@polymorphism-tech/morph-spec 8.2.1 → 8.3.0

package/framework/standards/ai-agents/observability-patterns.md

@@ -0,0 +1,156 @@
+# Observability Patterns — OpenTelemetry, enrichment, investigation
+
+> **Scope:** stacks=["dotnet"]
+> **Layer:** 1 (on-keyword)
+> **Keywords:** observability, telemetry, opentelemetry, logging, tracing, log investigation, troubleshooting, aspire dashboard, span, correlation id
+> **Read by Claude in:** implement (wiring telemetry into an agent) and troubleshoot (investigating a production issue)
+
+**Verified against:** MAF 1.0 GA. Last-verified: 2026-07-22.
+
+---
+
+## Por que isso importa
+
+Um agente MAF em produção não é uma função — é uma cadeia (LLM call → tool call → sub-agent → RAG/DB) que pode ter qualquer profundidade. Sem telemetria estruturada, um bug em produção ("o agente respondeu errado", "a automação não disparou") é indistinguível de uma caixa-preta: não dá pra saber qual chamada falhou, quanto cada etapa levou, ou qual tool retornou o quê. Este standard cobre as três partes do problema: **como instrumentar** (setup), **o que registrar** (enrichment) e **como investigar** (análise) — nessa ordem, porque setup sem enrichment produz traces vazios, e enrichment sem prática de investigação não ajuda ninguém no momento do incidente.
+
+---
+
+## Setup — OpenTelemetry
+
+```csharp
+using OpenTelemetry;
+using OpenTelemetry.Trace;
+using OpenTelemetry.Metrics;
+using Microsoft.Agents.AI;
+
+// Habilitar em agents individuais
+agent.WithOpenTelemetry();
+
+// Program.cs — setup completo
+builder.Services.AddOpenTelemetry()
+    .WithTracing(tracing =>
+    {
+        tracing.AddSource(AgentOpenTelemetryConsts.DefaultSourceName);   // MAF 1.0 GA source
+        tracing.AddSource(builder.Environment.ApplicationName);          // seus spans customizados (ver Enrichment)
+        tracing.AddAspNetCoreInstrumentation();
+        tracing.AddHttpClientInstrumentation();
+        tracing.AddOtlpExporter();
+    })
+    .WithMetrics(metrics =>
+    {
+        metrics.AddMeter(AgentOpenTelemetryConsts.DefaultSourceName);
+        metrics.AddAspNetCoreInstrumentation();
+        metrics.AddHttpClientInstrumentation();
+        metrics.AddRuntimeInstrumentation();
+    });
+
+builder.Logging.AddOpenTelemetry(logging =>
+{
+    logging.IncludeFormattedMessage = true;
+    logging.IncludeScopes = true;           // necessário para correlação — ver Enrichment
+    logging.AddOtlpExporter();
+});
+```
+
+> **Source name:** sempre `AgentOpenTelemetryConsts.DefaultSourceName` de `Microsoft.Agents.AI`.
+> **NÃO usar** o source string `"Experimental.[pacote-antigo].*"` — pertence à família `Microsoft.Extensions.AI` 9.x, incompatível com MAF 1.0 GA.
+> **Produção:** desabilite dados sensíveis — `EnableSensitiveTelemetryData = false` no agent — para não vazar prompt/response completos no backend de export.
+
+### Aspire Dashboard (dev/local)
+
+Visualiza em tempo real: fluxo completo de conversas entre agents, ciclos de tool call (Model → Tool → Model), token usage e latência por chamada, error traces e retry patterns. Para ambientes hospedados (Coolify), exporte via OTLP para o backend de observability do projeto — ver `infrastructure-docker-coolify-deploy` para o setup de Serilog/JSON console log que alimenta o mesmo pipeline de agregação.
+
+---
+
+## Enrichment — o que preencher em spans e logs
+
+Setup sem enrichment produz traces tecnicamente corretos mas operacionalmente inúteis — você vê "uma chamada ao LLM aconteceu", não "por que ela demorou 8s" ou "qual feature/task gerou essa chamada". Preencha sempre:
+
+### Correlação entre trace, feature e requisição
+
+```csharp
+using var activity = Activity.Current;
+activity?.SetTag("morph.feature", featureName);
+activity?.SetTag("morph.task_id", taskId);
+activity?.SetTag("morph.correlation_id", correlationId);   // mesmo ID em toda a cadeia de uma requisição de usuário
+```
+
+Sem um `correlation_id` propagado manualmente através de chamadas assíncronas/filas (Hangfire, background jobs), você não consegue religar "o usuário reclamou às 14:32" a um trace específico — o `TraceId` do OpenTelemetry sozinho não sobrevive a um boundary de fila.
+
+### Atributos de custo e uso do modelo
+
+```csharp
+activity?.SetTag("gen_ai.request.model", modelAlias);       // alias do Model Registry, nunca o nome hardcoded
+activity?.SetTag("gen_ai.usage.prompt_tokens", usage.PromptTokens);
+activity?.SetTag("gen_ai.usage.completion_tokens", usage.CompletionTokens);
+activity?.SetTag("gen_ai.response.finish_reason", finishReason);
