OpenAI-compatible API for chat integrations

BoardPulse AI exposes an OpenAI-compatible API surface at /openai/v1/. This lets you point any client that speaks the OpenAI ChatCompletion format — such as Open WebUI, LiteLLM, or your own tooling — directly at BoardPulse AI without any custom integration code. All endpoints under /openai/v1/ require a Bearer token. See Authentication for how to configure and pass the token.

Base path

http://localhost:8000/openai/v1

GET /openai/v1/models

Lists the AI models available through the OpenAI-compatible endpoint.

Authentication

Requires Authorization: Bearer <BOARDPULSE_OPENAI_COMPAT_KEY>.

Response

Returns an object in the standard OpenAI model list format.

string

Always "list".

object[]

List of available model cards.

Show OpenAIModelCard fields

string

The model identifier. BoardPulse AI exposes "boardpulse-executive".

string

Always "model".

integer

Unix timestamp of when the model entry was created.

string

Always "boardpulse".

Example

curl http://localhost:8000/openai/v1/models \
  -H "Authorization: Bearer boardpulse-dev-key"

{
  "object": "list",
  "data": [
    {
      "id": "boardpulse-executive",
      "object": "model",
      "created": 1714123456,
      "owned_by": "boardpulse"
    }
  ]
}

POST /openai/v1/chat/completions

Accepts a request in the standard OpenAI ChatCompletion format. BoardPulse AI extracts the last user message, runs it through the chat graph, and returns a formatted markdown response that includes the executive summary, a data table (if applicable), chart image (if available), export download links, and a collapsible SQL details section.

Authentication

Requires Authorization: Bearer <BOARDPULSE_OPENAI_COMPAT_KEY>.

Request body

string

default:"boardpulse-executive"

The model to use. Pass "boardpulse-executive" or any string — the value is forwarded to the underlying chat graph.

object[]

required

The conversation messages in OpenAI format.

Show OpenAIChatMessage fields

string

required

The message role. One of "system", "user", or "assistant". BoardPulse AI uses the last "user" message as the question.

string

required

The message text.

boolean

default:"false"

Streaming is accepted in the request schema but not implemented. All responses are returned as a single completion.

Response

The response follows the standard OpenAI ChatCompletion response format.

string

A unique identifier for the completion, prefixed with chatcmpl-boardpulse-.

string

Always "chat.completion".

integer

Unix timestamp of when the completion was created.

string

The model identifier echoed from the request.

object[]

The completion choices. BoardPulse AI always returns exactly one choice.

Show choice fields

integer

Always 0.

string

Always "stop".

object

string

Always "assistant".

string

The formatted markdown response, which may include the executive summary, a data table, a base64-encoded chart image, export download links, and a collapsible SQL details block.

object

Token usage counts. BoardPulse AI returns zeros for all fields since token counting is not implemented at this layer.

Show usage fields

integer

Always 0.

integer

Always 0.

integer

Always 0.

Example

curl -X POST http://localhost:8000/openai/v1/chat/completions \
  -H "Authorization: Bearer boardpulse-dev-key" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "boardpulse-executive",
    "messages": [
      {"role": "user", "content": "What were total sales by region last month?"}
    ]
  }'

{
  "id": "chatcmpl-boardpulse-1714123456",
  "object": "chat.completion",
  "created": 1714123456,
  "model": "boardpulse-executive",
  "choices": [
    {
      "index": 0,
      "finish_reason": "stop",
      "message": {
        "role": "assistant",
        "content": "Total sales last month were $1,284,500, led by the North region.\n\n### 📋 Resultados de la Consulta\n| Region | Total Sales |\n| --- | --- |\n| North | 512000 |\n| South | 398000 |\n| West | 374500 |\n\n⬇️ **Descargas:** [CSV](http://localhost:8000/api/v1/exports/report-1714123456.csv) | [XLSX](http://localhost:8000/api/v1/exports/report-1714123456.xlsx)\n\n<details><summary><b>🛠️ Detalles Técnicos (SQL)</b></summary>\n\n**SQL Generado:**\n```sql\nSELECT region, SUM(amount) AS total_sales FROM sales_orders GROUP BY region\n```\n\n**Provider:** openai\n</details>"
      }
    }
  ],
  "usage": {
    "prompt_tokens": 0,
    "completion_tokens": 0,
    "total_tokens": 0
  }
}

Connecting Open WebUI

To connect Open WebUI to BoardPulse AI:

In Open WebUI, go to Settings → Connections → OpenAI API.
Set the API base URL to http://localhost:8000/openai/v1.
Set the API key to the value of your BOARDPULSE_OPENAI_COMPAT_KEY environment variable.
Select the boardpulse-executive model and start chatting.

Open WebUI passes the authenticated user’s name and email via X-OpenWebUI-User-Name and X-OpenWebUI-User-Email headers, which BoardPulse AI includes in the technical details section of each response.

Overview

Endpoints

OpenAI-compatible API for chat integrations

Base path

GET /openai/v1/models

Authentication

Response

Example

POST /openai/v1/chat/completions

Authentication

Request body

Response

Example

Connecting Open WebUI

​Base path

​GET /openai/v1/models

​Authentication

​Response

​Example

​POST /openai/v1/chat/completions

​Authentication

​Request body

​Response

​Example

​Connecting Open WebUI

Base path

GET /openai/v1/models

Authentication

Response

Example

POST /openai/v1/chat/completions

Authentication

Request body

Response

Example

Connecting Open WebUI