules (card dealing, responder caps, reserved slots, cooldowns, stagger delays, auto-thinning). Kept only hardware capacity protection. Result: AIs respond within seconds instead of being silenced.
…n, activity/create - should-respond-fast: params.messageText crash (toLowerCase on undefined) when AI calls without messageText. Now returns graceful false result. - activity/join: activityId undefined → "Activity not found: undefined" Now validates activityId before DB lookup. - activity/create: recipeId undefined → "Recipe not found: undefined" Now validates recipeId before DB lookup. All three were AIs calling tools with missing params, getting either crashes or confusing error messages instead of clear validation errors.
Add code/task command as the entry point for the full coding agent pipeline. Wire PlanGovernance approval flow and CodeTaskDelegator into orchestrator. Add pending_approval status to CodingResult for high-risk plan gating.
CodeAgentOrchestrator.ensureWorkspace() now creates workspace directory and registers it in Rust backend before first code/* operation. Personas get writable workspace + read-only codebase access for discovery.
…ees, git write ops, iterative dev loop Phase 1: Orchestrator reads CLAUDE.md + architecture docs during discovery so AI plans follow project conventions. Phase 2: New code/verify command — runs tsc --noEmit via ExecutionSandbox, parses TypeScript errors, auto-verifies after write/edit steps. Added to write tier. Phase 3: WorkspaceStrategy abstraction routes sandbox (isolated dir) vs worktree (git sparse checkout on real repo). CodingTask extended with workspaceMode/sparsePaths. code/task command validates and passes through. Phase 4: Rust git_bridge extended with git_add/commit/push. IPC handlers + CodeDaemon methods for all git write ops. New code/git command (status/diff/log/add/commit/push) with SecurityTier gating. PlanFormulator knows commit action. Phase 5: Verify→re-plan iteration loop in orchestrator. When auto-verify fails, re-plans with error context (quick-fix mode), executes fix, re-verifies. Configurable via autoVerify/maxVerifyIterations on ExecutionOptions. 387 TypeScript tests (15 files), 362 Rust tests — all passing.
…spaces Rust: Notify-based blocking watch (no timeouts, no polling). Each ExecutionState gets Arc<Notify> — reader tasks call notify_one() on every output line. watch_execution() blocks on Notify until output arrives, classifies lines through CompiledSentinel (pre-compiled regex, first-match-wins), advances cursors, returns classified batch. Types (ts-rs exported): OutputClassification, SentinelAction, SentinelRule, ClassifiedLine, ShellWatchResponse — all generated from Rust structs to shared/generated/code/*.ts. IPC: code/shell-watch (async bridge via rt_handle.block_on after releasing DashMap lock) and code/shell-sentinel (synchronous, brief lock). TS bridge: RustCoreIPC, CodeDaemon, CodeDaemonServer wired. Commands: code/shell/watch and code/shell/sentinel generated via CommandGenerator with full server implementations delegating to CodeDaemon. Registered in schemas, constants, executors, structure. Workspace handle: sentinel(), watch(), execWatch(cmd, rules?, onLine?) — composed convenience that runs exec → sentinel → watch loop. Tests: 389 Rust lib tests pass, 183 TS unit tests pass (36 Workspace tests including 5 new watch/sentinel/execWatch tests), 0 failures.
…ls in system prompt DataDaemon.read<T>() now returns T|null (consistent with store/update), eliminating the .data.data unwrapping pattern across ~30 call sites. Recipe tools and strategy rules flow from JSON through ChatRAGBuilder into RAGContext and are injected into PersonaResponseGenerator system prompt as activity context for the LLM. PersonaUser supports per-room workspaces via Map<string, Workspace> with room-aware mode selection (worktree for code rooms). RecipeToolDeclaration type added to RecipeTypes, making the tools field in recipe JSONs visible to TypeScript.
…batch-micro-tune to GenomeJobCreate; fix VoiceService STT + GeminiLive cancel
Seed script (1265→622 lines): - Bulk load all users/rooms in single subprocess calls - Parallel Promise.all() for config updates (was 48+ sequential spawns) - Removed 550 lines of local function duplicates shadowing imports - Cleaned unused imports Code room: - Added CODE to DEFAULT_ROOMS in DefaultEntities.ts - Added code room to seed script with coding recipe - Added to ALL_EXPECTED_ROOMS for idempotent creation Auto-workspace bootstrapping: - PersonaToolExecutor auto-creates workspace when code/* tools invoked - Added ensureCodeWorkspace callback through MotorCortex→PersonaUser - Personas no longer need manual workspace creation to use code tools Validated: Together Assistant created index.html (365 bytes) using code/write and verified with code/tree. DeepSeek, Groq, Grok all using code/* tools correctly in the code room.
Two fixes for AI personas calling code/* tools with wrong parameter names: 1. PersonaToolDefinitions: PARAM_DESCRIPTION_OVERRIDES map provides meaningful descriptions (e.g. "Relative path to file within workspace") instead of generic "filePath parameter". Applied at tool definition build time so LLMs see useful descriptions in their function schemas. 2. PersonaToolExecutor: PARAM_CORRECTIONS map silently corrects common wrong names (path→filePath, contents→content, query→pattern, etc.) before validation. Covers code/write, code/read, code/edit, code/search, code/tree, code/git. Result: Together and Groq both successfully created calculator.html files using code/write with correct parameters after these fixes.
…tization, meta-tool exclusion - ToolRegistry: load descriptions via includeDescription:true (was empty strings) - PersonaToolExecutor: strip CDATA wrappers (Together), regex HTML entity decode (Groq), file_name correction - PersonaResponseGenerator: cap native tools at 64, prioritize recipe tools, exclude search_tools/list_tools/working_memory from native specs to prevent Claude meta-tool loops
…summaries, permissions, ToolNameCodec
Four bugs fixed in the persona tool execution pipeline:
1. Message accumulation: each tool loop iteration rebuilt messages from scratch,
so the model never saw previous tool calls or results. Now accumulates
assistant+result messages across iterations.
