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
{{ message }}
Repository navigation
fix(providers): tolerate null tool_calls in OpenAI-compatible responses - #144
Some OpenAI-compatible inference hosts (e.g. SGLang) send "tool_calls": null in chat-completion responses instead of omitting the field. Serde's #[serde(default)] only substitutes a value for missing keys, so an explicit null failed deserialization with:
json error: invalid type: null, expected a sequence at line 1 column 224
This broke every request (plain prompts and tool calls, streaming and non-streaming) against those hosts, including local SGLang deployments of Qwen3.
Fix
Treat an explicit null as "no tool calls" (empty list) in both response paths:
ChatMessage.tool_calls (non-streaming)
ChunkDelta.tool_calls (streaming)
via a small generic helper deser_null_as_empty_seq. Fully backward compatible — missing keys still default to empty, and real tool-call arrays deserialize as before.
Verification
cargo test -p code-providers --lib — 76/76 pass
cargo fmt --check and cargo clippy -p code-providers --lib clean
End-to-end against a live SGLang endpoint serving Qwen/Qwen3.8-27B:
plain prompt → pong
read_file tool call → file read and version returned
TS SDK → CLI → model round-trip → pong
Summary by CodeRabbit
Bug Fixes
OpenAI responses now handle tool-call fields set to null as empty, preventing deserialization errors for both standard and streaming responses.
Some OpenAI-compatible hosts (e.g. SGLang) send "tool_calls": null
instead of omitting the field. Serde's #[serde(default)] only fills
missing keys, so deserialization failed with 'invalid type: null,
expected a sequence' on every request.
Treat an explicit null as an empty tool-call list in both the
non-streaming (ChatMessage) and streaming (ChunkDelta) paths.
Backward compatible: missing keys still default to empty.
File: No configuration file from this repository applied
Commit: Unavailable
Layers: defaults
Review profile: CHILL
Plan: Advanced
Run ID: 63edff57-35ca-43ce-9fd8-b823a1381aae
Commits
Reviewing files that changed from the base of the PR and between ac72916 and d69df13.
Files selected for processing: 1
cli/README.md
No actionable comments were generated in the recent review.
Walkthrough
The OpenAI provider adds a generic deserializer that treats explicit null tool-call fields as empty vectors. It applies to both non-streaming responses and streaming deltas, while retaining defaults for missing fields.
Changes: OpenAI tool-call deserialization
Layer / File(s)
Summary
Null handling for tool-call sequences crates/code-providers/src/providers/openai.rs
A generic deserializer maps explicit null to an empty vector. Both ChatMessage.tool_calls and ChunkDelta.tool_calls use it alongside their missing-field defaults.
This change lets OpenAI-compatible hosts that return null tool_calls work instead of failing to parse. No merge-blocking risk was identified.
Pre-merge checks: 4
Check name
Status
Explanation
Title check
The title names the main change: tolerating explicit null tool_calls in OpenAI-compatible responses. It follows the conventional commit format and matches the raw summary, which adds a helper that tre…
Description check
The description covers the summary, the root cause (explicit null vs. serde default for missing keys), the fix in both ChatMessage.tool_calls and ChunkDelta.tool_calls, backward compatibility, and ver…
Linked Issues check
Check skipped because no linked issues were found for this pull request.
Out of Scope Changes check
Check skipped because no linked issues were found for this pull request.
Full details: Title check
Explanation
The title names the main change: tolerating explicit null tool_calls in OpenAI-compatible responses. It follows the conventional commit format and matches the raw summary, which adds a helper that treats null as an empty sequence for both non-streaming and streaming tool_calls fields. It is specific and concise.
Full details: Description check
Explanation
The description covers the summary, the root cause (explicit null vs. serde default for missing keys), the fix in both ChatMessage.tool_calls and ChunkDelta.tool_calls, backward compatibility, and verification steps. It does not follow the template's section structure: the Type of Change checkboxes are not marked (Bug fix is clearly the intended type), the Testing checklist is not ticked, the Checklist section is absent, and Additional Notes is missing. It also does not link a related issue, which the template asks for. Most required information is present, so the check passes, though the template sections should be completed.
Resolution
Optionally mark 'Bug fix' under Type of Change, tick the applicable Testing and Checklist items (for example unit tests pass and added tests, if tests were added), and add a related issue reference or an Additional Notes entry if one exists.
Finishing Touches
Generate unit tests (beta)
Commit to this branch
Create a new PR
Autofix · Keep fixing CodeRabbit findings and required CI, and resolving merge conflicts
Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out.
File: No configuration file from this repository applied
Commit: Unavailable
Layers: defaults
Review profile: CHILL
Plan: Advanced
Run ID: 511754b3-6e69-46a7-82d5-ed9c031ec13d
Commits
Reviewing files that changed from the base of the PR and between 728562d and ac72916.
Files selected for processing: 1
crates/code-providers/src/providers/openai.rs
Included review availability: This review used your included allowance. Your plan provides up to 1 included review per hour; 0 remain after this review.
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.
Summary
Some OpenAI-compatible inference hosts (e.g. SGLang) send
"tool_calls": nullin chat-completion responses instead of omitting the field. Serde's#[serde(default)]only substitutes a value for missing keys, so an explicitnullfailed deserialization with:This broke every request (plain prompts and tool calls, streaming and non-streaming) against those hosts, including local SGLang deployments of Qwen3.
Fix
Treat an explicit
nullas "no tool calls" (empty list) in both response paths:ChatMessage.tool_calls(non-streaming)ChunkDelta.tool_calls(streaming)via a small generic helper
deser_null_as_empty_seq. Fully backward compatible — missing keys still default to empty, and real tool-call arrays deserialize as before.Verification
cargo test -p code-providers --lib— 76/76 passcargo fmt --checkandcargo clippy -p code-providers --libcleanQwen/Qwen3.8-27B:pongread_filetool call → file read and version returnedpongSummary by CodeRabbit
nullas empty, preventing deserialization errors for both standard and streaming responses.