feat(inference): add chat completion message listing endpoint.

[skamenan7]

What does this PR do?

Adds GET /v1/chat/completions/{completion_id}/messages, the OpenAI-compatible endpoint for listing messages from a stored chat completion.

The route reads from the inference store, flattens input and output messages into a single paginated list with synthetic IDs ({completion_id}-{index}),and supports after, limit, and order query params. It follows the same pattern as GET /conversations/{id}/items.

according to https://github.com/llamastack/llama-stack/blob/main/docs/docs/api-openai/conformance.mdx#chat, the /messages route is missing.

Test Plan

Started a local server with Ollama (remote::ollama provider, llama3.2:3b),
then ran these against it:

Create a completion:

curl -s -X POST http://localhost:8321/v1/chat/completions \
  -H 'Content-Type: application/json' \
  -d '{"model":"ollama/llama3.2:3b","messages":[
    {"role":"system","content":"Be brief."},
    {"role":"user","content":"Say hello"}
  ]}' | jq '{id, model, content: .choices[0].message.content}'

{"id": "chatcmpl-899", "model": "ollama/llama3.2:3b", "content": "Hello!"}

List messages (the new route):

curl -s http://localhost:8321/v1/chat/completions/chatcmpl-899/messages | jq .

{
  "object": "list",
  "data": [
    {"id": "chatcmpl-899-0", "role": "system", "content": "Be brief."},
    {"id": "chatcmpl-899-1", "role": "user", "content": "Say hello"},
    {"id": "chatcmpl-899-2", "role": "assistant", "content": "Hello!"}
  ],
  "first_id": "chatcmpl-899-0",
  "last_id": "chatcmpl-899-2",
  "has_more": false
}

Pagination (limit=1, then cursor):

curl -s 'http://localhost:8321/v1/chat/completions/chatcmpl-899/messages?limit=1' | jq .

{"data": [{"id": "chatcmpl-899-0", "role": "system"}], "has_more": true}

curl -s 'http://localhost:8321/v1/chat/completions/chatcmpl-899/messages?after=chatcmpl-899-0' | jq .

{"data": [{"id": "chatcmpl-899-1", "role": "user"}, {"id": "chatcmpl-899-2", "role": "assistant"}], "has_more": false}

Invalid cursor returns 400:

curl -s 'http://localhost:8321/v1/chat/completions/chatcmpl-899/messages?after=bogus' | jq .

{"error": {"message": "Failed to list chat completion messages: cursor 'bogus' not found in completion 'chatcmpl-899'."}}

Multi-turn (4 input + 1 output = 5 messages):

curl -s http://localhost:8321/v1/chat/completions/chatcmpl-623/messages | jq '[.data[] | {id, role, content}]'

[
  {"id": "chatcmpl-623-0", "role": "system", "content": "You are a math tutor. Be brief."},
  {"id": "chatcmpl-623-1", "role": "user", "content": "What is 2+2?"},
  {"id": "chatcmpl-623-2", "role": "assistant", "content": "4"},
  {"id": "chatcmpl-623-3", "role": "user", "content": "What is 4+4?"},
  {"id": "chatcmpl-623-4", "role": "assistant", "content": "8"}
]

Unit tests

Unit tests in tests/unit/utils/inference/test_inference_store.py covering:

• Basic input/output message listing
• Pagination with limit and cursor
• Descending order
• Not-found and invalid-cursor error handling
• Multi-choice (n>1) global ID assignment
• Tool call preservation
• After-last-message empty page

uv run pytest tests/unit/utils/inference/test_inference_store.py -v -k "list_chat_completion_messages"

Integration test (against running server):

 uv run pytest tests/integration/inference/ -v -k "messages" \
    --text-model ollama/llama3.2:3b --stack-config=http://localhost:8321

Passes. Tests pagination with limit=1, after cursor, role and ID assertions.

[github-actions]

✱ Stainless preview builds

This PR will update the llama-stack-client SDKs with the following commit message.

feat(inference): add chat completion message listing endpoint.

Edit this comment to update it. It will appear in the SDK's changelogs.

✅ llama-stack-client-openapi studio · code · diff

Your SDK build had at least one "warning" diagnostic, but this did not represent a regression.
generate ⚠️

✅ llama-stack-client-python studio · code · diff

Your SDK build had at least one "warning" diagnostic, but this did not represent a regression.
generate ⚠️ → build ✅ → lint ❗ → test ✅

pip install https://pkg.stainless.com/s/llama-stack-client-python/00c5fd261ef2089ddee36d9c17922643c5d4d854/ogx_client-0.7.0a2-py3-none-any.whl

✅ llama-stack-client-go studio · conflict

Your SDK build had at least one new note diagnostic, which is a regression from the base state.

New diagnostics (1 note)

💡 Schema/EnumHasOneMember: This enum schema has just one member, so it could be defined using [`const`](https://json-schema.org/understanding-json-schema/reference/const).

✅ llama-stack-client-node studio · conflict

Your SDK build resulted in a merge conflict between your custom code and the newly generated changes, but this did not represent a regression.

---

This comment is auto-generated by GitHub Actions and is automatically kept up to date as you push.
If you push custom code to the preview branch, re-run this workflow to update the comment.
Last updated: 2026-04-28 16:23:25 UTC

[mergify]

This pull request has merge conflicts that must be resolved before it can be merged. @skamenan7 please rebase it. https://docs.github.com/en/pull-requests/collaborating-with-pull-requests/working-with-forks/syncing-a-fork

[mergify]

This pull request has merge conflicts that must be resolved before it can be merged. @skamenan7 please rebase it. https://docs.github.com/en/pull-requests/collaborating-with-pull-requests/working-with-forks/syncing-a-fork

[github-actions]

Recording workflow completed

Providers: azure

Recordings have been generated and will be committed automatically by the companion workflow.

View workflow run

Fork PR: Recordings will be committed if you have "Allow edits from maintainers" enabled.

[github-actions]

✅ Recordings committed successfully

Recordings from the integration tests have been committed to this PR.

View commit workflow

[mergify]

This pull request has merge conflicts that must be resolved before it can be merged. @skamenan7 please rebase it. https://docs.github.com/en/pull-requests/collaborating-with-pull-requests/working-with-forks/syncing-a-fork

[cdoern]

small review, initially looking at this I see new conformance issues we should fix

[cdoern]

same here

[skamenan7]

I kept this PR scoped to the actual schema fix, so I only changed the API side here.
There’s still one remaining Chat conformance item on GET /chat/completions/{completion_id}/messages. That looks like a schema-shape mismatch in the OpenAI conformance check, not an actual endpoint behavior issue. I didn’t make that conformance-side change here, but I can update it separately if you want.

[cdoern]

this is a net-new issue introduced by the new route. can we fix this? if not possible let me know.

[skamenan7]

yeah, fixed this in 2e6284b73.

The remaining item was coming from the OpenAI coverage diff on ChatCompletionMessageList.data.items rather than the route behavior itself. OpenAI wraps that item schema differently (allOf vs our bare $ref), so I normalized that in scripts/openai_coverage.py for the conformance check.

The /chat/completions/{completion_id}/messages entry is back to 0 issues now.

PTAL @cdoern the above change.