> ## Documentation Index
> Fetch the complete documentation index at: https://docs.presenton.ai/llms.txt
> Use this file to discover all available pages before exploring further.

# Environment variables

> Reference supported self-hosted settings by deployment concern.

Use environment variables for repeatable deployments and centrally managed secrets. The in-app Settings page is convenient for personal instances; environment variables are better for servers, automation, and locked-down provider credentials.

## Runtime and access

| Variable                     | Default      | Purpose                                                                    | Sensitive |
| ---------------------------- | ------------ | -------------------------------------------------------------------------- | --------- |
| `PRESENTON_HTTP_HOST_PORT`   | `5001`       | Compose host port mapped to container port `80`                            | No        |
| `CAN_CHANGE_KEYS`            | unset        | Set `false` to prevent users changing administrator-supplied provider keys | No        |
| `AUTH_USERNAME`              | unset        | Preseed the single administrator account                                   | No        |
| `AUTH_PASSWORD`              | unset        | Password used only to initialize or explicitly rotate credentials          | Yes       |
| `AUTH_OVERRIDE_FROM_ENV`     | `false`      | Apply a one-time credential rotation from environment values               | No        |
| `RESET_AUTH`                 | `false`      | Remove configured authentication during a recovery boot                    | No        |
| `DATABASE_URL`               | local SQLite | External PostgreSQL or MySQL connection URL                                | Yes       |
| `DISABLE_ANONYMOUS_TRACKING` | `false`      | Disable anonymous product telemetry                                        | No        |

## Text providers

Set `LLM` to the provider identifier, then configure that provider's model and credentials.

| Provider          | Required or common variables                                              |
| ----------------- | ------------------------------------------------------------------------- |
| OpenAI            | `LLM=openai`, `OPENAI_API_KEY`, `OPENAI_MODEL`                            |
| Google            | `LLM=google`, `GOOGLE_API_KEY`, `GOOGLE_MODEL`                            |
| Anthropic         | `LLM=anthropic`, `ANTHROPIC_API_KEY`, `ANTHROPIC_MODEL`                   |
| Ollama            | `LLM=ollama`, `OLLAMA_URL`, `OLLAMA_MODEL`                                |
| LM Studio         | `LLM=lmstudio`, `LMSTUDIO_BASE_URL`, `LMSTUDIO_MODEL`                     |
| OpenAI-compatible | `LLM=custom`, `CUSTOM_LLM_URL`, `CUSTOM_LLM_API_KEY`, `CUSTOM_MODEL`      |
| Azure OpenAI      | `LLM=azure` plus endpoint, API version, model/deployment, and credentials |
| Amazon Bedrock    | `LLM=bedrock`, region, model, and one supported AWS authentication method |
| LiteLLM           | `LLM=litellm`, `LITELLM_BASE_URL`, `LITELLM_MODEL`, optional API key      |

<Card title="Provider details" icon="microchip" href="/archive/v0.9.1-beta/self-hosted/core-concepts/providers">
  Review supported roles, local-versus-hosted data flow, and provider-specific guides.
</Card>

## Images and search

| Variable                                            | Purpose                                               |
| --------------------------------------------------- | ----------------------------------------------------- |
| `IMAGE_PROVIDER`                                    | Select stock or generated-image provider              |
| `PEXELS_API_KEY`, `PIXABAY_API_KEY`                 | Stock-image credentials                               |
| `DISABLE_IMAGE_GENERATION`                          | Disable presentation image generation                 |
| `ENABLE_PARALLEL_IMAGE_GENERATION`                  | Control concurrent image work                         |
| `WEB_GROUNDING`                                     | Enable web context by default                         |
| `WEB_SEARCH_PROVIDER`                               | Select `auto`, native search, SearXNG, Tavily, or Exa |
| `WEB_SEARCH_MAX_RESULTS`                            | Limit external results added to model context         |
| `SEARXNG_BASE_URL`, `TAVILY_API_KEY`, `EXA_API_KEY` | Search-provider connection details                    |

## Processing and memory

| Variable                | Default                           | Purpose                                         |
| ----------------------- | --------------------------------- | ----------------------------------------------- |
| `LITEPARSE_DPI`         | `120`                             | Rendering resolution during document extraction |
| `LITEPARSE_NUM_WORKERS` | `1`                               | Document-processing workers                     |
| `MEM0_ENABLED`          | `true`                            | Enable presentation-scoped memory               |
| `MEM0_DIR`              | `/app_data/mem0`                  | Memory storage root                             |
| `MEM0_LLM_MODEL`        | Ollama model or `llama3.1:latest` | Memory extraction model                         |
| `MEM0_LLM_BASE_URL`     | Ollama URL or host gateway        | Compatible model endpoint                       |
| `MEM0_EMBEDDER_MODEL`   | `BAAI/bge-small-en-v1.5`          | Local embedding model                           |

## Apply changes safely

* Keep secrets out of source control and shell history.
* Change one provider role at a time and run a short generation afterward.
* Restart or replace the container after environment changes.
* Back up `/app_data` and an external database before storage or migration changes.
* Treat empty values differently from deliberate `false` values.