Agentra + WisdomAI — Enterprise Agentic & Knowledge Platform

One industry-agnostic platform, two products: WisdomAI (extremely optimized enterprise RAG) and Agentra (multi-agent industry packs). LLM-agnostic, truly multi-tenant, built for mid-to-large enterprise across white- and blue-collar work.

Version 3.0.0-draft
Date 2026-06-12
Status Full platform blueprint (detailed)
Companion agentra-wisdomai-platform-design.md · decisions-log.md

Executive Summary

Agentra + WisdomAI is a single enterprise platform sold as two products separated by entitlement, not by build. WisdomAI is a permissions-aware knowledge assistant over an extremely optimized RAG pipeline — the low-risk land product. Agentra adds multi-agent industry packs that act over enterprise systems of record — the expand product, grounded on WisdomAI.

Grounded in a teardown of today's leaders — multi-agent (Microsoft Agent Framework, Vertex ADK, Bedrock AgentCore, Agentforce, LangGraph, Sierra, Moveworks, Decagon, Aisera), RAG (Glean, Vectara, Cohere, Contextual AI, Pinecone, Elastic, LlamaCloud, Azure/Vertex/AWS search, Voyage), gateways (LiteLLM, Portkey, Kong, Cloudflare), channels (Twilio, Genesys, Slack, Teams, WhatsApp), and vertical agents (project44, Sabre, Kantata, Zebra). The recurring lesson — "one contract, many adapters" — is the spine.

USP 1

Extremely optimized RAG

Contextual chunk embeddings + hybrid retrieval + cross-encoder reranking + agentic query planning + permissions-aware retrieval + cite-or-abstain grounding — continuously benchmarked. The stack is the moat.

USP 2

Industry packs

Production multi-agent packs for E-commerce, Travel, Logistics, NBFC, Retail & Professional Services — across white-collar and blue-collar/frontline workflows over the real systems of record.

At a Glance

2
Products, one platform
6
Industry packs
45+
Specialist agents (across packs)
15+
Channels (digital·voice·frontline)
∞
LLM providers (gateway-agnostic)
3
Tenancy tiers (pool·bridge·silo)
10
Architectural tenets
P95<3s
WisdomAI grounded answer
PropertyValue
Model accessOpenLLM Gateway — LLM-agnostic (Anthropic·OpenAI·Vertex·Bedrock·Mistral·self-hosted vLLM/NIM)
Embeddings / rerankVoyage voyage-context-3 (contextual) · Voyage rerank-2.5 / Cohere Rerank 3.5 (cross-encoder)
Vector storePinecone (serverless, namespaces) · Weaviate (one-shard-per-tenant, silo)
Data storesPostgreSQL (RLS) · Redis · Neo4j (graph) · ClickHouse (traces) · object storage
IntegrationMCP (tools/data) + A2A (agent peers) + tool gateway; read/write-tagged
Message bus / meshApache Kafka · Istio strict mTLS
IdentityAuth0 (external) · Okta/Entra (workforce) · SCIM · per-agent identity
TenancyBridge default · Silo upgrade (BYOK/CMK, region-pinned)
DeployKubernetes + Helm + Argo CD GitOps; SaaS · BYOC · self-hosted
ComplianceSOC 2 · ISO 27001 · GDPR · PCI-DSS + pack overlays (RBI, DOT/EU261, FMCSA)

Two Products, One Platform

ProductWhat it isTierCompetes withMetering
WisdomAIPermissions-aware knowledge assistant / optimized RAG with citations. The land product.knowledgeGlean-classseats · queries
AgentraMulti-agent industry packs that act over systems of record, by vertical. The expand product.agenticAisera / Sierra / Moveworksresolved actions · packs

A tenant lands on WisdomAI, then upgrades by purchasing Agentra packs — hot-loaded into the same data plane with no redeploy. Entitlement (tier + licensed packs) gates the mode; read-only tools on the knowledge tier, write/financial tools added on the agentic tier.

Architectural Principles

T1 · One contract, many adapters. OpenAI-shaped models, one channel envelope, MCP tools, signed tenant claim.
T2 · Control/data-plane split, live config. Routes, budgets, entitlements push live — no redeploy.
T3 · Tenant context at every stateful hop. Signed claim mandatory in every data/gateway call.
T4 · Durable, resumable agent state. Checkpoint each step; pause for approval at zero compute.
T5 · Two-tier memory, async consolidation. Raw session + background-extracted long-term memory.
T6 · Guardrails in parallel, below reasoning. Safety/policy/PII as supervisor agents.
T7 · Cite-or-abstain grounding. No claim without a citation; abstain over guess.
T8 · Permissions are architectural. ACL filters + pre-LLM re-check; never trust the model.
T9 · Trajectory-level observability. Trace every step; simulate pre-launch; command center live.
T10 · Graceful degradation. Every component has a fallback; frontline enforces compliance.

