fix(llm): complete path-less OpenAI-compatible base URLs

[jtguzman]

Problem

A stored local base URL of http://host:port made the OpenAI SDK POST to /chat/completions instead of /v1/chat/completions — the SDK appends the path relative to base_url.

The failure was silent in a way worth spelling out: LM Studio answers the wrong path with HTTP 200 and a body of {"error":"Unexpected endpoint or method. (POST /chat/completions)"}. Nothing raises, so the agent reports AI generated empty response instead of a routing mistake.

The credential validator could not catch it either — model listing rides LM Studio's native WebSocket API (ws://host:port/llm), which is happy without the /v1 suffix, so the URL validated green.

Fix

normalize_openai_base_url in services/llm/config.py completes the URL only when it carries no path, so an explicit gateway path such as /openai or /api/v2 is left alone. One implementation, applied at two points:

• services/llm/providers/openai.py — at the point of use, which makes an already-stored URL work without the operator re-entering it.
• nodes/model/_local_validator.py — when the URL is persisted, so new entries are stored canonically.

Verification

Live against an LM Studio server at host:1234 with no /v1 stored: the same provider path that returned an empty response now returns a completion (base_url -> http://host:1234/v1/, content 'ok', usage returned).

14 new tests in server/tests/llm/test_base_url_normalization.py cover bare host:port → /v1, an already-/v1 URL unchanged, a gateway path unchanged, empty/None, that the provider applies the completion whether the URL arrives via proxy_url or base_url, and that plain OpenAI still passes no base_url at all.

Full backend suite: 2969 passed, 2 failed — both pre-existing test_skill_icon_resolution.py failures unrelated to this change.

Out of scope, noted

An HTTP 200 carrying an {"error": ...} body should surface as a real error rather than an empty response. That is a separate change in openai.py.

🤖 Generated with Claude Code

[jtguzman]

Superado por zeenie-ai#115, que lleva el mismo arreglo rebasado sobre zeenie-ai/main en lugar del main del fork (esta rama arrastraba los 8 commits de Microsoft Graph que upstream todavía no tiene). Cierro y borro la rama; el tip queda archivado en el tag archivo/fix-lmstudio-openai-base-url.