2. Stale toolCalls: after regeneration, only aiResponse.text was updated — the
old toolCalls carried over, causing infinite re-execution. Now updates both
text and toolCalls from regenerated response.
3. Lean summary vagueness: tool result summaries said "completed" and closed
with "provide your analysis" (inviting more tool calls). Now produces
human-readable summaries ("Wrote 1218 bytes to hello.html") with
context-dependent stop signals that prevent re-calling succeeded tools.
4. Permission filtering: role permissions used inconsistent suffixes (:read,
:search, :write) that never matched the {category}:execute format tools
actually check. All roles now use {category}:execute consistently.
Additional improvements:
- ToolNameCodec: bidirectional encoder/decoder for API tool names. Models mangle
names (code__write, $FUNCTIONS.code_write, code-write) — codec handles all
variants via reverse lookup map populated at registration time.
- 3-tier tool prioritization: recipe tools → essentials → fill (was flat).
Tighter cap (32) when recipe tools present.
- Loop detection threshold lowered from 3 to 2 identical calls.
- Tools stripped from regeneration request when all calls succeeded, forcing
text-only response.
Layer 1 — Handle-based command execution: - CommandParams: background, timeout, onTimeout fields - CommandResult: handle, handleId, timedOut fields - Commands.execute(): fast path (direct) vs tracked path (delegate to ServerCommands) - ServerCommands: background fire-and-forget, timeout with cancel/continue/fail, await by handle ref, event subscriptions per handle lifecycle - Wired into JTAGSystemServer.connect() via dynamic import Layer 2 — Canonical agent tool loop: - ContentPart: tool_use + tool_result types in AIProviderTypesV2 - AnthropicAdapter: native tool_use/tool_result content block handling - BaseOpenAICompatibleAdapter: tool_calls parsing, finish_reason mapping, tool messages - MediaContentFormatter: tool_use/tool_result formatting for both protocols - PersonaResponseGenerator: while(finishReason==='tool_use') loop replaces broken PHASE 3.3.6 — model decides when to stop, full results returned, no summaries - PersonaToolExecutor: executeSingleTool() extracted, executeNativeToolCalls() calls it directly (no XML round-trip), prepareBatch() shared between paths - Safety caps: frontier 25, mid-tier native 10, XML/local 5 Quality: - generateSummary(): data-shape-driven (filePath, bytes, count, dimensions) instead of per-tool if-statements — single algorithm for all tools - Chat export filters metadata.toolResult messages by default - mediaToContentParts() helper deduplicates conversion in both loop paths - toolExecutionContext hoisted as loop-invariant
…olidate browser detection ModelContextWindows: add versioned Anthropic model IDs (claude-sonnet-4-5-20250929, claude-opus-4-20250514, claude-3-5-haiku-20241022), cloud API models (Groq, Together, Fireworks, Gemini 2.0, Grok 4, Qwen 3 Omni), inference speed entries for all cloud models, and date-suffix normalization so future versioned IDs resolve without explicit entries. Fixes 25x context window regression (8K → 200K for Claude). CodeToolSource: replace shallow workflow listing with substantive coding methodology — dynamic capability detection, Read→Edit→Verify→Iterate workflow, rules, anti-patterns. Budget increased from 5% to 8%. SystemOrchestrator: consolidate three browser-open locations into single detectAndManageBrowser() with ping-based detection and retry logic for WebSocket reconnection after server restart. Eliminates duplicate tab launches.
…s, spam rate-limiting Rust logger worker: replace per-message file.flush() with batched periodic flush (every 250ms or 200 messages). Reduces disk flushes from ~700/sec to ~4/sec. Add per-category rate limiter (100 msg/sec) that drops excess and logs a warning when burst ends. WorkerClient: remove synchronous appendFileSync debug logging that blocked the event loop on every IPC call. Break recursive timing loop by skipping TimingHarness for write-log/flush-logs IPC types. TimingHarness: replace appendFileSync with async WriteStream + periodic buffer flush (500ms). Add per-category enable/disable. New LogLevelRegistry: per-component runtime log level overrides (like audio mute buttons). Supports wildcard patterns (daemons/*). Wired into ComponentLogger.shouldLog() for filtering before IPC.
…d timing opt-in - Default log level changed from INFO to WARN (quiet by default) - Per-component overrides for daemon startup spam, PersonaUser loop, RAG pipeline - Fix logger worker busy-spin: Duration::ZERO caused 95.8% CPU when idle - Fix timing file opt-in: was opt-out (undefined !== 'false' = true), now requires explicit enable - Config template updated with JTAG_TIMING_ENABLED=false and WARN defaults
Single ping attempt falsely concluded no browser was connected (transient WebSocket reconnect timing), then opened a duplicate tab on every npm start. Now retries up to 3 times with 3s delays, matching SystemOrchestrator pattern.