System Context

Agentra + WisdomAI sits between three populations — external customers and internal employees across many channels, ops/admin users across consoles, and the third-party ecosystem (systems of record, partner agents) exposed entirely through MCP and A2A — with all model calls flowing through the OpenLLM Gateway.

flowchart LR
  subgraph Req["Requesters"]
    R1["External customers"]
    R2["Internal employees"]
    R3["Frontline / blue-collar"]
  end
  subgraph Ch["Channel Layer"]
    CH["Adapters → one envelope"]
  end
  subgraph Plat["Platform — per tenant data plane"]
    direction TB
    OR["Orchestration Core"]
    WIS["WisdomAI RAG (USP1)"]
    AGT["Agentra Packs (USP2)"]
    MEM["Memory"]
  end
  subgraph Ext["Systems of Record via MCP / A2A"]
    E1["OMS·WMS·TMS·PSS·GDS"]
    E2["Payments·LOS·LMS·Bureaus"]
    E3["CRM·POS·PSA·KB"]
    E4["Partner agents (A2A)"]
  end
  subgraph GW["OpenLLM Gateway"]
    G1["LLM-agnostic routing"]
  end
  subgraph CP["Control Plane"]
    P1["Provisioning"]
    P2["Capability Registry"]
    P3["Identity Auth0+Okta"]
    P4["Entitlements·Vault"]
  end
  R1 & R2 & R3 --> CH --> OR
  OR --> WIS & AGT
  WIS & AGT <--> MEM
  AGT <--> Ext
  WIS <--> Ext
  WIS & AGT --> GW
  CP -.-> Plat
  classDef r fill:#1a3a5e,stroke:#6ea8fe,color:#e6edf3
  classDef p fill:#1f2937,stroke:#6ea8fe,color:#e6edf3
  classDef e fill:#3b2e4a,stroke:#c79bf0,color:#e6edf3
  classDef c fill:#1f3a2f,stroke:#7ee0c0,color:#e6edf3
  class R1,R2,R3 r
  class OR,WIS,AGT,MEM,CH p
  class E1,E2,E3,E4 e
  class P1,P2,P3,P4,G1 c
    
Three requester populations · channel layer · per-tenant data plane · MCP/A2A ecosystem · LLM-agnostic gateway · global control plane.

Layered Architecture

flowchart TB
  EXP["EXPERIENCE — customer apps · employee apps · consoles · Frontline Mobile Shell"]
  CHN["CHANNEL LAYER — adapters → normalized InboundMessage (web·voice·IVR·WhatsApp·Slack·Teams·SMS·handheld·kiosk·push)"]
  ORC["ORCHESTRATION CORE — Triage → Context → Router/Supervisor → [Knowledge | Agentic] · durable state · guardrail supervisors"]
  WIS["WISDOMAI RAG ENGINE (USP1) — ingest·index·retrieve·rerank·ground·cite"]
  AGT["AGENTRA PACK RUNTIME (USP2) — specialist agents · reusable core · frontline shell"]
  INT["INTEGRATION — MCP servers (read/write-tagged) · A2A · tool gateway · capability registry"]
  GW["OPENLLM GATEWAY — LLM-agnostic · routing · fallback · cache · guardrails · virtual keys"]
  DAT["DATA & PLATFORM — vector · Postgres · Redis · graph · trace · bus · KMS/Vault · tenancy · security · obs"]
  EXP --> CHN --> ORC
  ORC --> WIS
  ORC --> AGT
  WIS --> INT
  AGT --> INT
  INT --> GW
  INT --> DAT
  GW --> DAT
  classDef l fill:#1f2937,stroke:#6ea8fe,color:#e6edf3
  classDef u fill:#16302a,stroke:#7ee0c0,color:#e6edf3
  class EXP,CHN,ORC,INT,GW,DAT l
  class WIS,AGT u
    
Two execution modes share one codebase: Knowledge mode (WisdomAI, read-only) and Agentic mode (Agentra, write + HITL). The mode is a configuration, never a fork.

