Feat: Add configurable AI provider base URL flag and update documentation (#3059)

* feat: add FLAG_ALLOW_UNSAFE_AI_BASE_URL for configurable AI provider base URLs

* feat: add FLAG_ALLOW_UNSAFE_AI_BASE_URL documentation

* fix: remove AI_ALLOWED_BASE_URLS from documentation and environment variable reference

---------

Co-authored-by: Amruth Pillai <im.amruth@gmail.com>
This commit is contained in:
SirSKillz
2026-05-13 19:01:37 -06:00
committed by GitHub
parent 6c4a4b2aa5
commit c71f3b0b92
8 changed files with 20 additions and 28 deletions
+1 -1
View File
@@ -180,7 +180,6 @@ Here's a complete list of environment variables you can configure:
| `BETTER_AUTH_API_KEY` | Better Auth dashboard API key | — |
| `BETTER_AUTH_URL` | Better Auth base URL override (advanced) | `APP_URL` |
| `BETTER_AUTH_SECRET` | Better Auth secret override (advanced) | `AUTH_SECRET` |
| `AI_ALLOWED_BASE_URLS` | Allowlist for custom AI provider base URLs | — |
| `SMTP_HOST` | SMTP Server Host (for email features) | — |
| `SMTP_PORT` | SMTP Server Port | `587` |
| `SMTP_USER` | SMTP Username | — |
@@ -196,6 +195,7 @@ Here's a complete list of environment variables you can configure:
| `FLAG_DISABLE_SIGNUPS` | Disables new user signups | `false` |
| `FLAG_DISABLE_EMAIL_AUTH` | Disables email/password login (SSO only) | `false` |
| `FLAG_DISABLE_IMAGE_PROCESSING` | Disables image processing | `false` |
| `FLAG_ALLOW_UNSAFE_AI_BASE_URL` | Enables local-network AI providers (e.g. Ollama) | `false` |
> **Note:** Some variables are only required for using related features (OAuth, SMTP, S3, etc.) and can be left unset if unused.
+3 -13
View File
@@ -90,10 +90,6 @@ OAUTH_SCOPES=""
# BETTER_AUTH_URL="https://auth.example.com"
# BETTER_AUTH_SECRET=""
# --- AI (optional) ---
# Comma-separated hostnames/origins for custom AI base URLs
# Example: api.openai.com,https://gateway.ai.vercel.com
AI_ALLOWED_BASE_URLS=""
# --- Email (optional) ---
# If all keys are disabled, the app logs the email to be sent to the console instead.
@@ -120,6 +116,7 @@ S3_FORCE_PATH_STYLE="false"
FLAG_DISABLE_SIGNUPS="false"
FLAG_DISABLE_EMAIL_AUTH="false"
FLAG_DISABLE_IMAGE_PROCESSING="false"
FLAG_ALLOW_UNSAFE_AI_BASE_URL="false"
```
</Step>
@@ -254,9 +251,6 @@ docker compose logs -f reactive-resume
<li>
S3 storage (<code>S3_&#42;</code>)
</li>
<li>
AI URL allowlist (<code>AI_ALLOWED_BASE_URLS</code>)
</li>
<li>
Feature flags (<code>FLAG_&#42;</code>)
</li>
@@ -337,15 +331,11 @@ openssl rand -hex 32
- `"false"` for virtual-hosted-style URLs (`https://bucket.endpoint`) common with AWS S3 / Cloudflare R2.
</Accordion>
<Accordion title="AI (optional)">
- **`AI_ALLOWED_BASE_URLS`**: Comma-separated hosts or origins allowed as custom AI API base URLs. - Use this when
routing AI requests through your own gateway/proxy. - Example: `api.openai.com,https://gateway.ai.vercel.com`
</Accordion>
<Accordion title="Feature Flags">
<Accordion title="Feature Flags">
- **`FLAG_DISABLE_SIGNUPS`**: Disables new signups (web app and server). Useful for private instances.
- **`FLAG_DISABLE_EMAIL_AUTH`**: Disables email/password login entirely. Also disables email verification, forgot password, and reset password flows. Users can still sign up via social auth (Google/GitHub/LinkedIn/Custom OAuth), unless FLAG_DISABLE_SIGNUPS is also set to true. Useful when only SSO is required.
- **`FLAG_DISABLE_IMAGE_PROCESSING`**: Disables image processing. This is useful if you are using a machine with limited resources, like a Raspberry Pi.
- **`FLAG_ALLOW_UNSAFE_AI_BASE_URL`**: Allows AI providers to be configured with `http://` URLs and private/loopback addresses (e.g. a local Ollama instance at `http://192.168.1.10:11434`). **Warning: enabling this on a multi-tenant deployment is an SSRF risk.** Only enable on a trusted, single-tenant self-hosted instance.
</Accordion>
</AccordionGroup>