You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
Successor to #596 at 7a07562. Keeps that head's fail-closed empty messages, spend/routing hoist, and request-knob honesty, and closes the buyer-visible gap those drafts still list as residual: tools / response_format + stream=true used to return 400 invalid_stream (or, before that, a billed JSON completion while the SDK waited for SSE).
TaskOrchestrator.proxy_completion_stream / ModelClient.proxy_stream_send SSE-proxy a single pool agent.
Mock frames are chat.completion.chunk whose concatenated content equals the non-stream JSON message.content.
Live providers are piped verbatim so delta.tool_calls survive (content-only parsers drop them).
Empty messages, seed / stop / n>1 / logprobs, batch routing, and stream_options.include_usage=true still fail closed before the first byte.
/v1/responses streaming stays rejected.
Buyer next action: send stream=true on tool-calling requests when the client reads SSE; omit stream_options.include_usage; still send a non-empty messages array of objects and known sync attribution.
Prefer this head over #582–#587, #589, #591–#597. Do not merge those in parallel. Independent non-author APPROVE + Full unit/Semgrep still required.
…closed otherwise
Chat history: message-level audio and legacy function_call are null/empty omit
no-ops; non-empty fail closed with named errors (including tools passthrough).
Tip substrate from #577 assistant refusal/annotations honesty. Local full unit:
940 passed.
…ed otherwise
OpenAI fine-tune style message weight is not applied on this gateway.
Accept null/0/1 as honest no-ops; reject other types and values with
invalid_message_weight. Tip substrate from #578. Local full unit: 943 passed.
…ion role
Reject unsupported message keys with named unknown_message_fields (not silent
strip or tools-passthrough smuggle). Reject legacy function role with
invalid_message_role migration to tool. Tip substrate from #579. Local full
unit: 947 passed.
OpenAI partial-assistant prefix flag is not applied on this gateway.
null/false are honest no-ops; true and non-booleans fail closed with
invalid_message_prefix. Tip substrate from #580. Local full unit: 950 passed.
…therwise
Named invalid_max_tool_calls on /v1/chat/completions instead of opaque
unknown_fields. Aligns with Responses max_tool_calls honesty; gateway has
no multi-step tool loop.
…losed otherwise
Legacy /v1/completions treated max_tool_calls as unknown_fields. Accept the
key for named invalid_max_tool_calls (null/empty/whitespace omit-equivalent),
matching chat/Responses honesty so SDKs get a clear migration path.
SDK clients often send include_usage/include_obfuscation as JSON null.
Drop null flag values before validation so null (and null+false mixes)
match omit / all-false no-ops on chat, Completions, and Responses.
True flags remain fail-closed with invalid_stream_options.
…age honesty
Null include_usage/include_obfuscation stay omit-equivalent, but unknown
stream_options keys no longer become no-ops just because their value is
null. Weight, prefix, refusal, annotations, developer role, empty
user/system content, and participant name now use the same named errors
on the tools passthrough path as on orchestration.
Co-authored-by: Seongho Bae <seonghobae@users.noreply.github.com>
Hoist stream, required model, stream_options, and temperature/top_p range
checks before proxy_completion so a tools or response_format body cannot
return a billed JSON completion when the SDK asked for SSE, or silently
pick a pool model when model is omitted.
Co-authored-by: Seongho Bae <seonghobae@users.noreply.github.com>
Hoist stream, required model, stream_options, and temperature/top_p range
checks before proxy_completion so a tools or response_format body cannot
return a billed JSON completion when the SDK asked for SSE, or silently
pick a pool model when model is omitted.
Co-authored-by: Seongho Bae <seonghobae@users.noreply.github.com>
Hoist attribution and routing validation before proxy_completion so a
tools or response_format body cannot bill a sync completion with an
unknown spend dimension or a batch/latency_tolerant hint. Tools
passthrough has no batch job plane.
Buyer next action: send known sync attribution; omit routing.channel=batch
and latency_tolerant=true on tool-calling requests.
Co-authored-by: Seongho Bae <seonghobae@users.noreply.github.com>
Call the orchestration message, max_tokens, attribution, and routing
validators before proxy_completion so a tools or response_format body
cannot bill a completion with no prompt, crash on a non-object entry,
or silently drop unknown spend/routing keys.
Co-authored-by: Seongho Bae <seonghobae@users.noreply.github.com>
Match _validate_messages: tools + user content 123 must 400 invalid_message
instead of billing a sync completion. Assistant/tool JSON null stays
omit-equivalent.
Buyer next action: send user/system content as a non-empty string or a
content-parts array.
Co-authored-by: Seongho Bae <seonghobae@users.noreply.github.com>
Match the tools-path honesty contract to the invalid_message behavior
so SDK clients send a string or content-parts array.
Co-authored-by: Seongho Bae <seonghobae@users.noreply.github.com>
Passthrough has no batch job plane. Reject routing.channel=batch and
latency_tolerant=true before proxy_completion so a tool-calling body
cannot bill a silent sync completion.