OpenLLM Gateway — LLM-Agnostic Model Access

Every model call — classifier, RAG synthesis, specialist reasoning, supervisor, guardrails, embeddings, reranking — flows through one gateway presenting an OpenAI-shaped contract and routing to whichever provider/model fits the task, tenant, cost, and latency. Built on an open core (LiteLLM/Portkey-class) wrapped with our multi-tenant control plane.

flowchart LR
  CL["Caller (OpenAI-shaped)"] --> A1["1 AuthN/Z
virtual key → tenant/scope"] A1 --> A2["2 Rate + budget
(Redis)"] A2 --> A3["3 INPUT guards
injection·moderation·PII"] A3 --> A4["4 Cache
exact → semantic (per-tenant)"] A4 --> A5["5 Router
cost·latency·complexity·pinned"] A5 --> A6["6 Provider adapter"] A6 --> P1["Anthropic"] A6 --> P2["OpenAI"] A6 --> P3["Vertex"] A6 --> P4["Bedrock"] A6 --> P5["self-hosted vLLM/NIM"] A6 --> A7["7 Resilience
retry·fallback·cooldown·LB"] A7 --> A8["8 OUTPUT guards + cache write"] A8 --> A9["9 Stream SSE + trace + spend"] classDef s fill:#1f2937,stroke:#6ea8fe,color:#e6edf3 classDef pv fill:#3b2e4a,stroke:#c79bf0,color:#e6edf3 class A1,A2,A3,A4,A5,A6,A7,A8,A9 s class P1,P2,P3,P4,P5 pv
Stateless data plane; control plane pushes routing/budgets/keys live. Provider keys never leave the vault; self-hosted models are just another adapter — true LLM-agnosticism.
Call typeDefault routingWhy
Triage / intentfast/cheap or on-cluster classifierhigh volume, latency-critical
RAG answer synthesismid-tier, grounding-tunedquality + cost balance
Supervisor / complex reasoningfrontier modelthe one frontier call per request
Guardrail / threat detectionsmall specialized classifierparallel, cheap, independent
Embeddings / rerankdedicated providers (Voyage/Cohere)retrieval quality

WisdomAI RAG Engine USP 1

The base product and grounding substrate of every agent. A superset of the strongest techniques found, ordered by impact. Compounding effect: contextual embeddings (~35% fewer retrieval failures) + cross-encoder reranking cut top-20 failure ~67% in published benchmarks; the grounding validator turns recall into trustworthy cited answers; permissions-aware retrieval makes it safe for regulated enterprises.

flowchart LR
  subgraph IN["Ingest"]
    I1["Connectors
content + ACL"] I2["Layout-aware parse"] I3["Parent-child chunk"] I4["Contextual embed"] I5["Index (vector + sparse)"] end subgraph QY["Query"] Q1["Rewrite / expand / plan"] Q2["ACL pre-filter"] Q3["Hybrid retrieve
dense+sparse RRF"] Q4["Multi-vector (opt)"] Q5["Cross-encoder rerank"] Q6["Pre-LLM permission re-check"] Q7["Grounded synthesis
cite-or-abstain"] Q8["Grounding validator"] end I1-->I2-->I3-->I4-->I5 Q1-->Q2-->Q3-->Q4-->Q5-->Q6-->Q7-->Q8 I5-.->Q3 classDef i fill:#16302a,stroke:#7ee0c0,color:#e6edf3 classDef q fill:#1f2937,stroke:#6ea8fe,color:#e6edf3 class I1,I2,I3,I4,I5 i class Q1,Q2,Q3,Q4,Q5,Q6,Q7,Q8 q
Permissions-aware (T8) and cite-or-abstain (T7) by construction. A continuous eval harness scores faithfulness, context precision/recall, and answer relevancy in CI.

Retrieval quality

  • Contextualized chunk embeddings (voyage-context-3)
  • Hybrid dense+sparse fused via RRF
  • Cross-encoder reranking (highest-ROI lever)
  • Agentic query planning / decomposition (multi-hop)
  • Corrective / Self-RAG self-correction

Enterprise safety

  • Index-time ACL capture (content + ACL together)
  • Unified principal store across all sources
  • Query-time ACL push-down filter (primary)
  • Pre-LLM permission re-check (defense in depth)
  • Cite-or-abstain + grounding validator

Channel Layer

A channel-agnostic runtime behind per-channel adapters and one normalized envelope. A differentiator: polished digital channels for white-collar users and rugged, voice-first, offline-tolerant channels for frontline/blue-collar workers.

