docs(kagent): Add Milvus resource constraints and WSL2 memory prerequisites for local Kind setup

Problem

Following kagent-feast-mcp/README.md

Step 1, deploying Milvus via Helm on a local Kind cluster running on WSL2 with ≤8GB RAM causes the Linux OOM killer to terminate the entire WSL VM.

Root cause chain:

1. The Milvus standalone Helm chart deploys 3 containers simultaneously: milvus-etcd, milvus-minio, and milvus-standalone
2. Without resource constraints, these containers request unbounded memory, baseline ~4-6GB combined
3. On WSL2 with default .wslconfig (50% of host RAM or ~4-8GB), this memory spike triggers the kernel OOM killer
4. The OOM killer reaps the WSL integration service itself, severing the Host-to-VM connection
5. The error surfaces as: Error code: Wsl/Service/0x8007274c — no Kubernetes logs are captured

Who this blocks: Any contributor on Windows/WSL2 with ≤16GB host RAM attempting to reproduce the Kagent PoC locally.

Reproduction

# 1. Start Kind cluster
kind create cluster --name docs-agent

# 2. Deploy Milvus per README Step 1
helm repo add zilliztech https://zilliztech.github.io/milvus-helm/
helm repo update
helm upgrade --install milvus zilliztech/milvus -n docs-agent \
  --set cluster.enabled=false \
  --set standalone.enabled=true \
  --set etcd.replicaCount=1 \
  --set etcd.persistence.enabled=false \
  --set minio.mode=standalone \
  --set minio.replicas=1 \
  --set pulsar.enabled=false \
  --set pulsarv3.enabled=false

# → WSL VM becomes unresponsive within 30-60s
# → Windows may hard restart
# → Error: Wsl/Service/0x8007274c

Proposed Fix

1. Add a values-local.yaml for constrained environments

Provide a Helm values override file that caps resource requests for local development:

# manifests/overlays/dev/values-local.yaml (or kagent-feast-mcp/values-local.yaml)
standalone:
  resources:
    requests:
      memory: "512Mi"
      cpu: "250m"
    limits:
      memory: "1Gi"
      cpu: "500m"

etcd:
  resources:
    requests:
      memory: "128Mi"
      cpu: "100m"
    limits:
      memory: "256Mi"
      cpu: "200m"

minio:
  resources:
    requests:
      memory: "128Mi"
      cpu: "100m"
    limits:
      memory: "256Mi"
      cpu: "200m"

Usage:

helm upgrade --install milvus zilliztech/milvus -n docs-agent \
  -f values-local.yaml \
  --set cluster.enabled=false \
  --set standalone.enabled=true \
  # ... rest of flags

2. Add WSL2 prerequisites to README

Add to kagent-feast-mcp/README.md Prerequisites section:

Windows/WSL2 Users

Ensure your .wslconfig allocates sufficient memory:

# %USERPROFILE%\.wslconfig
[wsl2]
memory=4GB
swap=2GB

For hosts with ≤8GB RAM, use the constrained Milvus Helm values:

helm upgrade --install milvus zilliztech/milvus -n docs-agent \
  -f values-local.yaml \
  # ... flags from Step 1

Validated

Successfully deployed and sustained all 3 Milvus pods (Running, 1/1) on an 8GB Windows host with memory=4GB WSL limit using the constrained values above. The pods survived host sleep/wake cycles and WSL restarts without triggering OOMKill.

Files

• kagent-feast-mcp/README.md — Prerequisites section + Step 1
• New file: kagent-feast-mcp/values-local.yaml (or appropriate location per repo convention)

Context

Discovered during local Kagent PoC validation — #59 (comment)

Start with kagent-feast-mcp/README.md, especially the Prerequisites section and Step 1, then check the repository convention for the proposed kagent-feast-mcp/values-local.yaml location. Done means the README documents WSL2 memory prerequisites and constrained Helm usage, and the values file supports the local Milvus deployment described in the issue.