- CandleGrpcAdapter: remove rogue appendFileSync, route through Logger.ts - clean:logs: fix find pattern to catch all .log files under personas/ - LoggingConfig: auto-enable persona when category enabled, auto-disable when last category disabled - logs/config: buildPersonaStatuses() queries all AI personas from DB, shows ON/OFF status per persona with source (explicit/default) and available categories in overview - Default logging OFF for all personas (beta-ready quiet baseline)
…te-limit - SessionDaemonServer: remove console.error debug dump on every session create (~200 lines of noise per deploy) - DatabaseHandleRegistry: remove per-handle console.log on open/alias - SqliteStorageAdapter: remove redundant path debug console.log - PersonaWorkerThread: consolidate 3-line worker start to 1 line - RustCoreIPC: make slow IPC rate-limit map static (shared across all 16 persona instances, not per-instance) Total log reduction: 14K lines → 3K lines (-79%), 2MB → 312KB (-84%)
…ire-and-forget IPC - ConversationHistorySource: single-flight request coalescing eliminates thundering herd (16 personas → 1 DB query per room), plus 2s TTL cache for subsequent access - PersonaIdentitySource: permanent identity cache (immutable at runtime) - GlobalAwarenessSource: 60s negative cache for "No memory corpus" IPC errors - SemanticMemorySource: 60s negative cache for "No memory corpus" IPC errors - AIProviderDaemon: fire-and-forget log_generation (was 2-5s blocking, now 0ms) - Hippocampus: fire-and-forget Rust corpus memory append - UnifiedConsciousness: fire-and-forget Rust corpus event append - PersonaInbox: signal-based pop() (was 100ms polling), binary insert O(log N), cached Date.now() in sort comparisons - InboxObserver: O(1) getSize() replaces O(N log N) peek(1000) for depth check Result: CPU 103% → 60% (41% reduction)
…ng, project-mode workspaces - New commands: code/shell/execute (async+blocking), code/shell/status, code/shell/kill - ToolOutputAdapter: ShellExecute/Watch/Status/Kill renderers for inline chat output cards - WorkspaceStrategy: project mode default (git worktrees), robust reuse/cleanup, git root auto-detection - PersonaUser: eager shell session creation on first code/* tool call - LLM param coercion: boolean (wait) and array (paths) survive string serialization from LLMs - Terminal widget scaffold at /terminal for dashboard view - Coding recipe updated with shell tools + strategy rules
…clogging chat - Add getToolCapability() to ToolFormatAdapter: classifies providers as native/xml/none - Gate tool injection in PersonaResponseGenerator: 'none' providers get zero tools - Add directed @mention filter in PersonaMessageEvaluator: early gate before expensive cognition work so non-mentioned personas stay silent immediately - Add toolCapability override field to ModelConfig for per-persona control
Evaluator pipeline (EVAL-PIPELINE): - task_create, plan_formulate, plan_log, state_update, state_snapshot, wm_store_observation, evaluate_and_respond, wm_store_reflection, plan_completion_log, state_cleanup Response pipeline (PIPELINE): - 3.1_rag, 3.2_format, 3.3a_slot, 3.3b_daemon_init, 3.3_inference, 3.4_agent_loop, 3.5_post Findings: CognitionLogger writes (plan_log, state_snapshot, plan_completion_log) eat 10-30s per response through Node.js DataCreate.execute(). Message posting takes 1.5-3.7s. The data layer is the bottleneck, not inference.
…ion writes Three fixes that cut data layer overhead from 10-30s to <1s per AI response: 1. Double event emission bug: DataDaemon.store() and static update() were emitting events AFTER calling instance create()/update() which already emit. Every write fired 2 WebSocket broadcasts — now fires 1. 2. CognitionLogger fire-and-forget: All 10 DataCreate.execute() calls per response cycle removed await, added suppressEvents: true, added .catch(). Cognition logs are observability data with no UI subscribers — blocking on them was pure waste. logPlanCompletion/logPlanAdjustment converted to synchronous void with fire-and-forget IIFE for internal read-modify-write. 3. suppressEvents wiring: DataDaemon.store() now accepts suppressEvents param, DataCreateServerCommand passes params.suppressEvents through default path. Measured: plan_log 9s→0ms, state_snapshot 6s→0ms, plan_completion 15s→0ms, message post 3.7s→320ms. All AI personas verified responding correctly.
Events.emit() was awaiting WebSocket delivery to all subscribers before returning. This blocked every DB write, status update, and telemetry event until all connected clients acknowledged receipt — pure waste. Changes: - DataDaemon: All instance methods (create, update, delete, batch, clear, truncate) now fire-and-forget on Events.emit(). DB write succeeds, event notification is non-blocking. Also fixed double-emit bug in static remove() (same pattern as store/update from prior commit). - PersonaResponseGenerator: All 9 Events.emit() calls (DECIDED_SILENT x5, STAGE_COMPLETE, AI_POSTED, voice routing, AI_ERROR) now fire-and-forget. - PersonaMessageEvaluator: All 6 Events.emit() calls (EVALUATING, DECIDED_SILENT x2, DECIDED_RESPOND, GENERATING, error STAGE_COMPLETE) now fire-and-forget. Events still fire and reach the browser — just don't block the server. Measured: Together 3.5_post 86ms (was 1-3.7s without contention). Remaining 1.5s on Groq is SQLite lock contention — Phase 1 (Rust IPC).