Co-authored-by: Seongho Bae <seonghobae@users.noreply.github.com>
Hoist the remaining chat request knobs before proxy_completion so an
OpenAI SDK tool-calling body cannot bill a sync completion for seed,
stop, n>1, logprobs, logit_bias, out-of-range token/penalty values,
unsupported reasoning_effort, or a non-default service_tier.
Buyer next action: omit those fields on tool-calling requests; they are
not applied on this gateway.
Co-authored-by: Seongho Bae <seonghobae@users.noreply.github.com>
Call _validate_messages before proxy_completion so tools + [] / omitted /
null / non-list / non-object messages raise invalid_message instead of
billing a completion with no prompt.
Buyer next action: always send a non-empty messages array of objects on
tool-calling requests.
Co-authored-by: Seongho Bae <seonghobae@users.noreply.github.com>
Hoist the orchestration request-knob validators before proxy_completion
so a tools or response_format body cannot bill a completion that silently
drops seed, stop, n>1, logprobs, logit_bias, or out-of-range penalties.
Co-authored-by: Seongho Bae <seonghobae@users.noreply.github.com>
stream_chat defaulted to 0.2 even after the HTTP path wrote
default_temperature from the request. A streamed invoice summary at
temperature=0.8 no longer silently falls back to 0.2.
Co-authored-by: Seongho Bae <seonghobae@users.noreply.github.com>
… proxy
SDK optional defaults serialize omitted fields as JSON null. Accepting those
keys without popping them is not omit-equivalent: proxy_completion forwards
the body and several providers reject a null JSON Schema object.
Pop the keys in place, keep non-null wrong types on invalid_tools, and assert
omit-real via mock echo on chat and Responses.
Co-authored-by: Seongho Bae <seonghobae@users.noreply.github.com>
OpenAI SDK tool callers send stream=true. Returning 400 was honest but
left every streaming tool client failing. Pipe a single pool agent as
chat.completion.chunk frames so concatenated mock content matches the
non-stream JSON body, and pipe live provider text so tool_calls deltas
survive. Keep include_usage, empty messages, and unsupported knobs
fail-closed.
Buyer next action: send stream=true on tool-calling requests; omit
stream_options.include_usage.
Co-authored-by: Seongho Bae <seonghobae@users.noreply.github.com>
Keep tools/response_format SSE proxy from this branch and take #601's
null tool.function omit-pop, request-knob HTTP cases, and streamed
route temperature. Conflict resolution keeps stream=true as SSE, not
400. CHANGELOG records the combined unreleased slice.
Co-authored-by: Seongho Bae <seonghobae@users.noreply.github.com>
The reason will be displayed to describe this comment to others. Learn more.
Stale comment
Next slice already ported — cherry-pick, do not open a parallel stack
This head is the current honesty + SSE-tools tip. Do not merge #582–#605 in parallel.
Remaining fail-open named in the last review: tools + unknown tool_calls entry/function keys still 200. That hole is closed on #591 (b7f81a6) and already ported as cherry-pick source a3b7da0 on cursor/bc-b3f7c89f-c445-4d73-a0cd-16d6a8452e1d-1c23 (tests/test_tool_calls_entry_keys_http_honesty.py). Also port 4e3bc12 if this merge dropped thread-local ModelClient.request_sampling.
Do not merge that cherry-pick branch as a 140-file stack. Put a3b7da0 on this tip, then hoist include_orchestration_trace / mode type checks before proxy_completion / SSE proxy.
Buyer next action: send only id / type / function / optional index on assistant tool_calls. Omit stream only if you still need JSON; SSE tools passthrough is this tip's unique work. After the key port: rebase #598a50911f KV allowlist onto this head.
Independent non-author APPROVE + Full unit / Semgrep still required. Do not self-approve.
Do not merge this draft in parallel with #600 / #601 / #608.
This is the SSE tools tip: proxy_completion_stream + live _stream_raw so provider delta.tool_calls survive. Keep it as the stream-path landing vehicle. Independent non-author APPROVE + Full unit / Semgrep still required. This automation will not APPROVE or merge.
Residual buyer gap on the mock path: _mock_raw_sse still emits content chunks and finish_reason=stop even when tools is present. Offline mock:// tool-calling streams therefore do not look like OpenAI streamed function calls. Live providers are already covered by test_stream_raw_pipes_tool_call_deltas_verbatim.
Do not open another SSE stack. Port only the mock tool_calls frames (function name from the first tool) onto this head. A discarded parallel at 364554f on cursor/bc-4cb50318-065a-46bd-aedf-e0abfaf37c2c-8155 has that mock plus tests/test_passthrough_tools_stream_sse_http_honesty.py — cherry-pick the mock, not the stack.
Buyer next action: on this tip, send stream=true with a pool model and a non-empty messages array; omit stream_options.include_usage. After this lands: unknown tool_calls entry keys (a3b7da0), then rebase #598 / #608 KV allowlist onto this tip.
