Configure an AI LLM

Before following this guide, we strongly recommend reading AI LLM to understand how provider configurations, repositories, AI API keys, and policies work together in Flashback. You can also review the AI LLM Management APIs if you want to automate setup.

This guide is experimental and may evolve with the platform. If something behaves differently in your workspace, please contact us on Discord.

Properties

Each AI LLM configuration uses the following properties:

• Configuration Name (required) Human-readable label of the provider connection in your workspace.

• AI LLM Type (required) Provider compatibility type used by Flashback to route and validate calls.

• API Endpoint (required) Base URL of the provider endpoint (for example https://api.openai.com/v1).

• API Key (optional) Optional key field for providers/setups that require an additional access key.

• API Secret (required) Main secret/token used to authenticate requests.

• Workspace (required, auto-selected) The current workspace where the configuration is created.

Access & security model

AI LLM credentials in Flashback follow the same security principles used for other connected resources:

• Secrets are encrypted at rest.

• Secrets are never returned in clear text by the APIs.

• Access is scoped by workspace permissions.

• Configuration names should be unique and explicit to avoid confusion during repository setup.

For the full security model (encryption/decryption behavior, scope, and best practices), see Security and Secret Encryption.

AI LLM Type

When creating a configuration, select the provider type that matches your endpoint:

• OPENAI: OpenAI-compatible endpoints.

• GOOGLE: Google AI/Gemini-compatible setup.

• ANTHROPIC: Anthropic Claude-compatible setup.

• AWS: AWS AI services (e.g., Bedrock-compatible usage through your chosen endpoint/proxy).

• OTHER: Any custom endpoint compatible with one of the supported API patterns (commonly OpenAI-compatible on-prem/private providers).

All provider types currently use the same form fields in the UI:

• Configuration Name

• API Endpoint

• API Key (optional)

• API Secret

• Workspace (auto-defined)

Instructions

Here is the step-by-step process to connect an AI LLM provider in Flashback Platform.

1

Open AI LLM management

In the left menu, go to AI (or AI → Overview), then open AI LLM. You can start creation from either:

• the + / Add action in Overview, or

• the + / Add action in AI LLM.

2

Start a new AI LLM configuration

Click Add AI LLM (or equivalent + action) to open the creation form.

3

Select the AI LLM type

Choose the provider compatibility group:

• OpenAI

• Google Cloud

• Anthropic Cloud

• AWS

• Other

Use Other for self-hosted/decentralized endpoints that are API-compatible with supported provider patterns.

4

Fill in configuration fields

Complete the form with your provider values:

• Configuration Name: clear unique name (example: prod-openai-gateway).

• API Endpoint: provider base URL.

• API Key: optional, only if your provider requires this extra key.

• API Secret: required provider secret/token.

• Workspace: already selected from your current workspace context.

Double-check endpoint protocol (https://), path (/v1 when needed), and token format before saving. Most validation failures come from endpoint typos or invalid/expired secrets.

5

Create the configuration

Click Create configuration. The AI LLM resource is now attached to your workspace and appears in the AI LLM list.

6

(Recommended) Validate and operate safely

After creation:

1. Validate connectivity (UI/API validation flow).

2. Attach the AI LLM to a Repository when needed.

3. Rotate keys regularly and update configuration when credentials change.

4. Remove unused configurations to keep workspace governance clean.

Troubleshooting tips

• Validation fails with authentication error: verify API Secret (and API Key if used), then rotate credentials.

• Provider returns endpoint/model errors: verify endpoint base URL and provider compatibility type.

• Configuration not visible where expected: check current workspace and your workspace permissions.

• Calls blocked by policy: review applied AI Policies at organization/workspace/repository scope.