flowchart LR
  subgraph DIG["Digital"]
    D1["Web · Mobile SDK"]
    D2["Email"]
    D3["Slack · Teams"]
    D4["WhatsApp·RCS·Apple"]
  end
  subgraph VOX["Voice"]
    V1["SIP / Media Streams"]
    V2["IVR"]
  end
  subgraph FL["Frontline"]
    F1["Handheld / scanner"]
    F2["In-cab / field app"]
    F3["Kiosk · SMS · push"]
  end
  AD["Channel Adapters
auth · normalize · identity · rich-element xlate"] EN["InboundMessage envelope
tenant-stamped · PII-scrubbed"] RT["Orchestration Core"] D1 & D2 & D3 & D4 --> AD V1 & V2 --> AD F1 & F2 & F3 --> AD AD --> EN --> RT classDef d fill:#1f2937,stroke:#6ea8fe,color:#e6edf3 classDef v fill:#3a2f24,stroke:#f0a868,color:#e6edf3 classDef f fill:#16302a,stroke:#7ee0c0,color:#e6edf3 classDef a fill:#3b2e4a,stroke:#c79bf0,color:#e6edf3 class D1,D2,D3,D4 d class V1,V2 v class F1,F2,F3 f class AD,EN a
One envelope for every channel. Cross-channel: identity linking, idempotency/signature verify, delivery reconciliation, WhatsApp/RCS/Apple template lifecycle, per-channel rate governors, multilingual/vernacular.
ChannelTransportSync/asyncRich formatTier
Web chat / mobile SDKHTTPS + WS/SSE (offline queue)SyncCards, quick replies, nativewhite + blue
EmailSMTP/IMAP/GraphAsyncHTML emailwhite
Slack / TeamsEvents API / Bot FrameworkSync-ishBlock Kit / Adaptive Cardswhite (employee)
WhatsApp / RCS / Apple MfBCloud API / RBM / MSPAsyncInteractive templates, rich cardswhite + blue
SMSCarrier via TwilioAsyncPlain textblue
Voice / SIPMedia Streams / Conversation Relay / AudioHookSync (real-time)STT/TTS, barge-inwhite + blue
IVRSIP + DTMF/speechSyncAudio menusblue
Warehouse handheld / scannerFirst-party SDK, offline replayAsync-ishScan/native UIblue
In-cab / field app · kiosk · pushSDK / HTTPS / APNs-FCMSync / asyncVoice/PTT, touch, notificationsblue

Multi-Agent Orchestration

Industry-agnostic core: Triage → Context Assembly → Router/Supervisor, with parallel guardrail supervisors and an async memory agent. Supervisor-router is the default; deterministic Sequential/Parallel/Loop workflows where predictability beats autonomy (NBFC KYC, travel refunds). The sequence below shows both modes.

sequenceDiagram
  participant CH as Channel
  participant OR as Orchestrator
  participant TR as Triage
  participant CTX as Context Assembly
  participant GATE as Mode Gate
  participant WIS as WisdomAI RAG
  participant SUP as Supervisor
  participant SP as Specialists (parallel)
  participant GRD as Guardrail Supervisors
  participant MEM as Memory
  CH->>OR: InboundMessage
  OR->>TR: classify
  TR-->>OR: intent · urgency · locale
  OR->>CTX: assemble (memory + identity)
  OR->>GATE: read entitlement
  alt knowledge tier (WisdomAI)
    GATE->>WIS: retrieve + rerank + ground
    WIS-->>OR: cited answer (or abstain)
  else agentic tier (Agentra)
    GATE->>SUP: dispatch (pack)
    SUP->>SP: Promise.allSettled (parallel)
    SP->>WIS: ground via RAG
    SP-->>SUP: AgentResult[]
    SUP->>SUP: synthesize + apply policies
    SUP->>GRD: validate (policy·PII·threat)
    GRD-->>SUP: pass / block / HITL
    SUP-->>OR: ResolutionDirective (+ actions)
  end
  OR->>CH: OutboundMessage (rich)
  OR-)MEM: persist (async)
    
Solid = synchronous; dashed = async. Knowledge mode answers; agentic mode fans out specialists, grounds on WisdomAI, guards, and acts with HITL on irreversible steps.

Durable runs (T4)

State checkpointed each step; HITL gates pause for days at zero compute; resume exactly-once on any worker.

Agent contract

Role · Data · Actions · Guardrails · Channel + capability manifest. Behavior authored as declarative instructions + policies.

Action authorization (T6)

Write tools need idempotency keys; financial side-effects double-gated; irreversible actions route to HITL — never auto-executed.

Integration Layer — MCP-First

Every external system is an MCP server (tool/data) or A2A peer; agents never import vendor SDKs. A tool gateway does auth + MCP↔API translation. Every tool is read/write-tagged — the single tag that drives the WisdomAI/Agentra entitlement gate and the audit/HITL rules.