+```
+
+Nomeação `gen_ai.*` segue a convenção semântica OpenTelemetry para GenAI — mantém compatibilidade se o backend de observability tiver dashboards prontos para essa convenção.
+
+### Logging estruturado com `ILogger` — sempre com scope
+
+```csharp
+using (_logger.BeginScope(new Dictionary<string, object>
+{
+    ["FeatureId"] = featureName,
+    ["TaskId"] = taskId,
+    ["AgentId"] = agentPersonaId
+}))
+{
+    _logger.LogInformation("Tool call {ToolName} started with args {Args}", toolName, args);
+    // ... chamada ...
+    _logger.LogInformation("Tool call {ToolName} completed in {ElapsedMs}ms", toolName, elapsed);
+}
+```
+
+- **Sempre** parâmetros nomeados no template (`{ToolName}`, não string interpolada) — permite querying estruturado no backend em vez de grep de texto livre.
+- **Nunca** logar o conteúdo completo do prompt/response em `Information` — isso é PII/dado sensível do cliente por padrão. Logar em `Debug` apenas, e só se `EnableSensitiveTelemetryData` estiver habilitado explicitamente para o ambiente.
+- **Nível de log por tipo de evento:** `Information` para início/fim de tool call e decisões do agent; `Warning` para retry/fallback de provider; `Error` apenas para falhas que geram resposta degradada ao usuário (não para toda exceção capturada e tratada).
+
+---
+
+## Investigação e análise — quando algo dá errado
+
+Workflow para diagnosticar um incidente relatado ("o agente X respondeu errado/lento/nada às HH:MM"):
+
+1. **Ancore no correlation_id ou janela de tempo.** Se o relato tem um identificador (número de WhatsApp, feature, conversationId), busque por ele nos atributos (`morph.correlation_id`, `gen_ai.*`) — não pelo texto do log. Sem identificador, restrinja pela janela de tempo + `morph.feature`/`AgentId` antes de olhar qualquer trace individual.
+2. **Suba do trace raiz para os spans filhos.** O span raiz mostra a duração total; os filhos (`Function Calling`, `IChatClient`) mostram onde o tempo/erro realmente está. Não assuma — meça. Um "agente lento" é quase sempre UM tool call específico (chamada de API externa, query sem índice), não o LLM.
+3. **Correlacione com métricas antes de teorizar.** Confira a tabela "What to Monitor" abaixo pelo período do incidente — um pico de `tool call failure rate` ou `token usage` aponta a causa mais rápido que ler traces individuais um a um.
+4. **Reproduza com Debug ligado, nunca em produção.** Se a causa não estiver clara nos atributos estruturados, habilite `Debug` (com dados sensíveis, se necessário) num ambiente de staging com o mesmo input — nunca habilite log de prompt/response completo em produção para investigar.
+5. **Documente a causa raiz em `decisions.md`** se a investigação levar a uma mudança de arquitetura (novo middleware de retry, novo timeout, novo cache) — não só no commit message.
+
+### What to Monitor
+
+| Metric | Purpose | Alert Threshold |
+|--------|---------|------------------|
+| Token usage per request | Cost control | > 10K tokens/request |
+| Latency per agent call | Performance | > 5s for simple tasks |
+| Tool call failure rate | Reliability | > 5% failures |
+| Cache hit rate | Cost optimization | < 50% (tune strategy) |
+| Model drift | Quality | Evaluation score drop > 10% |
+
+---
+
+## Checklist (verifiable by morph-eval)
+
+- [ ] OpenTelemetry configurado com `AgentOpenTelemetryConsts.DefaultSourceName` (traces + metrics + logging)
+- [ ] `EnableSensitiveTelemetryData = false` fora de dev/staging
+- [ ] Todo tool call relevante tem `correlation_id` propagado manualmente através de boundaries assíncronos (filas, background jobs)
+- [ ] Logging via `ILogger` com `BeginScope` estruturado, nunca string interpolada em templates de log
+- [ ] Nenhum log em `Information`/produção contém prompt/response completo
+- [ ] Alertas configurados para os thresholds da tabela "What to Monitor" relevantes ao projeto
+
+---
+
+## References
+
+- `ai-agents-production` — panorama de produção; esta seção era resumida lá, tratamento canônico agora aqui
+- `ai-agents-middleware-patterns` — middleware de telemetria no nível `IChatClient` (`TelemetryMiddleware`)
+- `infrastructure-docker-coolify-deploy` — Serilog/JSON console log para o pipeline de agregação em produção
+- [OpenTelemetry Semantic Conventions for GenAI](https://opentelemetry.io/docs/specs/semconv/gen-ai/)
+- [Agent Observability](https://learn.microsoft.com/agent-framework/user-guide/observability/)
+
+---
+
+*MORPH-SPEC by Polymorphism Tech — ai-agents/observability-patterns.md v1.0 (2026-07-22)*