Unknown assistant tool_calls entry/function fields, non-boolean
include_orchestration_trace, and unknown mode used to bill a JSON or SSE
completion on the tools proxy. They now return the same named 400s as the
orchestration path before the first byte.
Buyer next action: send only id/type/function/optional index on assistant
tool_calls; send a boolean or omit include_orchestration_trace; send
auto/route/conduct or omit mode.
Co-authored-by: Seongho Bae <seonghobae@users.noreply.github.com>
The reason will be displayed to describe this comment to others. Learn more.
Stale comment
Review
#606 at 0717921 still frames mock:// tools as delta.content + finish_reason=stop. A realistic invoice-lookup body (tools[0].function.name = lookup_balance, user text INV-9) must emit delta.tool_calls and finish as tool_calls so offline SDK clients match the live _stream_raw path this PR already preserves.
That helper change is on successor #61763089e4 (this branch plus the mock tool_calls frames). Prefer #617 over merging #606. Do not merge #582–#606 in parallel.
Buyer next action: point stream clients at #617; send stream=true and include the invoice id in the user message.
Independent non-author APPROVE + Full unit/Semgrep still required. This automation does not self-approve.
The reason will be displayed to describe this comment to others. Learn more.
Do not merge this tip. Live _stream_raw already preserves delta.tool_calls, but _mock_raw_sse still finishes every tools stream as stop with message.content chunks (contextual_orchestrator/orchestrator.py around the final _frame({}, finish="stop")). An invoice-lookup body (tools[0].function.name = lookup_balance) must emit delta.tool_calls and finish_reason=tool_calls so offline SDK clients match the billed provider stream.
The fix is already on successor #620 (8d7d534): JSON and SSE reconstruct to the same message.tool_calls, INV- binds from the user text (default INV-9), and tool_choice=none stays content/stop.
Buyer next action: review and land #620 instead of this head. Send stream=true on tool-calling requests and put the invoice id in the user message. Do not merge #606 / #609 / #610 / #613 in parallel. Independent non-author APPROVE is still required — this run cannot approve its own stack.
The reason will be displayed to describe this comment to others. Learn more.
Do not merge this tip. _mock_raw_sse still finishes tools as stop with content chunks, so an invoice lookup_balance stream does not match the live _stream_rawdelta.tool_calls shape.
Land #620 (baf2e84) instead. That tip already contains this stack plus:
mock JSON/SSE tool_calls (INV- binds from the user text, default INV-9; tool_choice=none stays content/stop)
streamed route top_p / penalties matching chat()
APA 7th citations (OpenAI function calling; Schick et al., 2023; Yao et al., 2023; Holtzman et al., 2020)
#617 is the same functional mock+sampling slice without those citations. Do not merge #606 / #609 / #610 / #613 / #617 in parallel with #620.
Buyer next action: review #620, send stream=true on tool-calling requests, and put the invoice id in the user message. Independent non-author APPROVE is still required.
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
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.
![cursor[bot]](https://avatars.githubusercontent.com/in/1210556?s=60&v=4){kind=small}
Summary
Successor to #596 at
7a07562. Keeps that head's fail-closed empty messages, spend/routing hoist, and request-knob honesty, and closes the buyer-visible gap those drafts still list as residual:tools/response_format+stream=trueused to return400 invalid_stream(or, before that, a billed JSON completion while the SDK waited for SSE).TaskOrchestrator.proxy_completion_stream/ModelClient.proxy_stream_sendSSE-proxy a single pool agent.chat.completion.chunkwhose concatenated content equals the non-stream JSONmessage.content.delta.tool_callssurvive (content-only parsers drop them).seed/stop/n>1/logprobs, batch routing, andstream_options.include_usage=truestill fail closed before the first byte./v1/responsesstreaming stays rejected.Buyer next action: send
stream=trueon tool-calling requests when the client reads SSE; omitstream_options.include_usage; still send a non-emptymessagesarray of objects and known sync attribution.Prefer this head over #582–#587, #589, #591–#597. Do not merge those in parallel. Independent non-author APPROVE + Full unit/Semgrep still required.
Test plan
python3 tests/test_passthrough_sse_tools_http_honesty.pypython3 tests/test_passthrough_stream_model_http_honesty.pypython3 tests/test_passthrough_attribution_routing_http_honesty.pypython3 tests/test_chat_messages_array_tools_passthrough_http_honesty.pypython3 tests/test_message_honesty_tools_passthrough_http_honesty.pypython3 tests/test_openai_passthrough.pytest_true_streaming.pypython3 tests/test_paper_contracts.pytest_self_check.pytest_conventions.pytest_api_contract.pyDocs
docs/rest_api_design.mdhonesty contract (APA: OpenAI, 2024; WHATWG SSE).docs/library_research.mdPonytail note: stdlib urllib, no httpx/sse-starlette.stream=trueon tool-calling requests.