flowchart LR
  subgraph AG["Specialist Agent / WisdomAI"]
    A1["execute()"]
    A2["mcp.callTool()"]
  end
  subgraph MC["MCP Client (core)"]
    M1["Discover via registry"]
    M2["Tenant credential resolve (Vault)"]
    M3["sideEffects gate (read/write/financial)"]
    M4["Audit · idempotency · retry · mTLS"]
  end
  subgraph SV["MCP Server (per vendor)"]
    S1["Vendor auth (OAuth refresh)"]
    S2["Rate limit · circuit breaker"]
    S3["Tool handler"]
  end
  VEN["Vendor API
OMS·PSS·GDS·LOS·POS·PSA…"] AG --> MC --> SV --> VEN classDef a fill:#1f2937,stroke:#6ea8fe,color:#e6edf3 classDef c fill:#16302a,stroke:#7ee0c0,color:#e6edf3 classDef s fill:#3b2e4a,stroke:#c79bf0,color:#e6edf3 classDef v fill:#3a2424,stroke:#f87171,color:#e6edf3 class A1,A2 a class M1,M2,M3,M4 c class S1,S2,S3 s class VEN v
Uniform auth, error contract, audit, rate-limit, and read/write gating. Adding a vendor = deploy an MCP server + registry entry; the capability registry hot-reloads so packs/connectors activate without redeploy.

Industry Packs USP 2

A pack turns the core into a vertical solution: specialist agents, intent taxonomy, RAG corpus, connectors to systems of record, policies, and compliance. Reusable core agents (Support, Returns/Claims, Exception, Knowledge/SOP, Document/OCR, Outreach, Resourcing) and a reusable Frontline Mobile Shell (offline-first, voice/PTT, multilingual, geo+photo capture) are built once; packs add connectors + pack-specific agents and compliance. Cards below show each agent with its collar, required systems, and P95 latency budget.

white = knowledge worker / digital · blue = frontline. Side-effect class shown where it drives gating: read (WisdomAI tier) · write · financial (double-gated).

Agentra

E-commerce Pack

Support / WISMO white
Order lookup, tracking, explain delay, escalate with summary.
OMS · carrier · helpdeskread · 900ms
Returns & Refunds white
Eligibility, RMA + label, refund, serial-returner fraud check.
OMS · payment gwfinancial · 1500ms
Order Reliability white
Detect split/OOS/address exceptions; proactive fix before customer notices.
OMS · WMSwrite · 1200ms
Catalog / Merchandising white
PDP copy/attributes/alt-text; cross-channel price/inventory mismatch.
commerce platformwrite · 1500ms
Warehouse Advisor blue
Slotting, pick-path, cycle-count guidance to the floor.
WMSread · 700ms
Last-Mile Delivery blue
Dynamic routing, photo POD, reattempt/redirect, customer ETA ping.
carrier / last-milewrite · 900ms

Systems: Shopify/commercetools · OMS (Manhattan/Fluent/Sterling) · WMS (Manhattan/Blue Yonder/Körber) · Stripe/Adyen · FedEx/UPS/Onfleet/Bringg. Compliance: PCI-DSS (tokenize), consumer return-rights, TCPA/CAN-SPAM.

Agentra

Travel Pack

Booking Servicing white
PNR lookup/modify/cancel; deterministic fare-rule evaluation (never asks LLM to interpret rules).
GDS · PSSwrite · 1500ms
Disruption / IRROPS white
Classify cause, proactively reaccommodate, duty-of-care. The marquee use case.
PSS · aviation realtimewrite · 1500ms
Refunds & Compensation white
Fare-rule refund + EU261/DOT eligibility; GDS+PSS+payment three-way orchestration.
GDS · PSS · paymentfinancial · 1500ms
Loyalty white
Balance, tier, redemption, missing-credit; tier-aware prioritization.
loyalty platformwrite · 700ms
Ground / Airport Ops blue
Gate/turn coordination, baggage exception, standby/upgrade clearing.
DCS · ops platformwrite · 1200ms
At-Counter Copilot blue
Surface reaccommodation options, queue-busting during IRROPS.
PSS · DCSread · 900ms

Systems: Amadeus/Sabre/Travelport · PSS (Altéa/SabreSonic) · DCS · loyalty. Compliance: DOT, EU261, IATA/PNR/APIS, PCI, ACAA. Fare-rule determinism via deterministic workflows.

Agentra

Logistics Pack

Freight Procurement white
Rate-shop, carrier select, tender.
TMSwrite · 1500ms
Visibility / Check-Calls white
Status, ETA, proactive delay alerts; automates check-calls.
RTTVP · TMSread · 800ms
Execution Recovery white
Detect in-transit exception, reroute / re-tender.
TMS · RTTVPwrite · 1300ms
Claims (OS&D) white
Over/short/damaged intake, documentation, liability, settlement.
TMS · ERPfinancial · 1200ms
Driver Assist blue
In-cab voice routing, detention/dwell capture, POD; HOS-aware, hands-free.
ELD/telematics · TMSwrite · 900ms
Yard / Dock & Dispatch blue
Dock scheduling, yard slotting; load assignment, driver match, route-risk.
YMS · WMS · ELDwrite · 1100ms

Systems: TMS (project44/Oracle OTM/Blue Yonder/McLeod) · RTTVP (project44/FourKites) · ELD (Samsara/Motive/Geotab) · ERP. Compliance: FMCSA HOS/ELD, hazmat, customs (CBP/C-TPAT) — enforced (T10).

