feat: Modify endpoints for OpenAPI compatibility#1340

Merged

Pouyanpi merged 32 commits into

NVIDIA-NeMo:developfrom

christinaexyou:openai-server-compat

Feb 5, 2026

Commits on Feb 5, 2026

feat: Modify endpoints for OpenAPI compatibility

christinaexyou
authored and
Pouyanpi
committed
fix: Colang 2.x doesn't support assistant messages

christinaexyou
authored and
Pouyanpi
committed
chore: Move OpenAPI schema and fix typos

christinaexyou
authored and
Pouyanpi
committed
Extend existing OpenAI types and add support for streaming chat completion

christinaexyou
authored and
Pouyanpi
committed
Add OpenAI docs and integration tests

christinaexyou
authored and
Pouyanpi
committed
Add model name to response body

christinaexyou
authored and
Pouyanpi
committed
Update poetry.lock

christinaexyou
authored and
Pouyanpi
committed
Rename OpenAI API schema definitions

christinaexyou
authored and
Pouyanpi
committed
test(server): move server tests to dedicated subfolder
Pouyanpi
committed
chore(deps): add extra server dependency and make aiofiles and openai optional
Pouyanpi
committed
fix(utils): handle non-dict JSON in streaming chunk
Pouyanpi
committed
remove non-server related changes
Pouyanpi
committed
refactor(api): move request schema to schemas module
Pouyanpi
committed
refactor(api): move config fields under guardrails key
Pouyanpi
committed
refactor(schema): drop implicit model to config_id mapping
Pouyanpi
committed
fix(api): improve error handling for config and state
Pouyanpi
committed
test: update test assertion to capture expected behavior
Pouyanpi
committed
refactor(api): require explicit model in chat completions
Pouyanpi
committed
feat(api): support model override in chat completions
Pouyanpi
committed
update openai.md with the recent changes
Pouyanpi
committed
update tests given refactoring
Pouyanpi
committed
fix(api): raise HTTPException for missing MAIN_MODEL_ENGINE
Pouyanpi
committed
Upgrade to `openai>=2.0.0`

christinaexyou
authored and
Pouyanpi
committed
refactor(schema): nest guardrails fields under `guardrails` for response
Pouyanpi
committed
fix(server): chunk id must be identical
Pouyanpi
committed
refactor(api): remove models endpoint
Pouyanpi
committed
openai 1 is also valid
Pouyanpi
committed
refactor(schema): migrate to Pydantic v2 validators
Pouyanpi
committed
docs(openai): remove models endpoint

christinaexyou
authored and
Pouyanpi
committed
nit(api): default MAIN_MODEL_ENGINE=openai

christinaexyou
authored and
Pouyanpi
committed
feat(api): support MAIN_MODEL_BASE_URL for main model
Pouyanpi
committed
remove doc
Pouyanpi
committed

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

feat: Modify endpoints for OpenAPI compatibility#1340

feat: Modify endpoints for OpenAPI compatibility#1340
Pouyanpi merged 32 commits into
NVIDIA-NeMo:developfrom
christinaexyou:openai-server-compat

Commits on Feb 5, 2026

feat: Modify endpoints for OpenAPI compatibility

fix: Colang 2.x doesn't support assistant messages

chore: Move OpenAPI schema and fix typos

Extend existing OpenAI types and add support for streaming chat completion

Add OpenAI docs and integration tests

Add model name to response body

Update poetry.lock

Rename OpenAI API schema definitions

test(server): move server tests to dedicated subfolder

chore(deps): add extra server dependency and make aiofiles and openai optional

fix(utils): handle non-dict JSON in streaming chunk

remove non-server related changes

refactor(api): move request schema to schemas module

refactor(api): move config fields under guardrails key

refactor(schema): drop implicit model to config_id mapping

fix(api): improve error handling for config and state

test: update test assertion to capture expected behavior

refactor(api): require explicit model in chat completions

feat(api): support model override in chat completions

update openai.md with the recent changes

update tests given refactoring

fix(api): raise HTTPException for missing MAIN_MODEL_ENGINE

Upgrade to `openai>=2.0.0`

refactor(schema): nest guardrails fields under `guardrails` for response

fix(server): chunk id must be identical

refactor(api): remove models endpoint

openai 1 is also valid

refactor(schema): migrate to Pydantic v2 validators

docs(openai): remove models endpoint

nit(api): default MAIN_MODEL_ENGINE=openai

feat(api): support MAIN_MODEL_BASE_URL for main model

remove doc