after install with helm chart librechat installation always return error

[Slach]

What happened?

I use following deploy approach
install ollama models with ollama-operator

kubectl apply --server-side=true -f https://github.com/nekomeowww/ollama-operator/raw/refs/heads/main/dist/install.yaml
kubectl apply -n ollama-operator-system -f librechat/ollama-models.yaml
kubectl apply -n ollama-operator-system -f librechat/ollama-services.yaml

install gateway database MCP server

rm -rf /tmp/gateway
helm pull oci://ghcr.io/centralmind/gateway-helm/gateway --destination /tmp/ --untar
kubectl create namespace gateway-system || true
helm upgrade --install gateway /tmp/gateway --namespace gateway-system --values "$CUR_DIR/gateway-values.yaml"

install librechat + librechat-rag-api

git clone https://github.com/danny-avila/LibreChat.git /tmp/LibreChat
helm repo add bitnami https://charts.bitnami.com/bitnami
helm repo add meilisearch https://meilisearch.github.io/meilisearch-kubernetes

# Create namespace for LibreChat
kubectl create namespace librechat || true

# Install LibreChat with Helm
cd /tmp/LibreChat/helm/librechat-rag-api && helm dependency build && cd -
helm upgrade --install librechat-rag-api /tmp/LibreChat/helm/librechat-rag-api \
  --namespace librechat \
  --values "$CUR_DIR/librechat-rag-api-values.yaml"

# Build Helm dependencies
cd /tmp/LibreChat/helm/librechat && helm dependency build && cd -

# Install LibreChat with Helm
helm upgrade --install librechat /tmp/LibreChat/helm/librechat \
  --namespace librechat \
  --values "$CUR_DIR/librechat-values.yaml"

librechat-values.yaml

librechat:
  configEnv:
    PLUGIN_MODELS: "devstral:latest"
    DEBUG_LOGGING: "true"
    DEBUG_CONSOLE: "true"

  configYamlContent: |
    version: "1.2.5"
    interface:
      customWelcome: "Welcome to Altinity ClickHouse LLM Prototype! Enjoy your experience."
      endpointsMenu: true
      modelSelect: true
      parameters: true
      sidePanel: true
      presets: true
      prompts: true
      bookmarks: true
      multiConvo: true
      agents: true    
    endpoints:
      custom:
      - name: ollama
        baseURL: "http://ollama-srv-devstral.ollama-operator-system.svc.cluster.local:11434"
        models:
          default: [ 'devstral:latest' ]
        agentOptions:
          model: "devstral:latest"
  
    mcpServers:
      gateway:
        type: sse
        url: "http://gateway.gateway-system.svc.cluster.local:8080/sse"
        chatMenu: true

librechat-rag-api-values.yaml

global:
  librechat:
    existingSecretName: ""
image:
  repository: danny-avila/librechat-rag-api-dev
  registry: ghcr.io
  pullPolicy: IfNotPresent
  tag: latest

rag:
  enabled: true
  configEnv:
    EMBEDDINGS_PROVIDER: ollama
    OLLAMA_BASE_URL: http://ollama-srv-graphite.ollama-operator-system.svc.cluster.local:11434/
    EMBEDDINGS_MODEL: granite-embedding:278m

all pods run

all ollama and gateway services available
and curl -vvv http://ollama-srv-graphite.ollama-operator-system.svc.cluster.local:11434/api/tags return expected model description

librechat UI run on 3080 port
kubectl port-forward -n librechat svc/librechat --address 0.0.0.0 3080

i can login with user which create with npm run create-user
but any chat message return error

inside librchat logs i see only following

2025-05-29T11:16:31.504Z debug: Custom config:
{
  version: "1.2.5",
    interface.customWelcome: "Welcome to Altinity ClickHouse LLM Prototype! Enjoy your experience.",
    interface.endpointsMenu: true,
    interface.modelSelect: true,
    interface.parameters: true,
    interface.sidePanel: true,
    interface.presets: true,
    interface.prompts: true,
    interface.bookmarks: true,
    interface.multiConvo: true,
    interface.agents: true,
    // 1 custom(s)
    endpoints.custom: [{"name":"ollama","baseURL":"http://ollama-srv-devstral.ollama-operator-system.svc.cluster.local:11434... [truncated]],
    gateway.type: "sse",
    gateway.url: "http://gateway.gateway-system.svc.cluster.local:8080/sse",
    gateway.chatMenu: true,
}
2025-05-29T11:16:31.504Z debug: [MCP][gateway] Transport sending: {"method":"tools/list","jsonrpc":"2.0","id":3}

