Add ManufacturingAgentSuite example

.pre-commit-config.yaml

@@ -82,6 +82,7 @@ repos:
     rev: v1.7.7
     hooks:
       - id: docformatter
+        language_version: python3.12
         args: [
             --in-place,
             --wrap-summaries=0, # 0 means disable wrap

ManufacturingAgentSuite/README.md

@@ -0,0 +1,68 @@
+# ManufacturingAgentSuite
+
+`ManufacturingAgentSuite` is a proposed OPEA GenAIExamples blueprint for
+route-isolated industrial action-card agents.
+
+The example demonstrates how one Gateway and Manufacturing Megaservice can
+route plant-floor evidence to five governed manufacturing workflows:
+
+| Route         | Workflow                                       | Output target            |
+| ------------- | ---------------------------------------------- | ------------------------ |
+| `maintenance` | Predictive maintenance / lao-shi-fu escalation | `maintenance_work_order` |
+| `iqc`         | Incoming and in-process quality control        | `qms_quality_event`      |
+| `changeover`  | SKU changeover verification                    | `changeover_checklist`   |
+| `wi`          | Released work-instruction guidance             | `wi_reference`           |
+| `hazard`      | EHS hazard observation                         | `ehs_case`               |
+
+## Architecture
+
+```text
+Plant evidence
+  -> Gateway
+  -> Manufacturing Megaservice
+  -> route registry: maintenance / iqc / changeover / wi / hazard
+  -> route-specific source evidence
+  -> deterministic evaluator
+  -> guardrails
+  -> bounded action card
+```
+
+The full WearEdge reference implementation also includes Qdrant RAG, OPEA
+embedding profiles, an official OPEA TEI path, benchmark evidence, and a browser
+demo console:
+
+```text
+https://github.com/davidmillerak2026-sys/wearedge-opea-manufacturing
+```
+
+## Quick Start On Xeon
+
+```bash
+cd ManufacturingAgentSuite/docker_compose/intel/cpu/xeon
+docker compose up -d
+curl http://localhost:8899/v1/agents
+curl http://localhost:8899/v1/agents/maintenance/demo
+curl http://localhost:8899/v1/scorecard
+```
+
+Optional OPEA TEI profile:
+
+```bash
+docker compose -f compose.yaml -f compose.opea-tei.yaml up -d
+```
+
+## Endpoints
+
+| Endpoint                       | Purpose                                         |
+| ------------------------------ | ----------------------------------------------- |
+| `GET /healthz`                 | Service health and configured embedding profile |
+| `GET /v1/agents`               | Route registry                                  |
+| `GET /v1/agents/{mode}/demo`   | Fixed demo request and bounded action card      |
+| `POST /v1/agents/{mode}/infer` | Route-specific inference contract               |
+| `GET /v1/scorecard`            | Five-route validation scorecard                 |
+
+## Guardrail Boundary
+
+The example must not claim autonomous restart, quality release, maintenance
+release, safety clearance, final root cause, customer acceptance, or remaining
+useful life. Restricted decisions remain human-confirmed.

ManufacturingAgentSuite/assets/flow.md

@@ -0,0 +1,33 @@
+# ManufacturingAgentSuite Flow
+
+```mermaid
+flowchart LR
+    Evidence["Plant-floor evidence"] --> Gateway["Gateway"]
+    Gateway --> Mega["Manufacturing Megaservice"]
+    Mega --> Registry["Route registry"]
+    Registry --> Maintenance["maintenance"]
+    Registry --> IQC["iqc"]
+    Registry --> Changeover["changeover"]
+    Registry --> WI["wi"]
+    Registry --> Hazard["hazard"]
+    Maintenance --> Eval["Evaluator"]
+    IQC --> Eval
+    Changeover --> Eval
+    WI --> Eval
+    Hazard --> Eval
+    Eval --> Guardrails["Guardrails"]
+    Guardrails --> Action["Bounded action card"]
+```
+
+OPEA component mapping:
+
+| OPEA concept  | ManufacturingAgentSuite role                                     |
+| ------------- | ---------------------------------------------------------------- |
+| Gateway       | Plant evidence/API entry point                                   |
+| Megaservice   | Route orchestration                                              |
+| Dataprep      | Route-specific manuals, quality plans, policies, and checklists  |
+| Retriever/RAG | Source-grounded evidence retrieval in the full reference package |
+| Vector DB     | Qdrant profile in the full reference package                     |
+| LLM service   | Pluggable LLM adapter; deterministic path for CI                 |
+| Guardrails    | Blocked claims and human-confirmation gates                      |
+| Evaluation    | Route scorecard                                                  |

ManufacturingAgentSuite/benchmark/README.md

@@ -0,0 +1,26 @@
+# ManufacturingAgentSuite Benchmark Notes
+
+The first PR benchmark should remain CI-friendly:
+
+```bash
+cd ManufacturingAgentSuite/docker_compose/intel/cpu/xeon
+docker compose up -d
+../../../../tests/test_compose_on_xeon.sh
+```
+
+The reference WearEdge package includes deeper evidence:
+
+- Intel Xeon AVX-512/AMX deterministic five-route CPU benchmark.
+- Google Cloud C3 Docker/Qdrant fresh-clone E2E.
+- Google Cloud C3 OPEA-compatible embedding profile E2E.
+- Google Cloud C3 official OPEA TEI profile E2E.
+
+Reference evidence:
+
+```text
+https://github.com/davidmillerak2026-sys/wearedge-opea-manufacturing/tree/main/evidence/benchmarks
+```
+
+Do not use this first PR to claim production LLM acceleration. The current
+hardware evidence is for the deterministic route pipeline and official OPEA TEI
+embedding path.