Agentra

NBFC / Lending Pack

Loan Origination white
Guided application, auto-fill from docs, drop-off reduction.
LOSwrite · 1500ms
KYC / AML white
ID extraction, Aadhaar/PAN verify, AML screen, dedupe.
KYC/AML · eKYCwrite · 1800ms
Underwriting Assist white
Bureau pull + bank-statement/alt-data, explainable risk grade.
bureaus · LOSread · 2000ms
Digital Collections white
Risk-segmented vernacular reminders, promise-to-pay, payment links.
LMS · paymentsfinancial · 1200ms
Field Collections Copilot blue
Visit plan, RBI-compliant script, geo-tag + receipt, cash reconciliation.
LMS · paymentsfinancial · 900ms
Field Verification blue
FI/residence/business checklist, photo/geo-stamp, on-site OCR, verdict to LOS.
LOS · KYCwrite · 1000ms
Heaviest compliance pack

RBI Digital Lending Directions 2025 (consent/disclosure/AI governance through recovery), data localization (borrower data resident in India → region-pinned gateway routing + silo tenancy), RBI Fair Practices Code (collection-hours/language/anti-harassment the agent enforces), DPDP Act 2023, CKYC, adverse-action explainability. Systems: Lentra/Jocata (LOS) · Finacle/FLEXCUBE/Finflux · CIBIL/CRIF · Aadhaar/Karza/Signzy/Perfios · UPI/NACH.

Agentra

Retail Pack

Omnichannel Service white
Order/return/loyalty across channels.
OMS · POS · CRMwrite · 900ms
Merchandising & Planning white
Assortment, markdown, planogram, demand forecast.
merch/planningwrite · 1800ms
Store Associate Copilot blue
Conversational SOP/job-aid + product Q&A (RAG).
KB · catalogread · 700ms
In-Aisle Sales blue
Real-time availability, pricing, cross/up-sell, endless aisle at the customer's side.
inventory · OMS · POSwrite · 900ms
Inventory / Shelf blue
Restock detection, cycle count, planogram check.
inventory · ESLwrite · 1000ms
BOPIS / Fulfillment blue
Pick-from-store, click-&-collect staging, substitutions.
OMS · inventorywrite · 1000ms

Systems: POS (Xstore/Square/Shopify) · Oracle Retail/Blue Yonder/Manhattan Active · CRM/clienteling (Salesforce/Tulip) · Zebra/Honeywell devices. Compliance: PCI at POS, age-restricted-sales, ADA, predictive-scheduling.

Agentra

Professional Services Pack

Proposal / RFP white
Proposal from delivery history with case studies, risks, optimal human–AI resource mix.
PSA · CRM · knowledgewrite · 2000ms
Resourcing / Staffing white
Detect over-allocation/timeline/budget risk; recommend/execute staffing moves.
PSAwrite · 1500ms
Knowledge Management white
Firm-wide projects/deliverables searchable intelligence (RAG).
knowledge reposread · 800ms
Time & Billing white
Auto-draft timesheets, leakage detection, draft invoices.
PSA · ERPfinancial · 1500ms
Project Delivery / PM white
Status rollups, RAID, milestone & margin alerts.
PSAwrite · 1200ms
Client Comms white
Drafts updates, QBR prep, meeting-to-action capture.
CRM · collaborationwrite · 1500ms

Systems: PSA (Kantata/Workday/Certinia/Rocketlane) · CRM · iManage/SharePoint/Confluence · NetSuite/Workday. Compliance: per-engagement confidentiality (sub-tenancy), client data-residency, SOC2/ISO, AI-deliverable IP.

Blue-collar design principles (cross-pack)

Enforced by the Frontline Mobile Shell so every pack inherits them: rugged/low-end Android & wearables; voice-first / barcode / push-to-talk, glanceable, interruptible; offline-first with background sync; multilingual/vernacular by default; onboarding-light with cross-shift context; and compliance-in-the-loop — the frontline agent enforces regulatory scripts/limits (RBI collection hours, HOS legality, age-restricted sales), not merely assists.