2025-05-29T11:19:21.145Z error: [handleAbortError] AI response error; aborting request: Agent not found
2025-05-29T11:19:21.146Z debug: [AgentController] Performing cleanup
2025-05-29T11:19:21.146Z debug: [AgentController] Cleanup completed

which Agent shall I add to allow usage MCP+Ollama?

Version Information

root@gpu-01:~# minikube image ls | grep libre
ghcr.io/danny-avila/librechat:v0.7.8
ghcr.io/danny-avila/librechat-rag-api-dev:latest

Steps to Reproduce

• install ollama-operator + models
• install gateway database MCP server helm chart
• install librechat + librechat-rag-api helm charts

What browsers are you seeing the problem on?

No response

Relevant log output

## cat ./api/logs/error-2025-05-29.log

{"level":"error","message":"[handleAbortError] AI response error; 
aborting request: Agent not found",
"stack":"Error: Agent not found
    at initializeClient (/app/api/server/services/Endpoints/agents/initialize.js:319:11)
    at process.processTicksAndRejections (node:internal/process/task_queues:95:5)
    at async AgentController (/app/api/server/controlle
rs/agents/request.js:105:20)
    at async controller (/app/api/server/routes/agents/chat.js:27:3)"}

Screenshots

No response

Code of Conduct

• I agree to follow this project's Code of Conduct

[Slach]

i tried to capture traffic inside librechat container

i don't see any 11434 port related traffic
all :8080/sse traffic return 200 OK

[danny-avila]

@hofq can you look into this?

[Slach]

@danny-avila @hofq could you suggest what exactly wrong with my librechat.yaml ? how to properly define agent which will connect to gateway MCP and OLLAMA models?

[hofq]

i'm not familiar with mcp support in Librechat, but as there is no seperate container i don't see how this has anything to do with the helm chart.

Correct me if i'm wrong tho @danny-avila

[Slach]

@hofq ok. this is maybe not helm chart related issue

guys, could you try look to my config and just suggest what i did get wrong? or it will just silence instead of answers?

[Slach]

@hofq @danny-avila how to setup agent in librechat.yaml properly? why provided helm chart can't

why following librechat.yaml config

endpoints:
  agents:
    disableBuilder: false
    maxRecursionLimit: 100
    recursionLimit: 50
  custom:
  - agentOptions:
      model: devstral:latest
    baseURL: http://ollama-srv-devstral.ollama-operator-system.svc.cluster.local:11434
    models:
      default:
      - devstral:latest
    name: ollama
interface:
  agents: true
  bookmarks: true
  customWelcome: Welcome to Altinity ClickHouse LLM Prototype! Enjoy your experience.
  endpointsMenu: true
  modelSelect: true
  multiConvo: true
  parameters: true
  presets: true
  prompts: true
  sidePanel: true
mcpServers:
  gateway:
    chatMenu: true
    type: sse
    url: http://gateway.gateway-system.svc.cluster.local:8080/sse
version: 1.2.5

always
return

{"level":"error","message":"[handleAbortError] AI response error; 
aborting request: Agent not found",
"stack":"Error: Agent not found
    at initializeClient (/app/api/server/services/Endpoints/agents/initialize.js:319:11)
    at process.processTicksAndRejections (node:internal/process/task_queues:95:5)
    at async AgentController (/app/api/server/controlle
rs/agents/request.js:105:20)
    at async controller (/app/api/server/routes/agents/chat.js:27:3)"}

[danny-avila]

@Slach you need to create an Agent to use the "Agents" feature.

[Slach]

@danny-avila can i define agent via librechat.yaml? Why when I try to create agent manually i don't see my properly provided ollama endpoint?
could you look to screenshot?

[Slach]

@danny-avila any suggestions?