Configure by goal

Use this page when you know what you want Presenton to do but do not yet know which settings make it happen.

Quick decision table

Goal	Required ports	Main settings or actions
Open the browser application	`5001:80` by default	Mount `/app_data`; select a text provider
Use a ChatGPT/Codex subscription	`5001:80` and `1455:1455`	`LLM=codex`, `CODEX_MODEL`; complete Sign in with ChatGPT
Use an API-key provider	`5001:80`	`LLM`, provider API key, and provider model
Use the REST API	`5001:80`	`AUTH_USERNAME`, `AUTH_PASSWORD`; HTTP Basic auth
Use MCP	`5001:80`	Admin credentials, login bearer token, client URL ending in `/mcp`
Use Ollama on the Docker host	`5001:80`; Ollama must be reachable from the container	`LLM=ollama`, `OLLAMA_URL`, `OLLAMA_MODEL`
Enable stock or generated images	`5001:80`	`IMAGE_PROVIDER` plus its credentials or endpoint
Generate without images	`5001:80`	`DISABLE_IMAGE_GENERATION=true`
Enable web research	`5001:80`	`WEB_GROUNDING=true`, `WEB_SEARCH_PROVIDER`, optional provider key
Lock provider settings	`5001:80`	`CAN_CHANGE_KEYS=false`
Persist presentations and settings	No extra port	Mount durable storage at `/app_data`
Use an external database	Database must be reachable from the container	`DATABASE_URL`, `MIGRATE_DATABASE_ON_STARTUP`
Disable anonymous telemetry	No extra port	`DISABLE_ANONYMOUS_TRACKING=true`

Use an existing ChatGPT or Codex subscription

Choose the Codex provider when you want to sign in with a free or paid ChatGPT account instead of supplying an OpenAI API key. The OAuth redirect is fixed to:

http://localhost:1455/auth/callback

Therefore Docker must publish both ports:

5001:80 for the Presenton browser application.
1455:1455 for the ChatGPT/Codex OAuth callback.

docker run -it --name presenton \
  --publish 5001:80 \
  --publish 1455:1455 \
  --env LLM=codex \
  --env CODEX_MODEL=gpt-5.5 \
  --volume "./app_data:/app_data" \
  ghcr.io/presenton/presenton:v0.9.0-beta

Then open Presenton, go to the text-provider settings, select Sign in with ChatGPT, complete authentication in the browser, and choose a supported Codex model.

If 1455:1455 is missing, the OpenAI sign-in page may open, but the redirect cannot reach the callback server inside the container. This port is not needed for normal API-key providers.

The current v0.9.0-beta application accepts gpt-5.5, gpt-5.4, gpt-5.4-mini, and gpt-5.3-codex-spark as Codex model identifiers. Availability can still depend on the signed-in account.

Docker Compose

The repository’s production, production-gpu, development, and development-gpu services already map 1455:1455.

LLM=codex
CODEX_MODEL=gpt-5.5

docker compose up production

Remote Docker host

Because the browser redirect targets localhost:1455, localhost means the machine running the browser. When Presenton runs on another host, use an approved network arrangement that makes both Presenton and the callback available locally. For an SSH-accessible server, one option is local port forwarding:

ssh -L 5001:localhost:5001 -L 1455:localhost:1455 user@presenton-server

While the tunnel is active, open http://localhost:5001 and complete the ChatGPT sign-in from that browser session.

Port 1455 is for Codex OAuth, not MCP. MCP uses the normal Presenton web port and /mcp path.

Use OpenAI with an API key

This path uses API billing and does not require port 1455.

docker run -it --name presenton \
  --publish 5001:80 \
  --env LLM=openai \
  --env OPENAI_API_KEY=YOUR_OPENAI_API_KEY \
  --env OPENAI_MODEL=gpt-4.1 \
  --env IMAGE_PROVIDER=dall-e-3 \
  --env DALL_E_3_QUALITY=standard \
  --env CAN_CHANGE_KEYS=false \
  --volume "./app_data:/app_data" \
  ghcr.io/presenton/presenton:v0.9.0-beta

CAN_CHANGE_KEYS=false hides and locks provider credentials in the UI. Leave it true or unset for a personal instance where settings should remain editable.

Use Google for text and images

One Google API key can supply the Gemini text and image modes:

LLM=google
GOOGLE_API_KEY=replace-with-your-key
GOOGLE_MODEL=models/gemini-2.0-flash
IMAGE_PROVIDER=gemini_flash
CAN_CHANGE_KEYS=false

Use LLM=vertex and the Vertex-specific authentication settings when you need Google Cloud project controls instead of the direct Gemini API.