Canonical End-to-End Flows

1 · WisdomAI knowledge query knowledge mode

Scenario: employee asks in Slack — "What's our refund SLA for enterprise customers in EMEA?"

sequenceDiagram
  participant EM as Employee (Slack)
  participant CH as Slack Adapter
  participant OR as Orchestrator
  participant TR as Triage
  participant WIS as WisdomAI RAG
  participant ACL as Permissions
  EM->>CH: refund SLA enterprise EMEA?
  CH->>OR: InboundMessage (internal, Okta id)
  OR->>TR: classify
  TR-->>OR: intent=knowledge.policy_lookup
  OR->>ACL: resolve principals (Okta groups)
  OR->>WIS: rewrite/expand → ACL pre-filter
  WIS->>WIS: hybrid retrieve + cross-encoder rerank
  WIS->>ACL: pre-LLM permission re-check
  WIS->>WIS: grounded synthesis + grounding validator
  WIS-->>OR: cited answer (or abstain)
  OR->>CH: Block Kit answer + source links
    
No specialists, no writes — the entire base product. Below support threshold → abstain, not guess (T7).

2 · Agentra e-commerce return agentic mode

Scenario: customer in web chat — "I want to return my damaged headphones from order #12345."

sequenceDiagram
  participant CU as Customer (web)
  participant OR as Orchestrator
  participant TR as Triage
  participant SUP as Supervisor
  participant RET as Returns & Refunds
  participant FRA as Fraud / Risk
  participant WIS as WisdomAI (policy)
  participant GRD as Guardrails
  participant OMS as OMS (MCP)
  participant PAY as Payment (MCP)
  CU->>OR: return damaged #12345
  OR->>TR: classify (sentiment -0.6)
  OR->>SUP: dispatch (pack=ecommerce)
  par parallel specialists
    SUP->>RET: eligibility + RMA
    RET->>OMS: getOrder / getReturnPolicy
    SUP->>FRA: serial-returner check
    SUP->>WIS: damage-return policy (cited)
  end
  SUP->>SUP: synthesize ResolutionDirective
  SUP->>GRD: policy (refund ≤ threshold?) + PII
  GRD-->>SUP: pass (within threshold)
  RET->>OMS: createReturn (write)
  RET->>PAY: refund (financial · idempotent · double-gate)
  SUP-->>OR: directive
  OR->>CU: RMA + label + refund (cited policy)
    
WisdomAI grounds the policy; Supervisor coordinates; financial action is double-gated and audited (§8.5).

3 · NBFC field collection blue-collar · agentic

Scenario: a field collections agent logs an overdue-EMI visit in the offline-capable Frontline Shell.

sequenceDiagram
  participant FA as Field Agent (offline shell)
  participant CH as Field-app Adapter
  participant OR as Orchestrator
  participant SUP as Supervisor
  participant FCC as Field Collections Copilot
  participant POL as RBI Compliance
  participant LMS as LMS (MCP)
  participant PAY as Payment (MCP)
  FA->>CH: log visit (queued offline)
  CH->>OR: on reconnect → InboundMessage (hi-IN)
  OR->>SUP: dispatch (pack=nbfc)
  SUP->>FCC: handle field visit
  FCC->>POL: enforce contact-hours + vernacular script + anti-harassment
  POL-->>FCC: compliant script (Hindi)
  FCC->>LMS: getAccount (read) → history
  FCC->>PAY: recordCollection (financial · cash recon · HITL on amount)
  FCC-->>OR: receipt + geo/photo to LMS
  OR->>CH: idempotent replay if connectivity drops
    
Region-pinned routing keeps borrower data in India; compliance is enforced infrastructure, not prompt advice (T10).

Latency Budgets

The Supervisor enforces per-agent budgets via deadlines; an over-budget specialist is dropped with a partial_result flag, never silently extended.

WisdomAI knowledge path — target P95 < 3s

HopComponentP50P95Notes
1Channel ingress30 ms50 msparse, PII, identity
2Triage100 ms180 msfast / on-cluster classifier
3Query understanding120 ms220 msrewrite/expand (+plan if compound)
4ACL pre-filter + hybrid retrieve180 ms320 msnamespace-scoped
5Cross-encoder rerank150 ms280 mstop-N
6Pre-LLM permission re-check30 ms60 mscandidate re-verify
7Grounded synthesis700 ms1100 msmid-tier model, single call
8Grounding validator200 ms350 msper-claim scoring
9Channel egress50 ms90 msrender rich
End-to-end1.6 s2.6 s~400 ms slack on 3s SLO

Agentra agentic path — target P95 < 6s

HopComponentP50P95Notes
1Channel ingress30 ms50 ms
2Triage120 ms200 mspack intent taxonomy
3Context Assembly250 ms400 msslowest memory/vendor read wins
4Supervisor dispatch50 ms80 msin-process
5Specialists (parallel)1300 ms1800 msmid-tier LLM + MCP + WisdomAI grounding
6Supervisor synthesis800 ms1200 msfrontier model, single call
7Guardrail supervisors150 ms280 msparallel: policy + PII + threat
8Action execution400 ms900 msMCP write + double-gate if financial
9Channel egress60 ms100 ms
End-to-end3.2 s5.0 s~1s slack; HITL pauses off-clock
HITL is off the latency clock

Approval gates pause the durable run (T4) — wall-clock to a human decision is not counted against the SLO. Gateway caching (semantic + exact) and WisdomAI answer caching materially reduce P50 on repeat queries.

