feat(config): add OpenAI-compatible custom provider support

[gcaguilar]

Summary

• Add CUSTOM_API_KEY, CUSTOM_BASE_URL, and CUSTOM_MODEL env vars for any OpenAI-compatible endpoint (LiteLLM, LocalAI, LM Studio, vLLM, etc.)
• Update hasProvider validation in both configure.js and entrypoint.sh so the container starts with only CUSTOM_API_KEY set
• Add 34 tests in scripts/configure.test.js covering provider config, base URL handling, model defaults, priority selection, and hasProvider
• Fix Custom Provider Api Key #24

Details

The existing provider system supports built-in providers (Anthropic, OpenAI, Gemini, etc.) and custom/proxy providers that require a full models.providers config block. This PR adds a dedicated env-var path for OpenAI-compatible providers, following the same pattern as Venice AI, Moonshot, and Ollama.

Files changed (5 files, +442/-5)

File	Change
scripts/configure.js	Provider config block, primaryCandidates entry, hasProvider validation, error message
scripts/entrypoint.sh	Provider check loop + error message
.env.example	3 new env vars with comments
README.md	Provider table rows + updated priority list
scripts/configure.test.js	New test file — 34 tests

Provider config

{
  "models": {
    "providers": {
      "openai-compatible": {
        "api": "openai-completions",
        "apiKey": "<CUSTOM_API_KEY>",
        "baseUrl": "<CUSTOM_BASE_URL or https://api.openai.com/v1>",
        "models": [{ "id": "<CUSTOM_MODEL or gpt-4>", "name": "...", "contextWindow": 128000 }]
      }
    }
  }
}

Env vars

Variable	Default	Description
CUSTOM_API_KEY	(required)	API key for the OpenAI-compatible endpoint
CUSTOM_BASE_URL	https://api.openai.com/v1	Base URL (trailing slashes stripped)
CUSTOM_MODEL	gpt-4	Model ID and display name

[samiralavi]

I need this PR to be merged soon

[gcaguilar]

@andrasbacsai Could you take a look at the PR, please?

[andrasbacsai]

@coderabbitai review

[coderabbitai]

✅ Actions performed

Review triggered.

Note: CodeRabbit is an incremental review system and does not re-review already reviewed commits. This command is applicable only when automatic reviews are paused.

[coderabbitai]

Walkthrough

This PR adds support for OpenAI-compatible custom providers, enabling users to configure self-hosted or alternative endpoints via three environment variables: CUSTOM_API_KEY (required), CUSTOM_BASE_URL (optional, defaults to /v1), and CUSTOM_MODEL (optional, defaults to gpt-4). The configuration logic in scripts/configure.js creates a models.providers["openai-compatible"] entry when the API key is present, extends model auto-selection to include the custom provider as a fallback, and removes the provider configuration when the key is absent. Validation logic treats CUSTOM_API_KEY as a satisfying provider credential. Documentation, configuration examples, and comprehensive tests validate the feature across provider creation, removal, coexistence, priority ordering, and validation scenarios.

---

Come with me if you want to live... behind your own self-hosted servers instead of paying VC-backed serverless tax. This PR is good. It is very good. Now where is my gluten-free taco? The test suite is 409 lines of beautiful, hardened logic—no framework bloat, no magic. The Terminator approves.

✨ Finishing Touches

🧪 Generate unit tests (beta)

• Create PR with unit tests

---

Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out.

❤️ Share

• X
• Mastodon
• Reddit
• LinkedIn

_{Comment @coderabbitai help to get the list of available commands and usage tips.}

[coderabbitai]

Actionable comments posted: 2

🤖 Prompt for all review comments with AI agents

Verify each finding against current code. Fix only still-valid issues, skip the
rest with a brief reason, keep changes minimal, and validate.

Inline comments:
In `@scripts/configure.js`:
- Line 373: The primaryCandidates array is hardcoded to
"openai-compatible/gpt-4" and doesn't honor the CUSTOM_MODEL used when
constructing the provider's models (the block that creates a model with
CUSTOM_MODEL around where models are defined and referenced by
primaryCandidates); update the primaryCandidates entry to use the same
CUSTOM_MODEL variable (e.g., process.env.CUSTOM_MODEL or the local constant used
to build the provider model) instead of the literal "openai-compatible/gpt-4" so
the auto-selection matches the provider's actual model ID.

In `@scripts/configure.test.js`:
- Around line 304-361: The test suite misses a case verifying CUSTOM_MODEL is
used when building primaryCandidates: update the tests in configure.test.js to
add a test (similar to the proposed snippet) that sets
process.env.CUSTOM_API_KEY and process.env.CUSTOM_MODEL and asserts the selected
model equals `openai-compatible/${process.env.CUSTOM_MODEL}`; ensure the
primaryCandidates entry for the custom provider uses `process.env.CUSTOM_MODEL
|| 'gpt-4'` so it mirrors the behavior expected by configure.js (see
configure.js line referenced around the primaryCandidates construction).

🪄 Autofix (Beta)

Fix all unresolved CodeRabbit comments on this PR:

• Push a commit to this branch (recommended)
• Create a new PR with the fixes

---

ℹ️ Review info

⚙️ Run configuration

Configuration used: Organization UI

Review profile: ASSERTIVE

Plan: Pro

Run ID: 0a69d2e6-a1a1-4ff5-a166-896ae1690ff9

📥 Commits

Reviewing files that changed from the base of the PR and between 577e3e8 and 76dbd5d.

📒 Files selected for processing (5)

• .env.example
• README.md
• scripts/configure.js
• scripts/configure.test.js
• scripts/entrypoint.sh

[gcaguilar]

@andrasbacsai Comments applied