ManufacturingAgentSuite/docker_compose/intel/cpu/xeon/README.md

@@ -0,0 +1,53 @@
+# Deploying ManufacturingAgentSuite on Intel Xeon
+
+This directory contains a minimal Docker Compose deployment for the proposed
+`ManufacturingAgentSuite` OPEA example.
+
+## Start
+
+```bash
+docker compose up -d
+```
+
+## Validate
+
+```bash
+curl http://localhost:8899/healthz
+curl http://localhost:8899/v1/agents
+curl http://localhost:8899/v1/agents/maintenance/demo
+curl http://localhost:8899/v1/scorecard
+```
+
+Expected scorecard result:
+
+```json
+{
+  "ok": true,
+  "routes": [
+    { "mode": "maintenance", "status": "pass" },
+    { "mode": "iqc", "status": "pass" },
+    { "mode": "changeover", "status": "pass" },
+    { "mode": "wi", "status": "pass" },
+    { "mode": "hazard", "status": "pass" }
+  ]
+}
+```
+
+## Optional OPEA TEI Profile
+
+```bash
+docker compose -f compose.yaml -f compose.opea-tei.yaml up -d
+```
+
+This starts Hugging Face TEI and the OPEA embedding microservice pattern with:
+
+```text
+TEI_EMBEDDING_ENDPOINT=http://tei-embedding-service:80
+EMBEDDING_COMPONENT_NAME=OPEA_TEI_EMBEDDING
+```
+
+## Stop
+
+```bash
+docker compose down
+```

ManufacturingAgentSuite/docker_compose/intel/cpu/xeon/compose.opea-tei.yaml

@@ -0,0 +1,40 @@
+# Copyright (C) 2026 Intel Corporation
+# SPDX-License-Identifier: Apache-2.0
+
+services:
+  manufacturing-agent-suite:
+    environment:
+      OPEA_EXAMPLE_PROFILE: opea-tei
+      OPEA_EMBEDDING_PROFILE: opea-tei
+      OPEA_EMBEDDING_URL: http://embedding:6000/v1/embeddings
+    depends_on:
+      qdrant:
+        condition: service_started
+      embedding:
+        condition: service_started
+
+  tei-embedding-service:
+    image: ghcr.io/huggingface/text-embeddings-inference:cpu-1.7
+    container_name: manufacturing-agent-suite-tei-embedding
+    ports:
+      - "6006:80"
+    volumes:
+      - "${MODEL_CACHE:-./data}:/data"
+    shm_size: 1g
+    command: --model-id ${EMBEDDING_MODEL_ID:-BAAI/bge-base-en-v1.5} --auto-truncate
+    restart: unless-stopped
+
+  embedding:
+    image: ${REGISTRY:-opea}/embedding:${TAG:-latest}
+    container_name: manufacturing-agent-suite-embedding
+    depends_on:
+      tei-embedding-service:
+        condition: service_started
+    ports:
+      - "6000:6000"
+    environment:
+      TEI_EMBEDDING_ENDPOINT: http://tei-embedding-service:80
+      EMBEDDING_COMPONENT_NAME: OPEA_TEI_EMBEDDING
+      HF_TOKEN: ${HF_TOKEN:-}
+      LOGFLAG: ${LOGFLAG:-False}
+    restart: unless-stopped

ManufacturingAgentSuite/docker_compose/intel/cpu/xeon/compose.yaml

@@ -0,0 +1,33 @@
+# Copyright (C) 2026 Intel Corporation
+# SPDX-License-Identifier: Apache-2.0
+
+services:
+  manufacturing-agent-suite:
+    image: python:3.12-slim
+    container_name: manufacturing-agent-suite
+    working_dir: /app
+    command: python /app/manufacturing_agent_suite.py
+    ports:
+      - "8899:8899"
+    environment:
+      OPEA_EXAMPLE_PROFILE: deterministic
+      OPEA_VECTOR_BACKEND: qdrant-profile
+      OPEA_EMBEDDING_PROFILE: deterministic
+    volumes:
+      - "../../../../manufacturing_agent_suite.py:/app/manufacturing_agent_suite.py:ro"
+    depends_on:
+      qdrant:
+        condition: service_started
+    restart: unless-stopped
+
+  qdrant:
+    image: qdrant/qdrant:v1.12.6
+    container_name: manufacturing-agent-suite-qdrant
+    ports:
+      - "6333:6333"
+      - "6334:6334"
+    restart: unless-stopped
+
+networks:
+  default:
+    driver: bridge