fix(api): treat padded tool_choice none as omit on mock tools by cursor[bot] · Pull Request #639 · ContextualWisdomLab/contextual-orchestrator

cursor · 2026-08-16T16:21:22Z

Successor to #620. Same SSE mock tool_calls + streamed route sampling tip, plus the padded-none honesty hole.

Unique vs #620 (`baf2e84`)

_selected_function_name treats whitespace-padded tool_choice (" none ", "\tnone\n") as none so mock lookup_balance stays content / stop.
_validate_chat_tool_choice writes the stripped token back so live providers see none, not the padded SDK string.
HTTP JSON + SSE honesty covers exact and padded none (tests/test_passthrough_sse_tools_http_honesty.py).
Holtzman et al. (2020) is now next to the buyer-facing nucleus paragraph in docs/rest_api_design.md.

Buyer next action

Send tool_choice=none when you want no tool call. Incidental padding is still omit-equivalent. Include the invoice id in the user text on tool-calling streams. Do not merge #620 in parallel — this tip already contains that stack plus the padded-none fix.

Do not merge in parallel

Prefer this over #620 / #606 / #617 / #609 / #610 / #613 / #615 for the stream-path tip. Keep KV/token #621, message-name #623, verify #618, and Responses stream #628 on their own tips. Independent non-author APPROVE + Full unit/Semgrep required.

Local checks printed ok: SSE tools honesty (including the new padded-none cases), padded tool_choice strip, chat tool_choice functions, paper/API/self-check/conventions.

…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 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>

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>

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>

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>

mock:// lookup_balance now returns assistant tool_calls and streams delta.tool_calls with finish_reason=tool_calls so offline SDK clients match the live _stream_raw path. Invoice identifiers bind from the user text (default INV-9). Cite Toolformer, ReAct, and OpenAI function calling (APA 7th) in docs/papers and the honesty test. Buyer next action: send stream=true on tool-calling requests and include the invoice id in the user message. Co-authored-by: Seongho Bae <seonghobae@users.noreply.github.com>

stream_chat recorded temperature but omitted the request-scoped top_p, presence_penalty, and frequency_penalty that chat() already forwards. A streamed invoice summary at top_p=0.1 no longer drops those knobs. Buyer next action: send the nucleus and penalty values you want on stream=true route requests; streaming matches the JSON path. Co-authored-by: Seongho Bae <seonghobae@users.noreply.github.com>

Streamed top_p / penalties now match chat(). Record the ICLR 2020 nucleus-sampling paper (APA 7th) so the buyer can see why those knobs must not silently drop on stream=true. Buyer next action: send the nucleus and penalty values you want on streamed route requests. Co-authored-by: Seongho Bae <seonghobae@users.noreply.github.com>

The HTTP validator already accepted " none " / tab-newline none as none but discarded the stripped token. Mock selection required an exact "none", so a padded none still emitted tool_calls after a 200. Strip in the selector and write the token back so live providers see none. JSON and SSE honesty tests cover exact and padded none. Buyer next action: send tool_choice=none to skip tools; incidental padding stays omit-equivalent. Co-authored-by: Seongho Bae <seonghobae@users.noreply.github.com>

Buyers type "invoice 4419" more often than "INV-4419". The mock binder only matched an INV- prefix, so a realistic prompt still defaulted to INV-9. Accept invoice / invoice id / invoice # plus digits. Buyer next action: put the invoice number in the user text; prefixed and bare forms both bind. Co-authored-by: Seongho Bae <seonghobae@users.noreply.github.com>

cursor · 2026-08-16T16:32:30Z

Bugbot is not enabled for your account, so this pull request was not reviewed.

Enable Bugbot in the Cursor dashboard to get automatic reviews on future PRs.

cursor

Stale comment

Review (#639 unique slice vs #620 baf2e84)

Padded tool_choice none stripping + validator writeback, bare-invoice invoice 4419 → INV-4419, and the Holtzman cite next to the nucleus paragraph are sound. HTTP JSON+SSE covers exact and padded none. Local tests/test_passthrough_sse_tools_http_honesty.py printed ok on this head.