Run text generation locally with Ollama

Run Ollama on the Docker host, pull the model there, and point Presenton at the host gateway:

ollama pull llama3.2:3b

LLM=ollama
OLLAMA_URL=http://host.docker.internal:11434
OLLAMA_MODEL=llama3.2:3b
START_OLLAMA=false

START_OLLAMA=false means Presenton does not start its own Ollama service. If host.docker.internal is unavailable on your Linux Docker setup, provide an equivalent reachable host gateway or run both services on the same Docker network. To avoid sending image prompts to an external provider:

DISABLE_IMAGE_GENERATION=true

For a more complete local visual workflow, configure ComfyUI instead of disabling images.

Use local text with stock images

Text and images do not need to use the same provider:

LLM=ollama
OLLAMA_URL=http://host.docker.internal:11434
OLLAMA_MODEL=llama3.2:3b

IMAGE_PROVIDER=pexels
PEXELS_API_KEY=replace-with-your-key

This keeps text generation local while sending image searches to Pexels.

Enable current-information web research

For supported OpenAI, Google, and Anthropic configurations, auto uses the provider’s native search behavior:

WEB_GROUNDING=true
WEB_SEARCH_PROVIDER=auto
WEB_SEARCH_MAX_RESULTS=5

For other text providers, choose an external search provider explicitly:

WEB_GROUNDING=true
WEB_SEARCH_PROVIDER=searxng
SEARXNG_BASE_URL=http://searxng:8080
WEB_SEARCH_MAX_RESULTS=5

Set WEB_GROUNDING=false when current public information is unnecessary or external research is not allowed.

Enable the REST API and MCP

Both interfaces use the normal web port. Configure the admin account first:

AUTH_USERNAME=admin
AUTH_PASSWORD=replace-with-a-long-password

REST API: http://localhost:5001/api/v1, authenticated with HTTP Basic.
MCP: http://localhost:5001/mcp, authenticated with a bearer token returned by /api/v1/auth/login.

No additional MCP port is required. The Electron desktop app does not expose MCP.

Generate with the API

Upload documents and generate PPTX or PDF output over REST.

Connect MCP

Request a bearer token and configure local or remote clients.

Improve document extraction

LiteParse defaults are suitable for most documents:

LITEPARSE_DPI=120
LITEPARSE_NUM_WORKERS=1

Increase LITEPARSE_DPI when scanned or image-heavy documents need clearer OCR. Increase workers only when the host has enough CPU and memory for concurrent parsing.

Use an external database

Without DATABASE_URL, Presenton uses SQLite under /app_data. For an external database:

DATABASE_URL=postgresql://presenton:password@postgres:5432/presenton
MIGRATE_DATABASE_ON_STARTUP=true

The database hostname must be reachable from the container. Continue mounting /app_data; presentations, uploads, templates, exports, authentication state, and memory are not all replaced by the external database.

Apply configuration changes

Environment changes require a container restart or recreation:

docker compose up --detach --force-recreate production
docker compose logs --tail 100 production

After every provider or networking change, generate a short test presentation and verify the browser, API, export, and MCP paths that your deployment intends to support.

Get started

Learn the basics

Features

Tutorials

API guides

Deploy and configure

Develop and contribute

This release

Quick decision table

Use an existing ChatGPT or Codex subscription

Docker Compose

Remote Docker host

Use OpenAI with an API key

Use Google for text and images

Run text generation locally with Ollama

Use local text with stock images

Enable current-information web research

Enable the REST API and MCP

Generate with the API

Connect MCP

Improve document extraction

Use an external database

Apply configuration changes

​Quick decision table

​Use an existing ChatGPT or Codex subscription

​Docker Compose

​Remote Docker host

​Use OpenAI with an API key

​Use Google for text and images

​Run text generation locally with Ollama

​Use local text with stock images

​Enable current-information web research

​Enable the REST API and MCP

Generate with the API

Connect MCP

​Improve document extraction

​Use an external database

​Apply configuration changes

Quick decision table

Use an existing ChatGPT or Codex subscription

Docker Compose

Remote Docker host

Use OpenAI with an API key

Use Google for text and images

Run text generation locally with Ollama

Use local text with stock images

Enable current-information web research

Enable the REST API and MCP

Improve document extraction

Use an external database

Apply configuration changes