Two bottlenecks eliminated from the response pipeline: 1. Semantic loop detection: Replaced AiDetectSemanticLoop.execute() command (embedding IPC + DB query, ~20 seconds) with inline Jaccard n-gram similarity against the already-loaded RAG conversation history. Same detection quality, ~0ms instead of ~20s. 2. RAG context update: Made fire-and-forget. updateRAGContext() does read-modify-write to store the trigger message in the context entity. This is bookkeeping — the pre-built RAG context already has current messages, so blocking on the store is pure waste. Pipeline now dominated by inference time only: Groq: 13.6s total (13.5s inference, 126ms post) Together: 14.8s total (13.5s inference, 1.3s post) DeepSeek: 28s total (27.2s inference, 889ms post)
- Replace SQLite query in post-inference validation with static in-memory cache (5s → 0ms per persona per response) - Post responses via DataDaemon.store() directly, bypassing JTAGClient → CommandDaemon → DataCreateServerCommand routing - Move voice TTS event emission before DB write so speech isn't blocked by SQLite contention (500-1500ms saved for voice)
…orage eviction RAG context build: 4.4s → 115ms (38x) via static caches + single-flight coalescing - ChatRAGBuilder: room cache (60s TTL, coalesced), user name cache, artifact cache - ConversationHistorySource: event-driven 30s cache, getCachedRawMessages for dedup - PersonaIdentitySource: batch pre-warm all personas in one query IPC bridge: sequential → concurrent request handling - handle_client now dispatches each request to rayon thread pool - Dedicated writer thread serializes responses back to socket - TS client already multiplexes via requestId — no client changes needed - Eliminates per-connection serialization that caused 1.5-2.7s delays Browser: fix localStorage QuotaExceededError causing stale cache - AsyncStorage evicts oldest 50% of entity cache entries on quota exceeded - Fixes bug where new data couldn't write, leaving 2-day-old stale cache
…njection - New workspace/list command: scans .git/continuum-worktrees/ on disk, merges WorkspaceStrategy in-memory state, returns git status per workspace. Supports --personaId filter. Generated via CommandGenerator. - ProjectContextSource RAG fix: falls back to main repo when no persona-specific workspace exists (solves chicken-and-egg — personas now see codebase context before first code/* tool use). Distinguishes "Shared Repository" vs "Your Workspace" with bootstrapping hint. - SentinelAutoConfig: auto-detects build/test/lint commands from code/shell/execute and injects sentinel classification rules. Wired into PersonaToolExecutor as fire-and-forget after shell execution starts. - Git cleanup: removed 39 stale undefined/* branches, 2 stale session worktrees, fast-forwarded 3 valid ai/ worktrees to HEAD.
…aemon worker Wire the existing Rust data-daemon worker (Unix socket IPC) into the DataDaemon's per-collection adapter routing. 10 write-heavy observability collections now go through Rust instead of TypeScript SQLite. Key changes: - DataDaemonServer: connectRustDataWorker() waits for socket, registers RustWorkerStorageAdapter for 10 collections, process.exit(1) if worker not available (no silent fallback) - DatabaseHandleRegistry: re-enable RustWorkerStorageAdapter import and 'rust' case in open() - DefaultStorageAdapterFactory: 'rust' case returns RustWorkerStorageAdapter - Rust data-daemon worker: PRAGMA table_info column filtering in data_create/data_update (and timed variants) to handle TypeScript entities with fields that don't map to SQL columns Verified: 1113 requests, 779 creates, 0 errors. Killing the Rust worker causes writes to fail (no fallback to TypeScript SQLite).
Three bugs fixed in the modular runtime refactor:
1. channel.rs: Return flat response structure for service-cycle-full
- Before: { "service_cycle": { should_process, ... } } (nested)
- After: { should_process, item, channel, wait_ms, stats, decision } (flat)
- TypeScript expected result.should_process directly
2. health.rs: Return correct field name for health check
- Before: { "status": "healthy" }
- After: { "healthy": true }
- Also fixed unit test to match
3. runtime.rs: Fix tokio block_on panic in route_command_sync
- Before: rt_handle.block_on() panics inside tokio context
- After: Spawn task + sync channel to bridge async -> sync
- Avoids "Cannot start a runtime from within a runtime"
Verified: All 405 tests pass, personas now respond to messages.
1. PersonaBrainWidget: toolsAvailable from ai/status (was hardcoded 12) - Added toolsAvailable to PersonaHealth type - AIStatusServerCommand queries PersonaToolRegistry - Widget displays actual available tools per persona 2. DiagnosticsWidget: dynamic log file discovery - Uses code/tree (Rust) to discover actual log files - Infers category from filename patterns - No more hardcoded hippocampus/mind/soul/body/cns list 3. DiagnosticsWidget: dynamic system health - Uses ping command to check service status - Memory, Hippocampus, RAG show actual operational state - Added CSS for operational/degraded/offline status colors Path forward: ts-rs for single source of truth (Rust → TypeScript types)
…y` types Architecture: - Rust enums with #[serde(tag = "source_type")] generate TypeScript discriminated unions - Each source type (Memory, Consciousness, Scene, Project, Custom) has strongly-typed params - RagSourceMetadata enum provides typed metadata per source type - ts-rs generates 21 TypeScript types from Rust - single source of truth Type System (Java/C# style OOP, not webdev): - RagSourceRequest: 5 variants with typed params (MemorySourceParams, SceneSourceParams, etc.) - RagSourceMetadata: corresponding typed metadata (MemorySourceMetadata, ConsciousnessSourceMetadata) - No Record<string, any> or unknown - all fields strongly typed - Pattern matching in Rust maps cleanly to TypeScript discriminated union narrowing Performance: - Rayon par_iter() loads all sources in parallel - Eliminates N sequential IPC round-trips (was 1000-1900ms per source) - Single batched rag/compose call replaces multiple memory/consciousness/custom calls Extensibility: - Scene source: video games, VR, 3D apps pass scene_id, description, objects, characters - Project source: code context with project_path, summary, recent_changes - Custom source: passthrough sections for any dynamic content - All pre-computed by TypeScript, passed through Rust for parallel aggregation Files: - modules/rag.rs: 777 lines - types, loaders, ServiceModule impl - modules/mod.rs: add pub mod rag - ipc/mod.rs: register RagModule with memory_manager dependency Generated (in .gitignore, rebuilt on cargo test): - shared/generated/rag/*.ts: 21 strongly-typed interfaces
Contributor
There was a problem hiding this comment.
