Add bounded tool-result metadata support to executed tool calls (#44336)

## What changed

- Add host-recorded `tool_result_metadata` snapshots with size limits, redacted debug output, and protection against deserialization from untrusted input. MCP capture remains disabled.
- Preserve accepted metadata across retries and delayed Code Mode results. Shed raw metadata before source evidence, calls, or completion markers when the request budget is exceeded.
- Send raw metadata only to allowed OpenAI and ChatGPT HTTPS endpoints, filtering by the resolved destination for HTTP and WebSocket requests.
- Encode source parse failures as an array entry with type `parse_failed` and an empty ID.

## Testing

Add coverage for metadata bounds and redaction, retained-call updates, budget shedding, disabled MCP capture, and HTTP/WebSocket destination filtering.

GitOrigin-RevId: b1a396323ad0623167cdd9e20ff9453f0d01bbf7
This commit is contained in:
ningyi-oai authored and copyberry committed 2026-09-09 23:46:17 +00:00
1 parent b5544d5732
commit 0df6366a87
18 files changed
+1198 -120

No files matched your search

+1
View File
@@ -44,6 +44,7 @@ pub use executed_tool_calls::ExecutedToolCall;
pub use executed_tool_calls::ExecutedToolCallArguments;
pub use executed_tool_calls::ExecutedToolCallTruncation;
pub use executed_tool_calls::MAX_TOOL_RESULT_SOURCE_FIELD_BYTES;
pub use executed_tool_calls::ToolResultMetadata;
pub use executed_tool_calls::ToolResultSource;
pub use executed_tool_calls::ToolResultSources;
pub use executed_tool_calls::bound_executed_tool_calls_for_prompt;