Qevron AI Gateway Setup

Qevron is Sosyabot's default LLM gateway — every text-generation call (prompts, inline assist, AI agents) routes through it, then on to whichever upstream model the admin enables in AI Models. The admin form at /admin/api-integration/qevron is where you point Sosyabot at your Qevron deployment.

Overview

Form field	Options key	Required
Base URL	qevron_api-base-url	Yes
API Key	qevron_api-key	Yes
Status	qevron_integration_status	—

ENV fallback: QEVRON_API_BASE_URL, QEVRON_API_KEY.

The full chat-completions URL Sosyabot calls is <Base URL>/v1/chat/completions — only the origin goes in Base URL; Sosyabot appends the path.

Step-by-step

1. Provision a Qevron deployment

Qevron is a self-hosted (or vendor-managed) OpenAI-compatible gateway. If you already run one, skip to step 2. Otherwise contact your Qevron operator for deployment URL and an API token.

2. Generate an API token

In your Qevron admin UI, issue a Bearer token scoped to chat completions. Copy the token — Qevron typically shows it once.

3. Confirm the public origin

The Base URL Sosyabot needs is the bare origin (no path, no trailing slash). Examples:

• http://172.19.16.57:3001 (internal)
• https://qevron.example.com (public)

Pasting into Sosyabot admin panel

/admin/api-integration/qevron:

1. Base URL: the public origin.
2. API Key: the Bearer token.
3. Status: Enable.
4. Save → Test connection sends a minimal probe to <Base URL>/v1/chat/completions and reports HTTP status.

Failure: Qevron not configured: set credentials from the admin panel (Integrations → Qevron) or via QEVRON_API_KEY env.

Per-model API key override

Each AI model registered in AI Models can carry its own API key (PUT /api/v1/ai-models/:id/api-key). When set, it overrides the global Qevron key for that specific model — useful when one model lives behind a different gateway or quota plan.

Verification

Test from /admin/data/ai-models → pick a text model → Test button. The test runs a one-shot prompt through Qevron and surfaces the response (or upstream error) inline.

You can also try inline AI from /app/ai-contents: select any model, click Generate, and confirm a response.

Troubleshooting

Symptom	Cause	Fix
Qevron not configured: ...	Admin + ENV missing	Set both Base URL and API Key.
Test connection returns 401	Wrong API token	Re-issue in Qevron, paste again.
Test connection returns 404	Base URL has a path or trailing slash	Use the bare origin. Sosyabot appends /v1/chat/completions.
Specific model fails but others work	Per-model API key invalid	Check the model's per-model key under AI Models; clear it to fall back to the global Qevron key.

ENV fallback

QEVRON_API_BASE_URL=https://qevron.example.com
QEVRON_API_KEY=<your bearer token>

After editing .env run ./service.sh restart api.

Related

OAuth Credentials overview

AI Models registry

AI Overview