Pull request overview
This PR introduces a large set of new JTAG “code/”, “skill/”, and “workspace/*” commands (with shared/browser/server packaging), updates workspace-path resolution to prefer persona uniqueId, and adjusts docs/ignore rules to support generated Rust↔TS bindings and runtime workspace data.
Changes:
- Added many new command modules (types + browser/server implementations) for skills lifecycle and workspace/code operations.
- Updated workspace path conventions and several commands to resolve paths via user
uniqueId. - Updated CLI timeout heuristics and repository docs/ignore rules for generated
ts-rstypes and runtime data.
Reviewed changes
Copilot reviewed 198 out of 416 changed files in this pull request and generated 13 comments.
Show a summary per file
| File | Description |
|---|---|
| src/debug/jtag/commands/workspace/list/shared/WorkspaceListTypes.ts | Adds shared types + executor for workspace/list |
| src/debug/jtag/commands/workspace/list/package.json | New command package metadata/scripts |
| src/debug/jtag/commands/workspace/list/browser/WorkspaceListBrowserCommand.ts | Browser delegates workspace/list to server |
| src/debug/jtag/commands/workspace/list/.npmignore | Publish ignore rules for command package |
| src/debug/jtag/commands/workspace/git/workspace/init/server/GitWorkspaceInitServerCommand.ts | Uses persona uniqueId in workspace path; coerces paths input |
| src/debug/jtag/commands/workspace/git/workspace/clean/server/GitWorkspaceCleanServerCommand.ts | Uses shared resolver for workspace path |
| src/debug/jtag/commands/workspace/git/status/server/GitStatusServerCommand.ts | Uses shared resolver for workspace path |
| src/debug/jtag/commands/workspace/git/shared/resolveWorkspacePath.ts | Adds shared workspace path resolution via DataDaemon |
| src/debug/jtag/commands/workspace/git/push/server/GitPushServerCommand.ts | Uses shared resolver for workspace path |
| src/debug/jtag/commands/user/create/shared/UserCreateTypes.ts | Extends model config with toolCapability |
| src/debug/jtag/commands/skill/validate/shared/SkillValidateTypes.ts | Adds shared types + executor for skill/validate |
| src/debug/jtag/commands/skill/validate/package.json | New command package metadata/scripts |
| src/debug/jtag/commands/skill/validate/browser/SkillValidateBrowserCommand.ts | Browser delegates skill/validate to server |
| src/debug/jtag/commands/skill/validate/README.md | Documents skill/validate |
| src/debug/jtag/commands/skill/validate/.npmignore | Publish ignore rules for command package |
| src/debug/jtag/commands/skill/propose/package.json | New skill/propose package metadata/scripts |
| src/debug/jtag/commands/skill/propose/browser/SkillProposeBrowserCommand.ts | Browser delegates skill/propose to server |
| src/debug/jtag/commands/skill/propose/.npmignore | Publish ignore rules for command package |
| src/debug/jtag/commands/skill/list/shared/SkillListTypes.ts | Adds shared types + executor for skill/list |
| src/debug/jtag/commands/skill/list/server/SkillListServerCommand.ts | Adds server implementation for skill/list |
| src/debug/jtag/commands/skill/list/package.json | New command package metadata/scripts |
| src/debug/jtag/commands/skill/list/browser/SkillListBrowserCommand.ts | Browser delegates skill/list to server |
| src/debug/jtag/commands/skill/list/README.md | Documents skill/list |
| src/debug/jtag/commands/skill/list/.npmignore | Publish ignore rules for command package |
| src/debug/jtag/commands/skill/generate/shared/SkillGenerateTypes.ts | Adds shared types + executor for skill/generate |
| src/debug/jtag/commands/skill/generate/server/SkillGenerateServerCommand.ts | Adds server implementation for skill/generate |
| src/debug/jtag/commands/skill/generate/package.json | New command package metadata/scripts |
| src/debug/jtag/commands/skill/generate/browser/SkillGenerateBrowserCommand.ts | Browser delegates skill/generate to server |
| src/debug/jtag/commands/skill/generate/README.md | Documents skill/generate |
| src/debug/jtag/commands/skill/generate/.npmignore | Publish ignore rules for command package |
| src/debug/jtag/commands/skill/activate/shared/SkillActivateTypes.ts | Adds shared types + executor for skill/activate |
| src/debug/jtag/commands/skill/activate/package.json | New command package metadata/scripts |
| src/debug/jtag/commands/skill/activate/browser/SkillActivateBrowserCommand.ts | Browser delegates skill/activate to server |
| src/debug/jtag/commands/skill/activate/README.md | Documents skill/activate |
| src/debug/jtag/commands/skill/activate/.npmignore | Publish ignore rules for command package |
| src/debug/jtag/commands/session/get-user/server/SessionGetUserServerCommand.ts | Adjusts DataDaemon.read handling |
| src/debug/jtag/commands/logs/config/shared/LogsConfigTypes.ts | Extends logs config result shape |
| src/debug/jtag/commands/development/code/shared/CodeCommandConstants.ts | Removes legacy dev code command constants |
| src/debug/jtag/commands/development/code/read/shared/CodeReadTypes.ts | Removes legacy dev code/read types |
| src/debug/jtag/commands/development/code/read/shared/CodeReadCommand.ts | Removes legacy dev code/read base class |
| src/debug/jtag/commands/development/code/pattern-search/shared/CodeFindTypes.ts | Removes legacy dev code/find types |
| src/debug/jtag/commands/development/code/pattern-search/shared/CodeFindCommand.ts | Removes legacy dev code/find base class |
| src/debug/jtag/commands/data/read/server/DataReadServerCommand.ts | Updates DataDaemon.read return handling |
