Unclassified

L0.5 Operating Layer Wiring Plan

Created: 2026-06-09
Status: IN PROGRESS
Authority: Nirmal (Founder) approval required for deployment

Executive Summary

This document outlines the systematic wiring of the L0.5 Operating Layer (Cloudflare Durable Execution substrate) to transform IntegrateWise from a reactive application into an autonomous operating system.

L0.5 Components:

  • Cloudflare Workflows → Durable, multi-step, stateful execution
  • Cloudflare Queues → Async processing with guaranteed delivery + DLQs
  • Cron Triggers → Scheduled, policy-bounded automation
  • Durable Objects → Stateful long-lived coordination (HITL gates, watchers)

Current State Analysis (Updated 2026-06-09 Phase 2)

✅ CONNECTED (Working)

  • pipeline-process queue → integratewise-pipeline consumer
  • signals queue → integratewise-intelligence consumer
  • intelligence-events queue → integratewise-intelligence consumer
  • knowledge-ingest queue → integratewise-knowledge consumer
  • memory-consolidation → integratewise-knowledge consumer
  • ops-dlq → integratewise-intelligence consumer
  • connector-sync queue → connector-sync consumer (webhook-driven + hourly cron)
  • intelligence-act queue → producer binding added (Phase 2) ✅

⚠️ NEEDS MANUAL WIRING (Phase 2 Code Deployed)

  • continuity-triage queue → consumer code deployed, needs Dashboard wiring ⏳

⚠️ CRON GAPS

  • connector-sync: Was 6 hours, needs webhook-driven (hourly backup)
  • continuity: Has handler, missing morning brief + backup crons
  • intelligence: No cron for morning brief or signal sweeps

⚠️ MISSING WORKFLOWS

  • TriageWorkflow → exists in code, needs activation
  • SyncWorkflow → missing, needed for connector delta
  • SignalWorkflow → missing, needed for spine→signal automation
  • MorningBriefWorkflow → missing, needed for 7am brief
  • ActWorkflow → missing, needed for approved→execute

TASK 1: FIX CRON TRIGGERS

1.1 connector-sync: Hourly + Webhook-Driven

File: services/connector-sync/wrangler.toml

Change:

# OLD
# crons = ["0 */6 * * *"]  # Every 6 hours

# NEW
[triggers]
crons = ["0 * * * *"]  # Every hour (backup sweep)
# Real-time syncs triggered by Nango webhooks at /webhooks/nango

Rationale: Nango webhooks (/webhooks/nango) provide real-time sync triggers. Hourly cron is backup sweep for missed events.

Status: ✅ DONE

1.2 continuity: Morning Brief + Daily Backup

File: services/continuity/wrangler.toml

Change:

# OLD
[triggers]
crons = ["0 */6 * * *"]  # TTL cleanup only

# NEW
[triggers]
crons = [
  "0 7 * * *",      # Morning brief at 7am
  "0 0 * * *",      # Daily reindex + backup at midnight
  "0 */6 * * *"     # TTL cleanup every 6 hours
]

Handler Logic:

async scheduled(event: ScheduledEvent, env: Env): Promise<void> {
  const cron = event.cron;

  if (cron === "0 7 * * *") {
    // Morning brief: trigger MorningBriefWorkflow
    // TODO: Wire to WORKFLOW service binding
  }
  else if (cron === "0 0 * * *") {
    // Daily backup: D1 → R2 + Vectorize reindex
    // TODO: Replace with AI Search (PRIORITY 4)
  }
  else if (cron === "0 */6 * * *") {
    // TTL cleanup (existing)
    await handleScheduled(event, env);
  }
}

Status: ✅ DONE (wrangler.toml), ⏳ PENDING (handler impl)

1.3 intelligence: Morning Brief + Signal Sweep

File: services/intelligence/wrangler.toml

Change:

# NEW
[triggers]
crons = [
  "0 7 * * *",      # Morning brief at 7am
  "0 */6 * * *"     # Signal analysis sweep every 6 hours
]

Status: ✅ DONE

TASK 2: FIX QUEUE WIRING

2.1 continuity-triage Queue

Problem: Producers exist (20), no consumer wired

Solution: Add consumer to continuity worker

File: services/continuity/wrangler.toml

[[queues.consumers]]
queue = "continuity-triage"
max_batch_size = 10
max_batch_timeout = 30

