Files
ragflow/conf
0xτensor 127aeac4aa fix: expose gpt-5.5 and gpt-5.4 in OpenAI model list (#14828)
### What problem does this PR solve?

OpenAI model catalogs used in provider selection flows were missing the
latest GPT models (`gpt-5.5` and `gpt-5.4`).
Because model availability is driven by seeded catalog data
(`conf/llm_factories.json` → DB seed → API response), these models were
not selectable in the UI or `/llm/list` responses.

This PR updates and synchronizes the OpenAI catalog definitions across
configuration sources and ensures the new models are correctly exposed
through the API layer and validated in tests.

---

### Type of change

* [x] New Feature (non-breaking change which adds functionality)

---

### Changes Made

* Added `gpt-5.5` and `gpt-5.4` to OpenAI catalog definitions in:

  * `conf/llm_factories.json`
  * `conf/models/openai.json` (chat + vision support)
* Ensured consistency between DB-seeded factory config and provider
model configuration
* Updated test coverage in:

  * `test_llm_list_unit.py`

    * seeded OpenAI catalog entries
* added response-level assertion validating `/llm/list` includes both
new model IDs under OpenAI grouping

---

### Root Cause

OpenAI model listings in selection flows are generated from catalog data
seeded via `conf/llm_factories.json`.
The catalog had not been updated to include the latest GPT models,
resulting in missing availability in UI and API responses.

---

### Testing

* Created isolated test environment:

  * `python -m venv .venv-review`
  * installed `pytest`
* Ran targeted and full test suite:

  * `test_list_app_grouping_availability_and_merge`:  passed
  * Full `test_llm_list_unit.py`:  10 passed

---

### Risks / Limitations

* Adding models to the catalog does not guarantee upstream provider
availability or account entitlement.
* Environments with pre-seeded DB catalogs may require reseed or refresh
to reflect updated configuration.

---

### Notes

* Changes are minimal and scoped strictly to catalog configuration and
related test coverage.
* Ensures `/llm/list` API remains aligned with expected latest OpenAI
model availability.
* Closes #14827
2026-05-12 18:03:47 +08:00
..
2024-01-15 08:46:22 +08:00
2026-03-24 20:08:36 +08:00