No correctness blocker in this slice. Do not self-approve. Do not merge #620 / #606 / #617 / #609 / #610 / #613 / #615 in parallel with this stream-path tip. Independent non-author APPROVE + Full unit/Semgrep still required before merge.

Buyer next action

Send tool_choice=none when you want no tool call; incidental padding is still omit-equivalent. Put the invoice number in the user text (INV-4419 or invoice 4419). After the first tool_calls hop, POST the tool output as role=tool with the same tool_call_id — that second hop is the remaining buyer gap and is landed on the successor tip, not here.

Follow-up (not this PR)

Empty/whitespace tool_choice with tools is accepted as omit, then the mock still emits lookup_balance unless the empty token is written back as none.

Bare-invoice bind is in-process only; add one _post_raw JSON+SSE pair for invoice 4419.

Second-hop role=tool continuation (LangChain/OpenAI SDK default) must synthesize content / stop from the observed values. That slice is the next stream-path landing vehicle — do not fold it back onto this tip after it opens.

Checks on this synchronize were still queued at review time. Treat CI wait as non-blocking; re-run only if a job fails with a real finding.

_{Sent by Cursor Automation: Fix Issues}

cursor

Review (#639 unique slice vs #620 `baf2e84`)

Padded tool_choice=none is content/stop on mock JSON and SSE. _validate_chat_tool_choice writes none / auto / required back. Bare invoice 4419 binds INV-4419 in proxy_completion. That unique work is implemented correctly and does not weaken the security gate.

Remaining buyer-visible holes on this head:

Write-back is untested. HTTP never asserts echo.tool_choice == "none". Deleting the write-back still passes.
Bare invoice bind is proxy_completion only. /v1/chat/completions and SSE are unlocked.
Mock tool loop. A role=tool observation still emits another lookup_balance. Official SDK / LangChain clients spin.

Do not merge #639 while a successor holds the stream-path tip. Prefer #660 (d074598, synthesize bound tool results + HTTP invoice + empty tool_choice write-back). #674 (594cb62) also stops the loop and locks echo write-back, but answers [agent] chat-mock instead of the observation amount — keep it behind #660. Do not merge #620 / #606 / #617 in parallel.

Buyer next action: send tool_choice=none to skip tools; put the invoice number in the user text; after the first tool_calls hop post role=tool with the same tool_call_id and read the answer on #660. Independent non-author review is still required. This automation cannot merge.

Local checks on the #674 successor printed ok for SSE tools honesty, paper/API/self-check/conventions, tool_call_id, and passthrough.

_{Sent by Cursor Automation: Fix Issues}

cursor · 2026-08-16T16:50:18Z

+            streamed, finish_reason = _reconstruct_tool_calls(sse)
+            assert finish_reason == "stop", choice
+            assert streamed == []
+            assert _reconstruct_content(sse) == reference["message"]["content"]


This HTTP loop asserts finish_reason=stop and no tool_calls, but never echo.tool_choice == "none". _selected_function_name strips on its own, so deleting the validator write-back at server.py:2969 still stays green while HTTPS would forward " none ".

Lock it here: assert json.loads(json_raw)["echo"]["tool_choice"] == "none".

#674 also writes the stripped token in _passthrough_upstream so proxy_completion echo matches. Prefer #660 for the stream-path tip (this slice plus bound tool-result synthesis). Do not merge this head in parallel.

cursor · 2026-08-16T16:50:19Z

+    assert result["choices"][0]["finish_reason"] == "tool_calls"
+    assert message["tool_calls"][0]["function"]["name"] == "lookup_balance"
+    assert json.loads(message["tool_calls"][0]["function"]["arguments"]) == {
+        "invoice_id": "INV-4419"


Bare invoice 4419 → INV-4419 is only asserted through proxy_completion on this head. A buyer hitting /v1/chat/completions or SSE is the path that matters.