Status: ⏳ PENDING

2.2 intelligence-act Queue

Problem: Consumer exists (intelligence), nothing produces to it

Solution: Wire Act service to publish to queue after HITL approval

File: services/intelligence/src/act/*.ts

// After HITL approval in govern flow:
await env.INTELLIGENCE_ACT_QUEUE.send({
  action_id: proposal.id,
  action_type: proposal.action_type,
  tenant_id: proposal.tenant_id,
  approved_by: approval.user_id,
  approved_at: new Date().toISOString(),
});

Status: ⏳ PENDING (needs queue producer binding)

2.3 connector-sync Queue

Problem: Consumer exists, nothing produces

Solution: Nango webhook handler triggers sync jobs

File: services/webhook-ingress/src/index.ts

Already wired! Nango webhook handler sends to connector-sync queue:

if (eventType === "connection.created" || eventType === "sync.success") {
  await env.CONNECTOR_SYNC_QUEUE.send({
    tenant_id: tenantId,
    provider: provider,
    sync_type: "delta",
    triggered_by: "webhook",
  });
}

Status: ✅ ALREADY WIRED (verify producer binding exists)

2.4 execution-outcomes Queue

Problem: Completely disconnected (0 producers, 0 consumers)

Decision:

  • Option A: Delete the queue (unused)
  • Option B: Wire to Act service for execution tracking

Recommendation: DELETE - execution tracking happens via D1 action_executions table

Status: ⏳ PENDING DECISION

TASK 3: CREATE AI SEARCH INSTANCE

3.1 Current L4 Library Architecture

Current Flow:

Knowledge content → R2 bucket → Manual vectorization → Vectorize index → D1 cache

Problems:

  • Manual indexing required
  • No built-in search API
  • No MCP endpoint
  • No chat completions interface

3.2 AI Search Migration

New Flow:

Knowledge content → AI Search index → /search API + /chat/completions + MCP

Benefits:

  • Managed indexing (auto-sync from R2/D1)
  • Built-in search and chat APIs
  • MCP-compatible endpoint
  • Versioning and rollback
  • Hybrid search (vector + keyword)

3.3 Implementation Steps

Step 1: Create AI Search Index

wrangler ai-search create integratewise-knowledge \
  --namespace=knowledge \
  --description="IntegrateWise L4 Library - org_memory + KB" \
  --embedding-model=@cf/baai/bge-base-en-v1.5

Step 2: Wire to Knowledge Worker

File: services/knowledge/wrangler.toml

[[ai_search_namespaces]]
binding = "AI_SEARCH"
namespace = "integratewise-knowledge"

Step 3: Migrate Content

// Bulk upload from existing org_memory
const documents = await fetchOrgMemory();
await env.AI_SEARCH.upsert(
  documents.map((doc) => ({
    id: doc.id,
    text: doc.content_md,
    metadata: {
      title: doc.title,
      tags: doc.tags,
      topics: doc.topics,
      governance_state: doc.governance_state,
    },
  }))
);

Step 4: Update Search Endpoints

File: services/knowledge/src/index.ts

// OLD: Manual Vectorize query
app.get("/search", async (c) => {
  const results = await vectorizeQuery(c.env.VECTOR_INDEX, query);
  // ...
});

// NEW: AI Search API
app.get("/search", async (c) => {
  const results = await c.env.AI_SEARCH.search(query, {
    limit: 10,
    includeVectors: false,
  });
  // ...
});

Status: ⏳ PENDING (AI Search in beta, verify availability)

TASK 4: DEPLOY MISSING WORKFLOWS

4.1 TriageWorkflow

Status: ✅ EXISTS in services/intelligence/src/workflows/triage-workflow.ts

Binding: ✅ ALREADY CONFIGURED in intelligence/wrangler.toml

Action: Verify it's being invoked correctly

4.2 Missing Workflows

MorningBriefWorkflow

  • Purpose: Generate 7am brief for /desk
  • Triggers: 7am cron (continuity, intelligence)
  • Steps:
    1. Fetch overnight signals
    2. Query Entity 360 changes
    3. Generate brief via AI
    4. Store in D1 cache
    5. Broadcast to connected clients

Status: ⏳ TODO

SignalWorkflow

  • Purpose: Spine change → signal generation → broadcast
  • Triggers: Pipeline writes, hourly sweep
  • Steps:
    1. Detect spine changes
    2. Run signal analyzer
    3. Create signal records
    4. Broadcast to workspaces

Status: ⏳ TODO

ActWorkflow

  • Purpose: Approved proposal → execute action
  • Triggers: HITL approval, intelligence-act queue
  • Steps:
    1. Fetch approved proposal
    2. Resolve target tool
    3. Execute via connector
    4. Track outcome
    5. Update proposal status

