Read an explicit null sequence as an empty one

[senamakel]

What changed

Every sequence field an OpenAI-compatible provider can send is now deserialized through Option<Vec<_>>, so an explicit null reads as an empty list instead of failing the decode.

#[serde(default)] covers an absent key and nothing else. A key present with the value null still reaches the Vec visitor and fails the whole response with invalid type: null, expected a sequence.

That is not hypothetical. Mistral-family endpoints spell "the model did not call a tool" as "tool_calls": null on every plain-text completion:

{"choices":[{"index":0,"finish_reason":"stop",
  "message":{"role":"assistant","tool_calls":null,"content":"Hi!"}}]}

Any agent role sitting on such a rung could not complete a single turn — the model answered perfectly and the caller saw a transport fault. Found on a Lean-specialised rung (labs-leanstral-1-5), where a sub-agent delegation failed 100% of the time with serialization error: invalid type: null, expected a sequence.

This is the same tolerance deserialize_arguments already applies one level down to a tool call's arguments, and for the same reason recorded in its doc comment: one provider's unexpected spelling of "nothing" must not fail the decode of everything around it.

Fields covered: choices and message.tool_calls (unary), choices and delta.tool_calls (streaming), the Responses API's output / content / summary, and the model listing's data.

Public API or behavior changes

None to the API. One behavior change, deliberate: a body carrying "choices": null now fails with Model("openai response contained no choices") rather than a serde error — the provider fact the caller can act on, rather than the transport being named for a body it read perfectly. Pinned by a test.

Validation

cargo fmt --all -- --check      ok
cargo clippy --all-targets -- -D warnings   ok, 0 warnings
cargo test -p tinyinference --lib           ok, 266 passed

cargo test --doc has 5 pre-existing failures in src/embeddings/ — present on main before this branch and untouched by it.

Three new tests: a_null_tool_calls_array_is_read_as_no_tool_calls, a_null_choices_array_is_read_as_no_choices, a_null_tool_calls_delta_is_read_as_no_fragments.

Summary by CodeRabbit

• Bug Fixes

  • Improved compatibility with OpenAI-compatible providers that return null for optional list fields.
  • Prevented deserialization failures for response choices, tool calls, streamed content, and model listings.
  • Improved error reporting when response candidates are unavailable.

• Tests

  • Added coverage for null-valued tool calls, choices, and streaming response fields.

[chatgpt-codex-connector]

Codex Review Summary

This comment shows the latest Codex review activity on this pull request.

Review	Status	Commit	Review trigger
📝 Code Review	✅ Completed 2026-08-31T11:44:07.414613Z	ab54e4e	PR opened

ℹ️ About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

• Open a pull request for review
• Mark a draft as ready
• Comment "@codex review" or "@codex security review".

Codex reacts with 👀 while any review is running, comments if it has suggestions, and reacts with 👍 once all reviews finish with no findings.

[coderabbitai]

No actionable comments were generated in the recent review. 🎉

ℹ️ Recent review info

⚙️ Run configuration

Configuration used: Organization UI

Review profile: CHILL

Plan: Pro Plus

Run ID: be808545-ab20-48f1-ab11-93c16fef95b0

📥 Commits

Reviewing files that changed from the base of the PR and between cc8aca4 and ab54e4e.

📒 Files selected for processing (3)

• crates/tinyinference/src/providers/openai/responses.rs
• crates/tinyinference/src/providers/openai/test.rs
• crates/tinyinference/src/providers/openai/types.rs

Included review availability: Your plan provides up to 1 included review per hour; 0 remain after this review.

---

📝 Walkthrough

Walkthrough

OpenAI response collection fields now deserialize explicit JSON null values as empty vectors. Tests cover non-streaming messages, choices, streaming tool-call deltas, response outputs, and model lists.

Changes

OpenAI null collection handling

Layer / File(s)	Summary
Null collection deserialization crates/tinyinference/src/providers/openai/types.rs, crates/tinyinference/src/providers/openai/responses.rs	Adds deserialize_null_as_empty and applies it to OpenAI choices, tool calls, model data, response output, content, and summary fields.
Null response coverage crates/tinyinference/src/providers/openai/test.rs	Adds tests for null tool calls, null choices, and null streaming tool-call deltas.