| src/debug/jtag/commands/data/list/server/DataListServerCommand.ts | Simplifies error messaging/logging |
| src/debug/jtag/commands/data/delete/server/DataDeleteServerCommand.ts | Types entityBeforeDelete and imports BaseEntity |
| src/debug/jtag/commands/data/create/server/DataCreateServerCommand.ts | Passes suppressEvents through to DataDaemon.store |
| src/debug/jtag/commands/collaboration/chat/export/server/ChatExportServerCommand.ts | Filters stored tool results from exports |
| src/debug/jtag/commands/collaboration/activity/join/server/ActivityJoinServerCommand.ts | Adds required param validation for activityId |
| src/debug/jtag/commands/collaboration/activity/create/server/ActivityCreateServerCommand.ts | Adds required param validation for recipeId |
| src/debug/jtag/commands/code/write/shared/CodeWriteTypes.ts | Adds shared types + executor for code/write |
| src/debug/jtag/commands/code/write/server/CodeWriteServerCommand.ts | Adds server implementation for code/write |
| src/debug/jtag/commands/code/write/package.json | New command package metadata/scripts |
| src/debug/jtag/commands/code/write/browser/CodeWriteBrowserCommand.ts | Browser delegates code/write to server |
| src/debug/jtag/commands/code/write/README.md | Documents code/write |
| src/debug/jtag/commands/code/write/.npmignore | Publish ignore rules for command package |
| src/debug/jtag/commands/code/verify/shared/CodeVerifyTypes.ts | Adds shared types + executor for code/verify |
| src/debug/jtag/commands/code/verify/package.json | New command package metadata/scripts |
| src/debug/jtag/commands/code/verify/browser/CodeVerifyBrowserCommand.ts | Browser delegates code/verify to server |
| src/debug/jtag/commands/code/verify/README.md | Documents code/verify |
| src/debug/jtag/commands/code/undo/shared/CodeUndoTypes.ts | Adds shared types + executor for code/undo |
| src/debug/jtag/commands/code/undo/server/CodeUndoServerCommand.ts | Adds server implementation for code/undo |
| src/debug/jtag/commands/code/undo/package.json | New command package metadata/scripts |
| src/debug/jtag/commands/code/undo/browser/CodeUndoBrowserCommand.ts | Browser delegates code/undo to server |
| src/debug/jtag/commands/code/undo/README.md | Documents code/undo |
| src/debug/jtag/commands/code/undo/.npmignore | Publish ignore rules for command package |
| src/debug/jtag/commands/code/tree/shared/CodeTreeTypes.ts | Adds shared types + executor for code/tree |
| src/debug/jtag/commands/code/tree/server/CodeTreeServerCommand.ts | Adds server implementation for code/tree |
| src/debug/jtag/commands/code/tree/package.json | New command package metadata/scripts |
| src/debug/jtag/commands/code/tree/browser/CodeTreeBrowserCommand.ts | Browser delegates code/tree to server |
| src/debug/jtag/commands/code/tree/README.md | Documents code/tree |
| src/debug/jtag/commands/code/tree/.npmignore | Publish ignore rules for command package |
| src/debug/jtag/commands/code/shell/watch/shared/CodeShellWatchTypes.ts | Adds shared types + executor for code/shell/watch |
| src/debug/jtag/commands/code/shell/watch/server/CodeShellWatchServerCommand.ts | Adds server implementation for code/shell/watch |
| src/debug/jtag/commands/code/shell/watch/package.json | New command package metadata/scripts |
| src/debug/jtag/commands/code/shell/watch/browser/CodeShellWatchBrowserCommand.ts | Browser delegates code/shell/watch to server |
| src/debug/jtag/commands/code/shell/watch/.npmignore | Publish ignore rules for command package |
| src/debug/jtag/commands/code/shell/status/shared/CodeShellStatusTypes.ts | Adds shared types + executor for code/shell/status |
| src/debug/jtag/commands/code/shell/status/server/CodeShellStatusServerCommand.ts | Adds server implementation for code/shell/status |
| src/debug/jtag/commands/code/shell/status/package.json | New command package metadata/scripts |
| src/debug/jtag/commands/code/shell/status/browser/CodeShellStatusBrowserCommand.ts | Browser delegates code/shell/status to server |
| src/debug/jtag/commands/code/shell/status/README.md | Documents code/shell/status |
| src/debug/jtag/commands/code/shell/status/.npmignore | Publish ignore rules for command package |
| src/debug/jtag/commands/code/shell/sentinel/shared/CodeShellSentinelTypes.ts | Adds shared types + executor for code/shell/sentinel |
| src/debug/jtag/commands/code/shell/sentinel/server/CodeShellSentinelServerCommand.ts | Adds server implementation for code/shell/sentinel |
| src/debug/jtag/commands/code/shell/sentinel/package.json | New command package metadata/scripts |
| src/debug/jtag/commands/code/shell/sentinel/browser/CodeShellSentinelBrowserCommand.ts | Browser delegates code/shell/sentinel to server |
| src/debug/jtag/commands/code/shell/sentinel/README.md | Documents code/shell/sentinel |
| src/debug/jtag/commands/code/shell/sentinel/.npmignore | Publish ignore rules for command package |
| src/debug/jtag/commands/code/shell/kill/shared/CodeShellKillTypes.ts | Adds shared types + executor for code/shell/kill |
| src/debug/jtag/commands/code/shell/kill/server/CodeShellKillServerCommand.ts | Adds server implementation for code/shell/kill |
| src/debug/jtag/commands/code/shell/kill/package.json | New command package metadata/scripts |
| src/debug/jtag/commands/code/shell/kill/browser/CodeShellKillBrowserCommand.ts | Browser delegates code/shell/kill to server |
| src/debug/jtag/commands/code/shell/kill/README.md | Documents code/shell/kill |
