Skip to main content
Run Repowise codebase intelligence on Eden AI, for both the LLM and the embedder, with an EU endpoint.

​
Overview

Repowise reads a repository and produces code health scores, generated documentation and git analytics. Eden AI is a built-in provider on both sides, which is unusual: one key covers the language model and the embedder used for semantic search.
  • 500+ models: reach OpenAI, Anthropic, Google, Mistral and more through one key
  • LLM and embedder from the same key: no second account for embeddings
  • EU endpoint: EDENAI_BASE_URL covers both the LLM and the embedder in one setting

​
Prerequisites

​
Setup

​
1. Export your key

​
2. Initialise with Eden AI

Reasoning models are supported with an explicit effort level:

​
Model naming

Eden AI addresses models as vendor/model, and Repowise passes the identifier through unchanged. The live catalogues are public and need no key:
The chat and embedding catalogues are separate endpoints. An embedding model will not appear in /v3/models, which is a common source of confusion when picking one.
The same key also selects edenai as the embedder. Pick a model with REPOWISE_EMBEDDING_MODEL:
The embedder can also be chosen explicitly, and Eden AI is one of the accepted values alongside openai, gemini, ollama, openrouter and mock:
A model’s vector width is a property of the gateway serving it, not of the model name alone. The same identifier can return a different number of dimensions through two different gateways, so let Repowise read the width from Eden AI rather than copying a figure from another provider’s table.

​
EU data residency

One variable covers both the LLM and the embedder:
The EU endpoint serves the subset of the catalogue available in the EU, so a model that works on the global endpoint is not guaranteed to be reachable through it. This is the main reason to reach for Eden AI here: the other hosted providers Repowise supports terminate outside the EU.

​
Troubleshooting

​
401 Unauthorized

Check that EDENAI_API_KEY is exported in the shell Repowise runs from, and that the value is an Eden AI key rather than a vendor key.

​
model not found

Copy the identifier exactly as the catalogue returns it, vendor prefix included, and make sure you read the right catalogue: chat models from /v3/models, embedding models from /v3/embeddings/models.

​
Semantic search returns nothing useful

Confirm the embedder actually resolved to edenai rather than falling back. An unset or unrecognised embedder value silently degrades to a keyless stub, which indexes without producing real vectors.

​
Next Steps