Setup Context Engine

Follow this guide to setup, install, and begin using the Context Engine in On Prem environments.

Context Engine documentation

See the full Context Engine documentation for more details on the Context Engine.

Prerequisites

Infrastructure Requirements

The machine running the Context Engine components should have at least:

• vCPU: 8

• 16 GB RAM

PostgreSQL Instance (Recommended Size):

• vCPU: 4

• Memory: 32 GB

• Disk: 100 GB

You may opt for smaller instances if working with a relatively small codebase.

PostgreSQL Version: 17

Required Configuration:

• pg_vector extension installed

  Note: This is pre-installed in managed cloud deployments.

• TCP port 5432 must be accessible from the Kubernetes cluster where the Context Engine will be deployed

• Two separate databases:

  • rag-indexer

  • metadata

• Two PostgreSQL users with full access to their respective databases

Supported Git Providers:

• GitHub (Cloud or Enterprise)

• GitLab

• Bitbucket Data Center

---

Models Supported

These models are used by the Context Engine and must be available for the Context Engine to be used:

• gpt-5

• claude-4.1-opus

Contact Qodo support if none of these AI models are supported by your system.

Contact Qodo support If you're using a custom LLM gateway.

---

1. Context Engine Installation

The Context Engine consists of a few components. Install them in the following order:

1. metadata-service

2. rag-indexer

3. context-retriever

4. context-retriever-mcp (only if you'd like to connect your codebase understanding capabilities to tools that are not Qodo Gen or Qodo Merge).

We recommend deploying all components in the same Kubernetes namespace.

Each component requires its own values.yaml file with configuration like the following.

1. metadata-service

Example metadata-values.yaml:

image:
  repository: artif-reg-self-hosted.codium.ai/proxy/codium-stack/us-central1-docker.pkg.dev/codium-production/codium-repo-self-hosted/qodo-engine-metadata
  tag: 1.33.1
  
deployments:
  enabled: true
  resources:
    requests:
      cpu: "500m"
      memory: "2Gi"
  ports:
    - containerPort: 8000
      name: http
  probes:
    liveness:
      httpGet:
        path: /api/v1/health
        port: http
      initialDelaySeconds: 15
      failureThreshold: 5
      periodSeconds: 30
      successThreshold: 1

service:
  enabled: true
  type: NodePort
  port: 8000
  ports:
    http:
      targetPort: http

# add externalSecrets block here if needed

volumes:
  qodo-aware-secrets:
    secret:
      secretName: qodo-aware-secrets

volumeMounts:
  qodo-aware-secrets:
    mountPath: /workspace/app/config_prod

Install with Helm:

helm upgrade --install metadata-service oci://artifacts-self-hosted.qodo.ai/codium-stack/stable/module -f ./metadata-values.yaml

2. rag-indexer

image:
  repository: artif-reg-self-hosted.codium.ai/proxy/codium-stack/us-central1-docker.pkg.dev/codium-production/codium-repo-self-hosted/qodo-engine-indexer
  tag: 1.33.1
  
deployments:
  enabled: true
  resources:
    requests:
      cpu: "500m"
      memory: "2Gi"
  ports:
    - containerPort: 3000
      name: http
  probes:
    liveness:
      httpGet:
        path: /api/v1/indexer/health
        port: http
      initialDelaySeconds: 15
      failureThreshold: 5
      periodSeconds: 30
      successThreshold: 1

service:
  enabled: true
  type: NodePort
  port: 3000
  ports:
    http:
      targetPort: http

# add externalSecrets block here if needed

volumes:
  qodo-aware-secrets:
    secret:
      secretName: qodo-aware-secrets

volumeMounts:
  qodo-aware-secrets:
    mountPath: /workspace/qodo_engine/shared_settings/secrets

cronJobs:
  reindex:
    schedule: "0 3 * * *"
    concurrencyPolicy: "Allow"
    suspend: false
    image:
      repository: curlimages/curl
      tag: 8.11.1
      pullPolicy: IfNotPresent
    command:
      - "/bin/sh"
      - "-c"
      - "curl -X POST -H 'Content-Type: application/json' -d '{}' http://indexer:3000/api/v1/indexer/reindex"

configMaps:
  main:
    REINDEX_CRONJOB_SCHEDULE: "0 3 * * *"
    ENVIRONMENT_TYPE: "ON_PREM"