| src/debug/jtag/commands/code/shell/kill/.npmignore | Publish ignore rules for command package |
| src/debug/jtag/commands/code/shell/execute/shared/CodeShellExecuteTypes.ts | Adds shared types + executor for code/shell/execute |
| src/debug/jtag/commands/code/shell/execute/server/CodeShellExecuteServerCommand.ts | Adds server implementation for code/shell/execute |
| src/debug/jtag/commands/code/shell/execute/package.json | New command package metadata/scripts |
| src/debug/jtag/commands/code/shell/execute/browser/CodeShellExecuteBrowserCommand.ts | Browser delegates code/shell/execute to server |
| src/debug/jtag/commands/code/shell/execute/.npmignore | Publish ignore rules for command package |
| src/debug/jtag/commands/code/search/shared/CodeSearchTypes.ts | Adds shared types + executor for code/search |
| src/debug/jtag/commands/code/search/server/CodeSearchServerCommand.ts | Adds server implementation for code/search |
| src/debug/jtag/commands/code/search/package.json | New command package metadata/scripts |
| src/debug/jtag/commands/code/search/browser/CodeSearchBrowserCommand.ts | Browser delegates code/search to server |
| src/debug/jtag/commands/code/search/README.md | Documents code/search |
| src/debug/jtag/commands/code/search/.npmignore | Publish ignore rules for command package |
| src/debug/jtag/commands/code/read/shared/CodeReadTypes.ts | Adds shared types + executor for code/read |
| src/debug/jtag/commands/code/read/server/CodeReadServerCommand.ts | Adds server implementation for code/read |
| src/debug/jtag/commands/code/read/package.json | New command package metadata/scripts |
| src/debug/jtag/commands/code/read/browser/CodeReadBrowserCommand.ts | Browser delegates code/read to server |
| src/debug/jtag/commands/code/read/.npmignore | Publish ignore rules for command package |
| src/debug/jtag/commands/code/history/shared/CodeHistoryTypes.ts | Adds shared types + executor for code/history |
| src/debug/jtag/commands/code/history/server/CodeHistoryServerCommand.ts | Adds server implementation for code/history |
| src/debug/jtag/commands/code/history/package.json | New command package metadata/scripts |
| src/debug/jtag/commands/code/history/browser/CodeHistoryBrowserCommand.ts | Browser delegates code/history to server |
| src/debug/jtag/commands/code/history/README.md | Documents code/history |
| src/debug/jtag/commands/code/history/.npmignore | Publish ignore rules for command package |
| src/debug/jtag/commands/code/git/shared/CodeGitTypes.ts | Adds shared types + executor for code/git |
| src/debug/jtag/commands/code/git/package.json | New command package metadata/scripts |
| src/debug/jtag/commands/code/git/browser/CodeGitBrowserCommand.ts | Browser delegates code/git to server |
| src/debug/jtag/commands/code/git/README.md | Documents code/git |
| src/debug/jtag/commands/code/edit/server/CodeEditServerCommand.ts | Adds server implementation for code/edit |
| src/debug/jtag/commands/code/edit/package.json | New command package metadata/scripts |
| src/debug/jtag/commands/code/edit/browser/CodeEditBrowserCommand.ts | Browser delegates code/edit to server |
| src/debug/jtag/commands/code/edit/.npmignore | Publish ignore rules for command package |
| src/debug/jtag/commands/code/diff/shared/CodeDiffTypes.ts | Adds shared types + executor for code/diff |
| src/debug/jtag/commands/code/diff/server/CodeDiffServerCommand.ts | Adds server implementation for code/diff |
| src/debug/jtag/commands/code/diff/package.json | New command package metadata/scripts |
| src/debug/jtag/commands/code/diff/browser/CodeDiffBrowserCommand.ts | Browser delegates code/diff to server |
| src/debug/jtag/commands/code/diff/.npmignore | Publish ignore rules for command package |
| src/debug/jtag/commands/ai/thoughtstream/server/ThoughtStreamServerCommand.ts | Updates DataDaemon.read usage for messages/users |
| src/debug/jtag/commands/ai/status/shared/AIStatusTypes.ts | Adds toolsAvailable metric |
| src/debug/jtag/commands/ai/status/server/AIStatusServerCommand.ts | Computes toolsAvailable via PersonaToolRegistry |
| src/debug/jtag/commands/ai/should-respond-fast/server/ShouldRespondFastServerCommand.ts | Adds required param validation for messageText |
| src/debug/jtag/commands/ai/rag/inspect/server/RAGInspectServerCommand.ts | Updates DataDaemon.read usage for trigger msg |
| src/debug/jtag/cli.ts | Removes legacy code constants; updates command mapping/timeouts |
| src/debug/jtag/challenges/task-manager/task-manager.ts | Adds TaskManager challenge module |
| src/debug/jtag/challenges/task-manager/task-manager.test.ts | Adds runnable tests for TaskManager |
| src/debug/jtag/api/data-seed/RoomDataSeed.ts | Adds “Code” room seed data |
| src/debug/jtag/.gitignore | Ignores runtime data + generated bindings |
| CLAUDE.md | Documents ts-rs Rust↔TS type boundary rules |
💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.
Comment on lines
+43
to
+45
| getByPriority(priority: 'low' | 'medium' | 'high'): Task[] { | ||
| return this.tasks.filter(t => t.priority !== priority); | ||
| } |
There was a problem hiding this comment.
TaskManager filtering logic is inverted in three places: getByPriority returns tasks that are NOT the requested priority, getPending returns all tasks (including completed), and removeCompleted keeps only completed tasks. Update these filters to match the method names/tests (priority should be equality, pending should be !completed, and removeCompleted should remove where completed === true).