package/framework/standards/ai-agents/production.md

@@ -179,54 +179,9 @@ public async Task<WeatherResponse> GetWeatherAsync(
 
 ## Observability / OpenTelemetry
 
-```csharp
-using OpenTelemetry;
-using OpenTelemetry.Trace;
-using OpenTelemetry.Metrics;
-using Microsoft.Agents.AI;
-
-// Enable on individual agents
-agent.WithOpenTelemetry();
-
-// Program.cs — full setup
-builder.Services.AddOpenTelemetry()
-    .WithTracing(tracing =>
-    {
-        tracing.AddSource(AgentOpenTelemetryConsts.DefaultSourceName);   // MAF 1.0 GA source
-        tracing.AddSource(builder.Environment.ApplicationName);
-        tracing.AddAspNetCoreInstrumentation();
-        tracing.AddHttpClientInstrumentation();
-        tracing.AddOtlpExporter();
-    })
-    .WithMetrics(metrics =>
-    {
-        metrics.AddMeter(AgentOpenTelemetryConsts.DefaultSourceName);    // MAF 1.0 GA meter
-        metrics.AddAspNetCoreInstrumentation();
-        metrics.AddHttpClientInstrumentation();
-        metrics.AddRuntimeInstrumentation();
-    });
-```
-
-> **Source name:** sempre `AgentOpenTelemetryConsts.DefaultSourceName` de `Microsoft.Agents.AI`.
-> **NÃO usar** o source string `"Experimental.[pacote-antigo].*"` — pertence à família `Microsoft.Extensions.AI 9.x`, incompatível com MAF 1.0 GA.
-
-### Aspire Dashboard
-
-Visualiza traces de agents incluindo:
-- Fluxo completo de conversas entre agents
-- Tool call cycles (Model → Tool → Model)
-- Token usage e latência por chamada
-- Error traces e retry patterns
-
-### What to Monitor
+Setup, enrichment (span attributes, correlation IDs, structured logging) e o workflow de investigação/análise de logs em produção têm **tratamento canônico** em **`ai-agents-observability-patterns`**. Não duplicado aqui.
 
-| Metric | Purpose | Alert Threshold |
-|--------|---------|----------------|
-| Token usage per request | Cost control | > 10K tokens/request |
-| Latency per agent call | Performance | > 5s for simple tasks |
-| Tool call failure rate | Reliability | > 5% failures |
-| Cache hit rate | Cost optimization | < 50% (tune strategy) |
-| Model drift | Quality | Evaluation score drop > 10% |
+> Em produção: source name sempre `AgentOpenTelemetryConsts.DefaultSourceName`; `EnableSensitiveTelemetryData = false` fora de dev/staging. Ver `ai-agents-observability-patterns` para o setup completo, o que preencher em cada span/log, e o workflow de investigação quando um incidente for reportado.
 
