[Documentation] librechat.yaml schema for custom endpoints (Ollama) is unclear

[joerg-rade]

Hi there and thanks for providing LibreChat!

I tried to compile a config file for using Phi on my laptop (with the help of Copilot).
It sent me in circles for ~1h or so, and after quite some prompting it claimed that the documentation regarding librechat.yaml could be improved. See below what it suggested.

Best regards
Jörg
---- 8< ----
📝 Description
While configuring LibreChat with a local Ollama model (Phi), I ran into repeated ZodError validation failures due to insufficient or misleading documentation. The current examples don’t clearly show the required structure for custom endpoints and models.

⚠️ Errors encountered
invalid_type on version → docs omit version, but schema requires it as a string.

unrecognized_keys on ollama/Ollama → schema doesn’t allow arbitrary keys under endpoints; must use custom.

invalid_type on models → docs show arrays or objects, but schema requires models.default as an array of strings.

Iterative errors (expected object, expected array, expected string) until the correct nesting was discovered.

✅ Working minimal config
yaml
version: "1.3.0"

endpoints:
custom:
- name: ollama
type: streamable-http
url: http://ollama:11434
models:
default:
- phi
🔎 Key gotchas
version is required at the top of the file.

Local endpoints must go under custom, not as top‑level keys.

models.default must be an array of strings (model names), not objects.

Example configs in docs should be updated to reflect this, especially for Ollama/local models.

📣 Suggested fix
Update the documentation and sample configs to show the correct schema for custom endpoints, including:

Required version field.

Proper custom nesting.

Correct models.default format.

[danny-avila]

Please see the quick start guide

https://www.librechat.ai/docs/quick_start/custom_endpoints