Status: ⏳ TODO

Deployment Sequence

Phase 1: Cron Fixes (Safe, No Breaking Changes) — ✅ DEPLOYED

  1. ✅ connector-sync: hourly cron (wrangler.toml)
  2. ✅ continuity: +7am, +midnight crons (wrangler.toml)
  3. ✅ intelligence: +7am, +6h crons (wrangler.toml)
  4. ✅ Implement continuity scheduled handler
  5. ✅ Implement intelligence scheduled handler
  6. ✅ DEPLOYED 2026-06-09 (manual cron wiring required via Dashboard)

Phase 2: Queue Wiring (Requires Testing) — ✅ DEPLOYED

  1. ✅ Add continuity-triage consumer (wrangler.toml + handler)
  2. ✅ Wire intelligence-act producer binding
  3. ✅ Verify connector-sync webhook flow (already wired)
  4. ✅ Delete execution-outcomes queue (done)
  5. ✅ DEPLOYED 2026-06-09 (manual queue consumer wiring required via Dashboard)

See: L05_PHASE2_DEPLOYMENT_SUMMARY.md for full deployment details.

Phase 3: AI Search (Requires Beta Access) — ⏳ NEXT

  1. Add continuity-triage consumer
  2. Wire intelligence-act producer
  3. Verify connector-sync webhook flow
  4. Delete execution-outcomes queue
  5. Deploy + test each queue independently

Phase 3: AI Search (Requires Beta Access)

  1. Create AI Search index
  2. Bulk upload org_memory content
  3. Update search endpoints
  4. Test MCP integration
  5. Deprecate manual Vectorize flow

Phase 4: Workflows (Requires Development)

  1. Implement MorningBriefWorkflow
  2. Implement SignalWorkflow
  3. Implement ActWorkflow
  4. Wire cron triggers
  5. Test durable execution + retries

Verification Commands

Check Deployed Crons

wrangler deployments list --name integratewise-connector-sync
wrangler deployments list --name continuity
wrangler deployments list --name intelligence

List Queues

wrangler queues list

Check Queue Consumers

wrangler queues consumer list continuity-triage
wrangler queues consumer list intelligence-act

Test Webhook Flow

curl -X POST 'https://ingress.dev.integratewise.ai/webhooks/nango' \
  -H 'Content-Type: application/json' \
  -d '{"type":"sync.success","connectionId":"iw-customer-zero","provider":"hubspot"}'

Success Criteria

L0.5 Fully Operational When:

  • ✅ All cron triggers deployed and running
  • ✅ All queue consumers wired with producers
  • ✅ Workflows executing durably with retries
  • ✅ AI Search replacing manual vectorization
  • ✅ Morning brief generated daily at 7am
  • ✅ Signal analysis sweep running every 6 hours
  • ✅ Connector sync webhook-driven with hourly backup
  • ✅ Act workflow executing approved proposals

Next Actions

PHASE 1 COMPLETE — READY FOR DEPLOYMENT:

All Phase 1 code changes are complete:

  1. ✅ Cron triggers added to wrangler.toml files (connector-sync, continuity, intelligence)
  2. ✅ Continuity scheduled handler implemented (7am brief, midnight backup, 6h cleanup)
  3. ✅ Intelligence scheduled handler implemented (7am brief, 6h signal sweep)

AWAITING NIRMAL APPROVAL TO DEPLOY:

# Unset Cloudflare API token (critical - see AGENTS.md Hard Rules)
unset CLOUDFLARE_API_TOKEN

# Deploy Phase 1 changes
wrangler deploy --env="" services/connector-sync
wrangler deploy --env="" services/continuity
wrangler deploy --env="" services/intelligence

AFTER PHASE 1 DEPLOYMENT:

  1. Monitor cron triggers in Cloudflare dashboard
  2. Verify scheduled handlers execute correctly
  3. Proceed to Phase 2: Queue Wiring

PHASE 2 QUEUE WIRING (After Phase 1 approval):

  1. Add continuity-triage consumer
  2. Wire intelligence-act producer
  3. Verify connector-sync webhook flow
  4. Delete execution-outcomes queue

PHASE 3 AI SEARCH (After Phase 2):

  1. Request AI Search beta access
  2. Create AI Search index
  3. Bulk upload org_memory content
  4. Update search endpoints

PHASE 4 WORKFLOWS (After Phase 3):

  1. Develop MorningBriefWorkflow
  2. Develop SignalWorkflow
  3. Develop ActWorkflow
  4. Wire workflow triggers

Document Status: Phase 1 READY FOR DEPLOYMENT
Last Updated: 2026-06-09T13:15:00Z
Next Review: After Phase 1 deployment + monitoring