Failure Modes & Graceful Degradation

Every component declares a fallback. The safe default is queue-for-human or honest abstention; hallucination is treated as worse than unavailability (T10).

ComponentFailureDetectionGraceful fallback
Gateway — providermodel/provider outageerror / latency SLAordered fallback chain to alternate provider; cooldown
Gateway — budgettenant over budgetRedis budget checksoft cap warns; hard cap → cheaper model or queue; never silent overspend
WisdomAI retrievalweak/empty contextCorrective-RAG scorereformulate / re-retrieve; if still weak → abstain
WisdomAI groundingunsupported claimvalidator < thresholddrop / correct / abstain — never ship ungrounded
Permissionsstale principal storefreshness checklate-binding live source check; deny on uncertainty
Specialist agentover latency budgetgRPC deadlinedrop result, set partial_result; synthesize from rest
Specialist agentintegration unhealthyregistry health + breakerroute intent to handoff; agent goes inactive (no code change)
Supervisor synthesisfrontier LLM failureerror after fallbackcold handoff with auto-assembled packet
Write / financial actionpartial failureidempotency + double-gateidempotent retry; persistent → human queue + audit alert
Frontline shellconnectivity lossoffline detectionoffline queue + idempotent replay; no data loss
Compliance guardrailpolicy violation attemptguardrail supervisorblock action; explain; route to human; audit

True Multi-Tenancy

Tiered bridge (default) with a silo upgrade path; a signed tenant_id claim is mandatory at every stateful hop. Used by NBFC (data localization), professional services (per-client confidentiality), and regulated travel/retail.

Data storePool (standard)Bridge (mid)Silo (enterprise)
RelationalShared + RLS by tenant_idSchema-per-tenantDB-per-tenant
Vector storeNamespace + JWT filterNamespace + dedicated indexDedicated index/cluster
Object storagePer-tenant prefix, IAM-scopedBucket-per-tenantDedicated bucket + KMS
Encryption keysPer-tenant DEKPer-tenant KEK in KMSBYOK/CMK (revocable)
LLM gatewayVirtual key + budget + cache ns+ per-tenant guardrail policy+ region-pinned / dedicated keys
ComputeShared, quota-governedShared + reservedDedicated compute / region

Automated provisioning as one workflow; gateway-enforced noisy-neighbor budgets; sub-tenancy for per-engagement isolation (ProServ). Entitlement (tier + packs) drives the mode gate and hot-loads packs with no redeploy.

Security & Compliance

Memory

Two-tier with async consolidation: working memory (Redis, raw session, hot path) + long-term semantic memory (vector, background-extracted, never blocking the live turn) + a customer/knowledge graph (Neo4j) for entity relationships and cross-channel/cross-pack history, plus a dynamic example store for few-shot steering. Per-tenant isolated; feeds both WisdomAI context and Agentra history-aware actions.

Observability & Closed-Loop Quality

Trajectory-level tracing (every tool call, reasoning step, retrieval, agent hop) via OpenTelemetry → ClickHouse. RAG eval harness (faithfulness, context precision/recall, relevancy) + trajectory evals in CI. A simulation test center generates synthetic edge cases per pack pre-launch; a live command center shows tool firings, failures, latency/cost/CSAT post-launch. Traces feed a human-gated closed-loop pipeline (prompt revisions, fine-tune candidates, RAG-gap reports).

Architectural Decisions (v3)

ADRDecision
032"One contract, many adapters" spine (model · channel · tool · tenant)
033OpenLLM Gateway — mandatory, LLM-agnostic model access on an open core
034WisdomAI optimized RAG pipeline (USP 1)
035Industry packs + reusable core agents + Frontline Mobile Shell (USP 2)
036Detailed channel layer: digital + voice + frontline behind one envelope
037Tiered bridge→silo multi-tenancy; signed tenant claim enforced everywhere

Carried forward & binding from v2: ADR-027 (service-experience positioning; autonomy deferred), ADR-028 (generalized requester), ADR-030 (WisdomAI + Agentra productization), ADR-031 (read/write tool tagging).

Deliberately deferred

Fully autonomous / unattended agents, scheduling, event triggers, durable multi-hour autonomous workflows, non-human/agent identity + OBO. The platform keeps a human requester in the loop or a human approving the action — focus stays on the two USPs.