Estimated code review effort: 2 (Simple) | ~10 minutes

Merge Risk: ⚪ Minimal · up to ab54e

This localized change makes explicit null sequence fields decode as empty collections, with targeted tests and passing checks; no actionable merge-blocking risk remains, so it is merge-ready after normal checks.

Poem

A rabbit hops through JSON rain
Null arrays leave no broken chain
Empty vectors fill the tray
Tests keep stray decode errors away
OpenAI paths now parse the way

🚥 Pre-merge checks | ✅ 5

✅ Passed checks (5 passed)

Check name	Status	Explanation
Description Check	✅ Passed	Check skipped - CodeRabbit’s high-level summary is enabled.
Title check	✅ Passed	The title clearly and concisely describes the main change: treating explicit null sequence values as empty sequences during deserialization.
Docstring Coverage	✅ Passed	Docstring coverage is 100.00% which is sufficient. The required threshold is 80.00%. Docstring coverage is scoped to functions touched by this diff. Analyzed 5 functions across 3 files.
Linked Issues check	✅ Passed	Check skipped because no linked issues were found for this pull request.
Out of Scope Changes check	✅ Passed	Check skipped because no linked issues were found for this pull request.

✨ Finishing Touches

📝 Generate docstrings

• Create stacked PR
• Commit on current branch

Warning

Your free Security trial is over. An organization admin can activate Security or dismiss this notice.

---

_{Comment @coderabbitai help to get the list of available commands.}

[tinysweeper]

tinysweeper found nothing blocking. Approving.

             $0.0140 · 85,077 in / 1,741 out · 20,745 cached (24%) · deepseek/deepseek-v4-flash, openrouter/openai/text-embedding-3-small, z-ai/glm-5.2 · 302 embedded
critique:    $0.0031 · 33,958 in / 365 out   · 0 cached (0%)       · deepseek/deepseek-v4-flash
security:    $0.0070 · 32,603 in / 433 out   · 17,145 cached (53%) · deepseek/deepseek-v4-flash, z-ai/glm-5.2
tests:       $0.0011 · 12,537 in / 114 out   · 0 cached (0%)       · deepseek/deepseek-v4-flash
description: $0.0026 · 4,821 in  / 534 out   · 3,600 cached (75%)  · z-ai/glm-5.2

[tinysweeper]

How this change flows

2 changed behaviours across 15 relationships. 6 surrounding behaviours are shown (60 graph nodes walked). 43 further behaviours left out to keep the diagram readable.

flowchart LR
  n0["...400_unions_with_existing_baseline_degrade<br/>changed"]:::changed
  n1["ChatCompletionChunk<br/>changed"]:::changed
  n2["...ne_knobs_default_to_supported_wire_shapes"]:::impacted
  n3["translates_request_to_openai_json_shape"]:::impacted
  n4["with_response_format"]:::impacted
  n5["ingest"]:::impacted
  n6["model"]:::impacted
  n7["translate_request"]:::impacted
  n0 -->|calls| n4
  n0 -->|tests| n4
  n2 -->|calls| n4
  n2 -->|tests| n4
  n2 -->|calls| n6
  n2 -->|tests| n6
  n2 -->|calls| n7
  n2 -->|tests| n7
  n3 -->|calls| n4
  n3 -->|tests| n4
  n3 -->|calls| n6
  n3 -->|tests| n6
  n3 -->|calls| n7
  n3 -->|tests| n7
  n5 -->|uses| n1
  classDef changed fill:#0d4429,stroke:#238636,color:#e6edf3
  classDef impacted fill:#161b22,stroke:#6e7681,color:#c9d1d9
  classDef flagged fill:#5a1e02,stroke:#d93f0b,color:#ffffff
  classDef blocking fill:#67060c,stroke:#f85149,color:#ffffff

Loading

_{Green: changed behaviour. Grey: surrounding behaviour. Arrows name the call, use, implementation, or test relationship. Orange: has findings. Red: has a finding that blocks the merge.}