Comment on lines
+53
to
+62
| getPending(): Task[] { | ||
| return this.tasks; | ||
| } | ||
|
|
||
| /** Remove all completed tasks. Returns count of removed tasks. */ | ||
| removeCompleted(): number { | ||
| const before = this.tasks.length; | ||
| this.tasks = this.tasks.filter(t => t.completed); | ||
| return before - this.tasks.length; | ||
| } |
There was a problem hiding this comment.
TaskManager filtering logic is inverted in three places: getByPriority returns tasks that are NOT the requested priority, getPending returns all tasks (including completed), and removeCompleted keeps only completed tasks. Update these filters to match the method names/tests (priority should be equality, pending should be !completed, and removeCompleted should remove where completed === true).
| 'help': 'commandName', | ||
| [CODE_COMMANDS.READ]: 'path', | ||
| [CODE_COMMANDS.FIND]: 'pattern', | ||
| 'code/read': 'path', |
There was a problem hiding this comment.
The new code/read command types use filePath (not path). With the current mapping, the CLI will populate the wrong parameter name and code/read will fail validation on the server. Change the mapping to use filePath for code/read.
Suggested change
| 'code/read': 'path', | |
| 'code/read': 'filePath', |
| status: data.status ?? '', | ||
| scope: data.scope ?? '', | ||
| createdById: data.createdById ?? '', | ||
| limit: data.limit ?? 0, |
There was a problem hiding this comment.
limit is documented as defaulting to 20, and the server implementation uses const limit = params.limit ?? 20. However, the params factory forces limit to 0 when omitted, which bypasses the server default and can produce empty results. Prefer leaving limit as undefined when not provided (or default it to 20 here consistently).
Suggested change
| limit: data.limit ?? 0, | |
| limit: data.limit, |
Comment on lines
+42
to
+45
| timeoutMs: data.timeoutMs ?? 0, | ||
| ...data | ||
| }); | ||
|
|
There was a problem hiding this comment.
Defaulting timeoutMs to 0 changes runtime behavior for sync executions: the server later does ... ?? (wait ? 30000 : undefined), but 0 is not nullish, so it prevents the intended 30s default when wait=true. Consider omitting timeoutMs unless explicitly provided (or set it to 30000 when wait=true).
Suggested change
| timeoutMs: data.timeoutMs ?? 0, | |
| ...data | |
| }); | |
| ...data | |
| }); |
Comment on lines
+74
to
+76
| personaId: data.personaId ?? '', | ||
| includeGitStatus: data.includeGitStatus ?? true, | ||
| ...data |
There was a problem hiding this comment.
personaId is optional in WorkspaceListParams, but the factory converts “omitted” into an empty string. This can subtly change behavior if the server treats empty string as a real filter value rather than “no filter”. Prefer leaving personaId as undefined when not provided (and similarly avoid defaulting optional strings to '' unless the wire protocol requires it).
Suggested change
| personaId: data.personaId ?? '', | |
| includeGitStatus: data.includeGitStatus ?? true, | |
| ...data | |
| ...data, | |
| includeGitStatus: data.includeGitStatus ?? true, |
| /** | ||
| * Supported git operations. | ||
| */ | ||
| export type GitOperation = 'status' | 'diff' | 'log' | 'add' | 'commit' | 'push'; |
There was a problem hiding this comment.
CodeGitParams.operation is typed as string even though you define a stricter GitOperation union. Also, CodeGitParams re-declares userId even though it’s already part of the base CommandParams (and is frequently auto-injected). Consider typing operation: GitOperation and removing the redundant userId field from the params interface (rely on CommandParams.userId).
Comment on lines
+23
to
+27
| export interface CodeGitParams extends CommandParams { | ||
| /** Persona/workspace handle */ | ||
| userId: string; | ||
| /** Git operation to perform */ | ||
| operation: string; |
There was a problem hiding this comment.
CodeGitParams.operation is typed as string even though you define a stricter GitOperation union. Also, CodeGitParams re-declares userId even though it’s already part of the base CommandParams (and is frequently auto-injected). Consider typing operation: GitOperation and removing the redundant userId field from the params interface (rely on CommandParams.userId).
| - **compiled**: `boolean` - Whether TypeScript compilation succeeded | ||
| - **testsRun**: `number` - Number of tests executed | ||
| - **testsPassed**: `number` - Number of tests that passed | ||
| - **errors**: `object` - Array of error messages from compilation or tests |
There was a problem hiding this comment.
The SkillValidateResult.errors field is defined as string[] in shared/SkillValidateTypes.ts, but the README documents it as object. Update the README type to string[] (or “string[] (array of strings)”).
Suggested change
| - **errors**: `object` - Array of error messages from compilation or tests | |
| - **errors**: `string[]` - Array of error messages from compilation or tests |
| Returns `CodeShellStatusResult` with: | ||
|
|
||
| Returns CommandResult with: | ||
| - **sessionId**: `string` - Shell session identifier |
There was a problem hiding this comment.
The result type uses shellSessionId (explicitly named to avoid collision with CommandResult.sessionId), but the README documents the field as sessionId. Update the README to match the actual result shape (shellSessionId).
Suggested change
| - **sessionId**: `string` - Shell session identifier | |
| - **shellSessionId**: `string` - Shell session identifier |
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Summary
Major architectural improvements:
ServiceModuletrait system replaces 2,598-line IPC dispatcher with dynamic module registrationKey Changes
Rust Runtime (Phase 0-3 of plan)
ServiceModuletrait +ModuleRegistryfor command routingMessageBusfor inter-module pub/subSharedComputefor lazy-compute-once cachingRAG System
rag/composeendpointanytypes - full type safetyData Layer
Test plan
npm run build:tspassescargo checkpassesnpm startdeploys and AI personas respondrag/batchtimings)