Elastic Ramen (Elastic Console Plugin)

Experimental — this feature is under active development and may change without notice.

An OpenAI-compatible proxy that routes requests through Kibana-configured AI connectors. External tools (coding agents, CLI tools, IDE extensions) can talk to any Elasticsearch AI connector using the standard OpenAI chat completions API.

Enabling

The plugin is gated behind both a feature flag and an advanced setting:

1. Feature flag: elasticConsole.enabled must be true. In local dev, add to kibana.dev.yml:
   yamlCopy

   feature_flags.overrides:
     elasticConsole.enabled: true
   

2. Advanced setting: Go to Stack Management > Advanced Settings, search for Elastic Ramen, and toggle elasticRamen:enabled to true.

Until both are enabled, every route returns 404.

Authentication

All API routes require a valid Kibana session or API key. The easiest way to get started is the one-click setup endpoint or UI page (see below).

Setup endpoint

POST /internal/elastic_ramen/setup

Returns:

{
  "elasticsearchUrl": "https://my-cluster.es.cloud:443",
  "kibanaUrl": "https://my-cluster.kb.cloud:443",
  "apiKeyEncoded": "base64-encoded-api-key"
}

The API key is scoped to the calling user's privileges and expires after 30 days.

Setup UI

Navigate to /app/elasticRamen in Kibana. Click Generate credentials to create an API key. The page will:

1. Attempt to auto-deliver credentials to a local agent at http://localhost:14642/config
2. Fall back to displaying the credentials for manual copy if no local agent is found

Required headers

All routes are internal Kibana APIs. Clients must include:

Example: curl

curl -X POST "https://my-kibana:5601/internal/elastic_ramen/v1/chat/completions" \
  -H "Authorization: ApiKey YOUR_API_KEY" \
  -H "Content-Type: application/json" \
  -H "x-elastic-internal-origin: kibana" \
  -H "kbn-xsrf: true" \
  -d '{
    "model": "default",
    "messages": [{"role": "user", "content": "Hello!"}],
    "stream": false
  }'

API Endpoints

List models

GET /internal/elastic_ramen/v1/models

Returns available AI connectors in OpenAI model list format:

{
  "object": "list",
  "data": [
    {
      "id": "connector-id-1",
      "object": "model",
      "owned_by": ".gen-ai",
      ...
    }
  ]
}

Chat completions

POST /internal/elastic_ramen/v1/chat/completions

OpenAI-compatible chat completions endpoint. Supports:

• Streaming ("stream": true) — returns text/event-stream with SSE chunks
• Non-streaming ("stream": false) — returns a single JSON response
• Tools/function calling — pass tools array and tool_choice
• Multi-turn conversations — include full message history with assistant and tool role messages
• Image content — base64 data URIs in user message content arrays

Request body

{
  "model": "default",
  "messages": [
    { "role": "system", "content": "You are a helpful assistant." },
    { "role": "user", "content": "Hello!" }
  ],
  "stream": true,
  "temperature": 0.7,
  "tools": [],
  "tool_choice": "auto"
}

Connector resolution

The model field is used to resolve which AI connector to use:

1. If an x-connector-id header is present, that connector ID is used
2. Otherwise, the model value is tried as a connector ID
3. If neither works, the default inference connector is used
4. As a last resort, the first available connector is used

Use "model": "default" to always use the default connector.

Conversations

CRUD endpoints for managing chat conversations stored in Elasticsearch.

List conversations

GET /internal/elastic_ramen/conversations?agent_id=<optional>

Returns conversations for the current space, sorted by updated_at descending (max 100). The conversation_rounds field is excluded from list results.

Get conversation

GET /internal/elastic_ramen/conversations/:id

Returns a single conversation with full conversation_rounds.

Create conversation

POST /internal/elastic_ramen/conversations

Body:

{
  "agent_id": "my-agent",
  "title": "New conversation",
  "conversation_rounds": []
}

Returns { "id": "generated-uuid" }.

Update conversation

PUT /internal/elastic_ramen/conversations/:id

Body (all fields optional):

{
  "title": "Updated title",
  "conversation_rounds": [...]
}

Returns { "id": "conversation-id" }.

Configuration for external tools

OpenAI SDK (Python)

from openai import OpenAI

client = OpenAI(
    base_url="https://my-kibana:5601/internal/elastic_ramen/v1",
    api_key="YOUR_API_KEY",
    default_headers={
        "x-elastic-internal-origin": "kibana",
        "kbn-xsrf": "true",
    },
)

response = client.chat.completions.create(
    model="default",
    messages=[{"role": "user", "content": "Hello!"}],
)
print(response.choices[0].message.content)

OpenAI SDK (TypeScript/Node)

import OpenAI from 'openai';

const client = new OpenAI({
  baseURL: 'https://my-kibana:5601/internal/elastic_ramen/v1',
  apiKey: 'YOUR_API_KEY',
  defaultHeaders: {
    'x-elastic-internal-origin': 'kibana',
    'kbn-xsrf': 'true',
  },
});

const response = await client.chat.completions.create({
  model: 'default',
  messages: [{ role: 'user', content: 'Hello!' }],
});
console.log(response.choices[0].message.content);

Claude Code / other agents

Set the base URL to https://my-kibana:5601/internal/elastic_ramen/v1 and use the API key from the setup endpoint. Include x-elastic-internal-origin: kibana in all requests, and include kbn-xsrf: true for non-GET requests.

Development

The plugin lives at x-pack/platform/plugins/shared/elastic_console/.

Type check

yarn test:type_check --project x-pack/platform/plugins/shared/elastic_console/tsconfig.json

Lint

node scripts/eslint --fix x-pack/platform/plugins/shared/elastic_console/