 ---
 
@@ -262,6 +217,7 @@ Visualiza traces de agents incluindo:
 - `ai-agents-sweet-spot` — decisão de arquitetura antes de middleware
 - `ai-agents-providers-model-registry` — Model Registry para `_modelRegistry.GetChatClient(...)`
 - `ai-agents-middleware-patterns` — **tratamento canônico** de middleware (código dos 3 níveis, telemetria, resilência, anti-hallucination)
+- `ai-agents-observability-patterns` — **tratamento canônico** de observability (setup OpenTelemetry, enrichment de spans/logs, workflow de investigação)
 - `ai-agents-mcp-tools` — agente MAF consumindo MCP servers externos
 - `ai-agents-mcp-server` — projeto expondo capacidades como MCP server
 - `ai-agents-rag-custom-pgvector` — implementação completa de busca semântica (referida em Caching)

package/framework/templates/REGISTRY.json

@@ -109,55 +109,6 @@
       "technology": "handlebars",
       "validatePlaceholders": true
     },
-    {
-      "id": "context",
-      "kind": "template",
-      "name": "Project Context",
-      "path": "context/CONTEXT.md",
-      "location": "framework",
-      "category": "context",
-      "phase": "proposal",
-      "required": false,
-      "description": "Project-level context with tech stack, active features, agents, standards",
-      "outputName": "CONTEXT.md",
-      "placeholders": [
-        "AGENT_COUNT",
-        "AGENT_TEAMS_DISPLAY_MODE",
-        "AGENT_TEAMS_THRESHOLDS",
-        "APP_INSIGHTS",
-        "AZURE_LOCATION",
-        "CONTAINER_ENV",
-        "COOLIFY_APP_URL",
-        "COOLIFY_SERVER",
-        "DEVOPS_AREA_PATH",
-        "DEVOPS_BUILD_PIPELINE",
-        "DEVOPS_ORG",
-        "DEVOPS_PROJECT",
-        "DEVOPS_RELEASE_PIPELINE",
-        "DEVOPS_WIKI_NAME",
-        "DOTNET_VERSION",
-        "EF_VERSION",
-        "INFERRED_STANDARDS",
-        "MORPH_VERSION",
-        "NEON_PROJECT_ID",
-        "NEON_REGION",
-        "NEXTJS_VERSION",
-        "OPENAI_MODEL",
-        "PROJECT_DESCRIPTION",
-        "PROJECT_NAME",
-        "PROJECT_TYPE",
-        "REPOSITORY_URL",
-        "SQL_DATABASE",
-        "SQL_SERVER",
-        "TIMESTAMP",
-        "UI_LIBRARY"
-      ],
-      "stackSpecific": false,
-      "engine": "handlebars",
-      "deprecated": false,
-      "technology": "handlebars",
-      "validatePlaceholders": true
-    },
     {
       "id": "architecture",
       "kind": "template",
@@ -176,50 +127,6 @@
       "technology": "handlebars",
       "validatePlaceholders": true
     },
-    {
-      "id": "context-feature",
-      "kind": "template",
-      "name": "Feature Context",
-      "path": "context/CONTEXT-FEATURE.md",
-      "location": "framework",
-      "category": "context",
-      "phase": "proposal",
-      "required": false,
-      "description": "Feature-level context with progress, agents, tasks, decisions, validations",
-      "outputName": "CONTEXT-FEATURE.md",
-      "placeholders": [
-        "ACTIVE_AGENTS_COUNT",
-        "AGENT_TEAMS_DECISION",
-        "COMPLEXITY",
-        "CONTRACTS_PATH",
-        "CURRENT_TASK",
-        "DECISIONS_PATH",
-        "DESCRIPTION",
-        "FEATURE_ID",
-        "FEATURE_NAME",
-        "MORPH_VERSION",
-        "PHASE",
-        "PHASE_DESCRIPTION",
-        "PROPOSAL_PATH",
-        "RECAP_PATH",
-        "SPEC_PATH",
-        "STATUS",
-        "TASKS_BLOCKED",
-        "TASKS_COMPLETED",
-        "TASKS_IN_PROGRESS",
-        "TASKS_PATH",
-        "TASKS_PROGRESS",
-        "TASKS_TOTAL",
-        "TEAM_LEAD",
-        "TIMESTAMP",
-        "VALIDATION_TIMESTAMP"
-      ],
-      "stackSpecific": false,
-      "engine": "handlebars",
-      "deprecated": false,
-      "technology": "handlebars",
-      "validatePlaceholders": true
-    },
     {
       "id": "spec-examples",
       "kind": "template",

