Fix incorrect formatting in OpenAI responses API for tool

[noorbhatia]

Fix #124

[Copilot]

Pull request overview

Fixes incorrect OpenAI Responses API request-body formatting for tool-call round trips in OpenAILanguageModel (Fix #124), aligning it with Responses API conventions rather than Chat Completions conventions.

Changes:

• Formats function_call_output.output as a direct value (plain string for .text, JSON array for .blocks) instead of JSON-stringifying “content blocks”.
• Flattens image_url in tool output blocks to a string URL (Responses format).
• Converts Chat Completions-style raw assistant tool_calls messages into Responses function_call input items.

---

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.

[Copilot]

The pattern matches against dict["role"] / dict["tool_calls"] are missing optional unwrapping (?). Since dict[...] is optional (as used elsewhere in this file, e.g. case let .string(type)? = obj["type"]), this won’t compile. Update these to case .string(let role)? = ... / case .array(let toolCallsArr)? = ... (or unwrap with guard let).

Suggested change

	case .string(let role) = dict["role"],
	role == "assistant",
	case .array(let toolCallsArr) = dict["tool_calls"]
	{
	for tc in toolCallsArr {
	guard case .object(let tcDict) = tc,
	case .string(let tcId) = tcDict["id"],
	case .object(let fnDict) = tcDict["function"],
	case .string(let fnName) = fnDict["name"],
	case .string(let fnArgs) = fnDict["arguments"]
	case .string(let role)? = dict["role"],
	role == "assistant",
	case .array(let toolCallsArr)? = dict["tool_calls"]
	{
	for tc in toolCallsArr {
	guard case .object(let tcDict) = tc,
	case .string(let tcId)? = tcDict["id"],
	case .object(let fnDict)? = tcDict["function"],
	case .string(let fnName)? = fnDict["name"],
	case .string(let fnArgs)? = fnDict["arguments"]

[Copilot]

Same optional-pattern issue inside the tool call loop: tcDict["id"], tcDict["function"], fnDict["name"], and fnDict["arguments"] are dictionary subscripts and need ? in the pattern match (or explicit unwrapping). As written, this block won’t compile.

Suggested change

	case .string(let tcId) = tcDict["id"],
	case .object(let fnDict) = tcDict["function"],
	case .string(let fnName) = fnDict["name"],
	case .string(let fnArgs) = fnDict["arguments"]
	case .string(let tcId)? = tcDict["id"],
	case .object(let fnDict)? = tcDict["function"],
	case .string(let fnName)? = fnDict["name"],
	case .string(let fnArgs)? = fnDict["arguments"]

[Copilot]

This PR changes request-body formatting for Responses tool-call round trips, but there’s no deterministic unit test asserting the generated input items (e.g., .raw assistant+tool_calls → function_call items, and .tool → function_call_output with a plain string output). Adding a unit test around Responses.createRequestBody would prevent regressions without requiring a live API key.