Install with Helm:

helm upgrade --install indexer oci://artifacts-self-hosted.qodo.ai/codium-stack/stable/module -f ./indexer-values.yaml

3. context-retriever

image:
  repository: artif-reg-self-hosted.codium.ai/proxy/codium-stack/us-central1-docker.pkg.dev/codium-production/codium-repo-self-hosted/qodo-engine-context-retriever
  tag: 1.33.1
  
deployments:
  enabled: true
  resources:
    requests:
      cpu: "500m"
      memory: "2Gi"
  ports:
    - containerPort: 8001
      name: http
  probes:
    liveness:
      httpGet:
        path: /v1/context/health
        port: http
      initialDelaySeconds: 15
      failureThreshold: 5
      periodSeconds: 30
      successThreshold: 1

service:
  enabled: true
  type: NodePort
  port: 8001
  ports:
    http:
      targetPort: http

volumes:
  qodo-aware-secrets:
    secret:
      secretName: qodo-aware-secrets

# add externalSecrets block here if needed

volumeMounts:
  qodo-aware-secrets:
    mountPath: /app/qodo_engine/shared_settings/secrets

envFrom:
  configMapRefs:
    - name: main

configMaps:
  main:
    SERVICE_CONFIG__AUTH_BACKEND_PROVIDER: "noauth"
    ENVIRONMENT_TYPE: "ON_PREM"
    test: "TEsts"

Install with Helm:

helm upgrade --install context-retriever oci://artifacts-self-hosted.qodo.ai/codium-stack/stable/module -f ./context-retriever-values.yaml

1. Trigger initial indexing (if you prefer to not wait until the cronjob is triggered): kubectl create job indexer-reindex-manual --from=cronjob/indexer-reindex

2. Verify indexer pod (that is part of the deployment, not the job) logs, to confirm indexing has completed (can take anywhere from minutes to up to a day, depends on how many repositories are configured and how big they are).

4. context-retriever-mcp

Note: Set context-retriever-mcp only if you'd like to connect your codebase understanding capabilities to tools that are not Qodo Gen or Qodo Merge.

image:
  repository: artif-reg-self-hosted.codium.ai/proxy/codium-stack/us-central1-docker.pkg.dev/codium-production/codium-repo-self-hosted/qodo-engine-context-retriever
  tag: 1.5.0
  
deployments:
  enabled: true
  resources:
    requests:
      cpu: "500m"
      memory: "2Gi"
  ports:
    - containerPort: 8001
      name: http
  probes:
    liveness:
      httpGet:
        path: /health
        port: http
      initialDelaySeconds: 15
      failureThreshold: 5
      periodSeconds: 30
      successThreshold: 1

ingress:
  enabled: true
  host: "" # add client host here
  annotations:
    # add relevant annotations here
  hosts:
    - paths:
        - path: /mcp
          pathType: Prefix

replicaCount: 1

service:
  enabled: true
  type: NodePort
  port: 8001
  ports:
    http:
      targetPort: http

volumes:
  qodo-aware-secrets:
    secret:
      secretName: qodo-aware-secrets

# add externalSecrets block here if needed

volumeMounts:
  qodo-aware-secrets:
    mountPath: /app/qodo_engine/shared_settings/secrets

envFrom:
  configMapRefs:
    - name: main

configMaps:
  main:
    SERVICE_CONFIG__EXPOSE_MCP_SERVER: "true"
    ENVIRONMENT_TYPE: "ON_PREM"
    SERVICE_CONFIG__AUTH_BACKEND_PROVIDER: "local"

---

2. Finish Setup on your version control

Finish setting up Qodo Aware on your version control:

GitHub

GitLab

Bitbucket Data Center

2.1 Custom CA

If you are using On prem git provider with HTTPS connection from qodo-engine you will need to configure the custom CA like this:

configMaps:
  main:
    CONFIG__CERTIFICATE_BUNDLE: "/etc/ssl/custom/ca.pem"

volumeMounts:
  custom-ca:
    mountPath: /etc/ssl/custom/ca.pem
    subPath: ca.pem

volumes:
  custom-ca:
    secret:
      secretName: customer-ca-secret

---

Enabling Context Engine functionality in the Git Integration

Git Integration

1. Update qodo-merge secret, append the following

[rag_arguments]
enable_rag=true

[context_retrieval_service]
url = "<service name>.<namespace>" # set this to context-retriever service address

1. Restart pods to pick up the new secret.