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@polymorphism-tech/morph-spec",
-  "version": "8.2.1",
+  "version": "8.3.0",
   "description": "MORPH-SPEC: NLH (Natural Language Harness) for spec-driven development with Claude Code",
   "keywords": [
     "claude-code",

package/src/commands/observability/telemetry.js

@@ -0,0 +1,75 @@
+/**
+ * `morph-spec telemetry [--json] [--feature <name>] [--limit <n>]`
+ *
+ * The reader surface over the harness event stream (.morph/logs/events.jsonl,
+ * written by the telemetry-log hook). Reconstructs the whole execution path of
+ * an NLH session: top skills, personas dispatched, standards (knowledge base)
+ * read, the tool mix, and the recent prompts (user + sub-agent).
+ *
+ * Read-only, always exits 0 (fail-open — an empty stream is an answer).
+ */
+
+import chalk from 'chalk';
+import { loadTelemetry, aggregateTelemetry } from '../../lib/telemetry/telemetry-reader.js';
+
+// The knowledge-base count reflects EXPLICIT reads only. The core standards
+// load via @-refs in rules (Layer 0/1 auto-load) — those are injected by
+// Claude Code's rules system, not via the Read tool, so no hook ever sees them.
+// A low count here means "few on-demand reads", NOT "standards barely used".
+const STANDARDS_CAVEAT =
+  'Só reads explícitos (via Read) são observáveis; standards auto-carregados por @-ref nas rules (Layer 0/1) não passam por tool call e ficam invisíveis aqui.';
+
+function section(title, rows, fmt) {
+  if (!rows.length) return;
+  console.log('\n' + chalk.white.bold(title));
+  for (const r of rows) console.log('  ' + fmt(r));
+}
+
+/**
+ * @param {{ json?: boolean, feature?: string, limit?: number, cwd?: string }} [options]
+ */
+export function telemetryCommand(options = {}) {
+  const events = loadTelemetry(options.cwd || process.cwd());
+  const summary = aggregateTelemetry(events, { feature: options.feature });
+
+  if (options.json) {
+    console.log(JSON.stringify({ ...summary, notes: { standards: STANDARDS_CAVEAT } }));
+    process.exit(0);
+  }
+
+  if (summary.total === 0) {
+    const scope = options.feature ? ` para a feature "${options.feature}"` : '';
+    console.log(chalk.gray(`Telemetria vazia${scope} — nenhum evento capturado ainda em .morph/logs/events.jsonl.`));
+    process.exit(0);
+  }
+
+  const scope = options.feature ? chalk.cyan(` · feature ${options.feature}`) : '';
+  console.log(chalk.white.bold(`Telemetria do harness — ${summary.total} eventos${scope}`));
+  if (summary.features.length) {
+    console.log(chalk.gray(`Features: ${summary.features.join(', ')}`));
+  }
+
+  section('Skills', summary.skills, s => `${chalk.green(s.name.padEnd(24))} ${s.count}`);
+  section('Personas despachadas', summary.personas, p => `${chalk.magenta(p.name.padEnd(24))} ${p.count}`);
+  section('Base de conhecimento (standards lidos explicitamente)', summary.standards,
+    s => `${chalk.yellow(String(s.count).padStart(3))}×  ${s.path}`);
+  if (summary.standards.length || summary.tools.some(t => t.name === 'Read')) {
+    console.log(chalk.gray(`  ⚠ ${STANDARDS_CAVEAT}`));
+  }
+  section('Mix de tools', summary.tools, t => `${t.name.padEnd(16)} ${t.count}`);
+
+  const limit = Number.isInteger(options.limit) && options.limit > 0 ? options.limit : 10;
+  const prompts = summary.prompts.slice(-limit);
+  if (prompts.length) {
+    console.log('\n' + chalk.white.bold(`Prompts (últimos ${prompts.length})`));
+    for (const p of prompts) {
+      const tag = p.source === 'user' ? chalk.blue('user ') : chalk.magenta('agent');
+      const where = p.feature ? chalk.gray(` [${p.feature}/${p.phase || '-'}]`) : '';
+      const oneLine = String(p.text || '').replace(/\s+/g, ' ').trim();
+      const preview = oneLine.length > 120 ? oneLine.slice(0, 117) + '...' : oneLine;
+      console.log(`  ${tag}${where} ${preview}`);
+    }
+  }
+
+  process.exit(0);
+}