#660 and #674 both add test_http_chat_tools_stream_binds_bare_invoice_number. Prefer #660: after the first tool_calls hop it synthesizes the observation (balance_usd=128.50) instead of looping lookup_balance or returning [agent] chat-mock. Do not duplicate that test here.

seonghobae and others added 29 commits August 16, 2026 21:18

fix(api): treat message prefix null/false as omit; fail-closed on true

82038d3

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.

fix(api): treat chat max_tool_calls null/empty as omit; fail-closed o…

1a196b0

…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.

docs(api): tell buyers non-string user content fails closed on tools

f6dfae5

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>

docs: tell buyers to omit batch routing on tool-calling requests

dbfa0e5

Co-authored-by: Seongho Bae <seonghobae@users.noreply.github.com>

docs: record passthrough knob and stream temperature fixes in CHANGELOG

7e9a339

Co-authored-by: Seongho Bae <seonghobae@users.noreply.github.com>

cursor Bot requested a review from seonghobae August 16, 2026 16:21

cursor Bot mentioned this pull request Aug 16, 2026

fix(api): pop null tool.function optional fields omit-real before proxy #638

Open

3 tasks

seonghobae marked this pull request as ready for review August 16, 2026 16:32

seonghobae enabled auto-merge (squash) August 16, 2026 16:32

cursor Bot mentioned this pull request Aug 16, 2026

fix(api): pop null json_schema description/strict omit-real before proxy #646

Closed

7 tasks

cursor Bot commented Aug 16, 2026

View reviewed changes

Comment thread contextual_orchestrator/server.py

Comment thread contextual_orchestrator/orchestrator.py

Comment thread tests/test_passthrough_sse_tools_http_honesty.py

This was referenced Aug 16, 2026

fix(api): synthesize mock content after bound tool results #660

Open

fix(api): stop mock tool loop after role=tool observation #674

Closed

cursor Bot commented Aug 16, 2026

View reviewed changes

cursor Bot mentioned this pull request Aug 16, 2026

fix(api): bind invoice no/nr/inv# on mock lookup_balance #682

Closed

opencode-agent Bot disabled auto-merge August 16, 2026 19:24

Navigation Menu

Uh oh!

fix(api): treat padded tool_choice none as omit on mock tools - #639

fix(api): treat padded tool_choice none as omit on mock tools#639
cursor[bot] wants to merge 30 commits into
mainfrom
cursor/bc-1a54e378-9680-4d5d-aa2c-459ad8ed9086-2a63

cursor Bot commented Aug 16, 2026

Uh oh!

cursor Bot commented Aug 16, 2026

Uh oh!

cursor Bot left a comment •

edited

Loading

Uh oh!

Uh oh!

Uh oh!

Uh oh!

cursor Bot left a comment

Uh oh!

cursor Bot Aug 16, 2026

Uh oh!

cursor Bot Aug 16, 2026

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

2 participants

Uh oh!

Conversation

cursor Bot commented Aug 16, 2026

Unique vs #620 (baf2e84)

Buyer next action

Do not merge in parallel

Uh oh!

cursor Bot commented Aug 16, 2026

Uh oh!

cursor Bot left a comment • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Choose a reason for hiding this comment

Review (#639 unique slice vs #620 baf2e84)

Buyer next action

Follow-up (not this PR)

Uh oh!

Uh oh!

Uh oh!

Uh oh!

cursor Bot left a comment

Choose a reason for hiding this comment

Review (#639 unique slice vs #620 baf2e84)

Uh oh!

cursor Bot Aug 16, 2026

Choose a reason for hiding this comment

Uh oh!

cursor Bot Aug 16, 2026

Choose a reason for hiding this comment

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

2 participants

Unique vs #620 (`baf2e84`)

cursor Bot left a comment •

edited

Loading

Review (#639 unique slice vs #620 `baf2e84`)

Review (#639 unique slice vs #620 `baf2e84`)