Repository navigation
Conversation
caidao22
changed the title
All GPT models serve ARGO's Responses API natively, so give them both OpenAI providers and enable prefer_same_format: /v1/responses reaches ARGO unconverted instead of round-tripping through Chat Completions. Requires llm-rosetta>=0.14 (argo--openai_responses shim and server.prefer_same_format).
ARGO treats a body-level user as an override of the bearer, so a client-supplied one is replaced, not defaulted.
caidao22
force-pushed
the
feature/native-responses-routing
branch
from
October 4, 2026 01:45
d46448e to
9e70fc4
Compare
This branch has not been deployed
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
What
/v1/responsesis converted down to Chat Completions and back, which dropsthings the Chat schema has no place for. ARGO serves the Responses API
natively for GPT models, so for those this PR forwards to ARGO's own
Responses endpoint instead — as config only, no handler, no new module.
Every non-embedding OpenAI-family model gets both providers and
prefer_same_formatpicks the one matching the client's wire format.Auth fix
ARGO treats a body-level
useras an override of the bearer, so thesame-format route (which carries the body through, unlike the conversion
path that dropped the field) would let a client authenticate as someone
else.
_apply_identityoverwrites it with the resolved user.Impact: under username-passthrough this closes body-based impersonation —
the bearer now always wins. Otherwise it kills the bogus-
user-returns-200-with-warning-text failure mode. Cost: clients using OpenAI's
userforend-user attribution get it silently replaced with the ARGO username.
Blocked on llm-rosetta 0.14
Needs the
argo--openai_responsesshim,.donetool arguments, andprefer_same_format— all on master, none released (PyPI is 0.13.0, wherethis raises
KeyError: 'provider'at startup). The floor is already>=0.14, so installs fail loudly until the release lands.84 tests pass, no skips.