package/src/core/state/state-manager.js

@@ -472,7 +472,6 @@ async function ensureFeature(featureName) {
       status: "draft",
       createdAt: new Date().toISOString(),
       updatedAt: new Date().toISOString(),
-      activeAgents: [],
       approvalGates: {
         proposal: { approved: false, timestamp: null, approvedBy: null },
         uiux: { approved: false, timestamp: null, approvedBy: null },
@@ -486,7 +485,6 @@ async function ensureFeature(featureName) {
         pending: 0
       },
       checkpoints: [],
-      dispatchLog: [],
       trustConfig: {},
       contextBundles: [],
       fileChanges: []
@@ -533,7 +531,6 @@ export async function createFeature(featureName, options = {}) {
     workType: workType || 'feature',
     createdAt: now,
     updatedAt: now,
-    activeAgents: [],
     approvalGates: {
       proposal: { approved: false, timestamp: null, approvedBy: null },
       uiux: { approved: false, timestamp: null, approvedBy: null },
@@ -549,7 +546,6 @@ export async function createFeature(featureName, options = {}) {
       pending: 0
     },
     checkpoints: [],
-    dispatchLog: [],
     trustConfig: {},
     contextBundles: [],
     fileChanges: []
@@ -649,73 +645,6 @@ export async function addCheckpoint(featureName, note) {
   return checkpoint;
 }
 
-/**
- * Add agent to feature
- * @param {string} featureName - Feature name
- * @param {string} agentId - Agent ID
- * @returns {boolean} True if added, false if already exists
- */
-export async function addAgent(featureName, agentId) {
-  await ensureFeature(featureName);
-  const state = loadState(); // Load AFTER ensuring feature exists
-
-  if (!state.features[featureName].activeAgents.includes(agentId)) {
-    state.features[featureName].activeAgents.push(agentId);
-    state.features[featureName].updatedAt = new Date().toISOString();
-    saveState(state);
-    return true;
-  }
-
-  return false;
-}
-
-/**
- * Remove agent from feature
- * @param {string} featureName - Feature name
- * @param {string} agentId - Agent ID
- * @returns {boolean} True if removed, false if not found
- */
-export function removeAgent(featureName, agentId) {
-  const state = loadState();
-
-  if (!state.features[featureName]) {
-    throw new Error(`Feature '${featureName}' not found.`);
-  }
-
-  const index = state.features[featureName].activeAgents.indexOf(agentId);
-  if (index > -1) {
-    state.features[featureName].activeAgents.splice(index, 1);
-    state.features[featureName].updatedAt = new Date().toISOString();
-    saveState(state);
-    return true;
-  }
-
-  return false;
-}
-
-/**
- * Log a dispatch event to state.json.
- * Appends entry to feature.dispatchLog, capped at 50 entries (oldest trimmed).
- *
- * @param {string} featureName - Feature name
- * @param {{ timestamp: string, phase: string, mode: string, agents: Array<{ id: string, subagentType: string, group: string, tasks: string[] }> }} entry - Dispatch event to log
- */
-export function logDispatch(featureName, entry) {
-  const state = loadState();
-  const feature = state.features?.[featureName];
-  if (!feature) return;
-
-  if (!Array.isArray(feature.dispatchLog)) feature.dispatchLog = [];
-  feature.dispatchLog.push(entry);
-
-  // Cap at 50 entries
-  if (feature.dispatchLog.length > 50) {
-    feature.dispatchLog = feature.dispatchLog.slice(-50);
-  }
-
-  saveState(state);
-}
-
 /**
  * Normalize output type from kebab-case to camelCase (BUG-002 fix - Enhanced)
  * Supports both kebab-case and camelCase input formats for all output types
@@ -1103,60 +1032,3 @@ export function getApprovalHistory(featureName) {
   // Sort by timestamp
   return history.sort((a, b) => new Date(a.timestamp) - new Date(b.timestamp));
 }
-
-/**
- * Find which agent was assigned a specific task in the latest dispatch.
- * Searches only the last dispatchLog entry.
- *
- * @param {string} featureName
- * @param {string} taskId
- * @returns {string|null} Agent ID or null if not found
- */
-export function resolveAgentForTask(featureName, taskId) {
-  const state = loadState();
-  const feature = state.features?.[featureName];
-  if (!feature) return null;
-
-  const log = feature.dispatchLog;
-  if (!Array.isArray(log) || log.length === 0) return null;
-
-  const latest = log[log.length - 1];
-  for (const agent of latest.agents || []) {
-    if (Array.isArray(agent.tasks) && agent.tasks.includes(taskId)) {
-      return agent.id;
-    }
-  }
-  return null;
-}
-
-/**
- * Update timing fields on a specific agent in the latest dispatch entry.
- * Fail-open: returns silently if feature/dispatch/agent not found.
- *
- * @param {string} featureName
- * @param {string} agentId - Agent ID to update
- * @param {Object} updates - Fields to merge (startedAt, completedAt, tasksCompleted)
- */
-export function updateDispatchAgentTiming(featureName, agentId, updates) {
-  const state = loadState();
-  const feature = state.features?.[featureName];
-  if (!feature) return;
-
-  const log = feature.dispatchLog;
-  if (!Array.isArray(log) || log.length === 0) return;
-
-  const latest = log[log.length - 1];
-  const agent = (latest.agents || []).find(a => a.id === agentId);
-  if (!agent) return;
-
-  Object.assign(agent, updates);
-
-  // If all agents now have completedAt, set entry.completedAt to the latest
-  const allComplete = (latest.agents || []).every(a => a.completedAt);
-  if (allComplete && latest.agents.length > 0) {
-    const timestamps = latest.agents.map(a => new Date(a.completedAt).getTime());
-    latest.completedAt = new Date(Math.max(...timestamps)).toISOString();
-  }
-
-  saveState(state);
-}

package/src/dashboard/app/pages/LiveState.jsx

@@ -98,33 +98,6 @@ function TaskList({ taskList }) {
   );
 }
 
-function DispatchLog({ dispatchLog }) {
-  if (!dispatchLog?.length) return null;
-
-  return (
-    <div className="mt-3">
-      <p className="text-gray-500 text-[10px] uppercase tracking-wide mb-1.5">Dispatch Log</p>
-      <div className="space-y-1 border-l-2 border-gray-700 pl-3">
-        {dispatchLog.map((entry, i) => (
-          <div key={i} className="relative">
-            <span className="absolute -left-4 top-1 w-2 h-2 rounded-full bg-gray-600 border border-gray-500" />
-            <div className="text-[10px] text-gray-500">
-              {entry.timestamp && (
-                <span className="font-mono mr-2">
-                  {new Date(entry.timestamp).toLocaleTimeString([], { hour: '2-digit', minute: '2-digit' })}
-                </span>
-              )}
-              <span className="text-blue-400 font-mono">{entry.agent || entry.agentId}</span>
-              {entry.task && <span className="text-gray-400 ml-1">→ {entry.task}</span>}
-              {entry.action && <span className="text-gray-400 ml-1">{entry.action}</span>}
-            </div>
-          </div>
-        ))}
-      </div>
-    </div>
-  );
-}
-
 function FeatureCard({ featureId, feature }) {
   const status = feature.status || 'draft';
   const phase = feature.phase || feature.currentPhase || PHASE_ORDER[0];
@@ -134,7 +107,6 @@ function FeatureCard({ featureId, feature }) {
     : typeof feature.taskList === 'object' && feature.taskList !== null
       ? Object.values(feature.taskList)
       : [];
-  const dispatchLog = feature.dispatchLog || [];
 
   return (
     <div className="bg-gray-900 border border-gray-700 rounded-xl p-4 space-y-3">
@@ -169,21 +141,6 @@ function FeatureCard({ featureId, feature }) {
           <TaskList taskList={taskList} />
         </div>
       )}
-
-      {/* Dispatch log */}
-      <DispatchLog dispatchLog={dispatchLog} />
-
-      {/* Active agents */}
-      {feature.activeAgents?.length > 0 && (
-        <div>
-          <p className="text-gray-500 text-[10px] uppercase tracking-wide mb-1.5">Active Agents</p>
-          <div className="flex flex-wrap gap-1">
-            {feature.activeAgents.map((a) => (
-              <span key={a} className="px-2 py-0.5 bg-gray-800 border border-gray-700 rounded text-[10px] font-mono text-gray-300">{a}</span>
-            ))}
-          </div>
-        </div>
-      )}
     </div>
   );
 }