refactor: decompose kiro_executor.go into streaming and transform modules

[KooshaPari]

Summary

• Extracted kiro_streaming.go (~2990 LOC) from the monolithic kiro_executor.go:

  • ExecuteStream + executeStreamWithRetry — entry points for streaming requests
  • parseEventStream, readEventStreamMessage, skipEventStreamHeaderValue, extractEventTypeFromBytes — AWS Event Stream binary parser
  • streamToChannel — core SSE-to-channel converter with thinking mode, tool use, and usage tracking
  • Full web search MCP handler: handleWebSearchStream, handleWebSearch, callMcpAPI, webSearchHandler, fetchToolDescription, callKiroAndBuffer, callKiroDirectStream, sendFallbackText, executeNonStreamFallback, CloseExecutionSession

• Extracted kiro_transform.go (~470 LOC) from kiro_executor.go:

  • kiroEndpointConfig type, kiroDefaultRegion const
  • extractRegionFromProfileARN, buildKiroEndpointConfigs, resolveKiroAPIRegion, getKiroEndpointConfigs — region/endpoint resolution
  • isIDCAuth, buildKiroPayloadForFormat, sanitizeKiroPayload — payload transformation
  • kiroCredentials, getMetadataString, getEffectiveProfileArnWithWarning, determineAgenticMode — auth/model helpers
  • mapModelToKiro, kiroModelFingerprint — model ID mapping

• kiro_executor.go is now ~1270 LOC (down from 4691), containing: core KiroExecutor struct, HTTP client pool + retry logic, Execute/executeWithRetry, CountTokens, Refresh, Prepare/HttpRequest, and token persistence helpers.

No public API changes. All symbols remain in the executor package.

Task 3 Note

The task requested migration of openrouter, xai, deepseek, and vertex auth providers to BaseTokenStorage. Upon inspection:

• openrouter, xai, deepseek — auth directories do not exist in this codebase
• vertex — stores service account JSON (not OAuth tokens), structurally incompatible with BaseTokenStorage

No Task 3 commits were made. Candidates for future migration: kimi and iflow (same SaveTokenToFile pattern as providers already migrated).

Test plan

• go build ./pkg/llmproxy/executor/ passes (pre-existing qwen dependency error unrelated)
• gofmt -l reports no formatting issues on the three kiro files
• No duplicate symbol declarations across the three files
• Existing tests in kiro_executor_*_test.go continue to compile and pass

🤖 Generated with Claude Code

[gemini-code-assist]

Warning

You have reached your daily quota limit. Please wait up to 24 hours and I will start processing your requests again!

[coderabbitai]

Warning

Rate limit exceeded

@KooshaPari has exceeded the limit for the number of commits that can be reviewed per hour. Please wait 14 minutes and 3 seconds before requesting another review.

⌛ How to resolve this issue?

After the wait time has elapsed, a review can be triggered using the @coderabbitai review command as a PR comment. Alternatively, push new commits to this PR.

We recommend that you space out your commits to avoid hitting the rate limit.

🚦 How do rate limits work?

CodeRabbit enforces hourly rate limits for each developer per organization.

Our paid plans have higher rate limits than the trial, open-source and free plans. In all cases, we re-allow further reviews after a brief timeout.

Please see our FAQ for further information.

📥 Commits

Reviewing files that changed from the base of the PR and between ffe4146 and c615f6b.

📒 Files selected for processing (3)

• pkg/llmproxy/executor/kiro_executor.go
• pkg/llmproxy/executor/kiro_streaming.go
• pkg/llmproxy/executor/kiro_transform.go

✨ Finishing Touches

🧪 Generate unit tests (beta)

• Create PR with unit tests
• Post copyable unit tests in a comment
• Commit unit tests in branch refactor/phase2-streaming-transform

---

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.

❤️ Share

• X
• Mastodon
• Reddit
• LinkedIn

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

[KooshaPari]

Review: kiro_executor.go reduced from 4.7K to 1.3K LOC. Streaming and transform modules cleanly extracted. LGTM.

[Copilot]

Pull request overview

Refactors the Kiro executor implementation by decomposing the previously monolithic kiro_executor.go into focused streaming and transform modules, keeping everything within the executor package and preserving existing exported APIs.

Changes:

• Extracts streaming execution, AWS Event Stream parsing, SSE chunk emission, and MCP web search handling into kiro_streaming.go.
• Extracts endpoint/region resolution, payload shaping, auth/model helper utilities into kiro_transform.go.
• Leaves kiro_executor.go as the smaller “core” executor with request execution, retry, refresh, and token persistence helpers.

Reviewed changes

Copilot reviewed 1 out of 3 changed files in this pull request and generated 4 comments.

File	Description
pkg/llmproxy/executor/kiro_streaming.go	Streaming request path, Event Stream parsing, SSE conversion, and MCP web search flow extracted into a dedicated module.
pkg/llmproxy/executor/kiro_transform.go	Region/endpoint resolution and request/payload/model transformation helpers extracted into a dedicated module.
pkg/llmproxy/executor/kiro_executor.go	Core executor retained after moving streaming/transform concerns into new files (per PR description).

Comments suppressed due to low confidence (1)

pkg/llmproxy/executor/kiro_transform.go:296

• There is a standalone comment block describing findRealThinkingEndTag, but the function itself is not present in this file. This reads like a leftover from the pre-refactor executor and is likely to confuse future readers. Either move this comment next to the function implementation (wherever it now lives) or remove it.

// findRealThinkingEndTag finds the real </thinking> end tag, skipping false positives.
// Returns -1 if no real end tag is found.
//
// Real </thinking> tags from Kiro API have specific characteristics:
// - Usually preceded by newline (.\n</thinking>)
// - Usually followed by newline (\n\n)
// - Not inside code blocks or inline code
//
// False positives (discussion text) have characteristics:
// - In the middle of a sentence
// - Preceded by discussion words like "标签", "tag", "returns"
// - Inside code blocks or inline code
//
// Parameters:
// - content: the content to search in
// - alreadyInCodeBlock: whether we're already inside a code block from previous chunks
// - alreadyInInlineCode: whether we're already inside inline code from previous chunks

// determineAgenticMode determines if the model is an agentic or chat-only variant.

---

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

[Copilot]

The docstring for getKiroEndpointConfigs says it supports region resolution from the auth metadata "region" field, but resolveKiroAPIRegion() explicitly does not consult that field (and the comment below also says it should not be used). Please update the comment to match the actual region sources (api_region + profile_arn + default), or implement support for the "region" metadata if that’s intended.

Suggested change

	// Supports dynamic region based on auth metadata "api_region", "profile_arn", or "region" field.
	// Supports reordering based on "preferred_endpoint" in auth metadata/attributes.
	//
	// Region priority:
	// 1. auth.Metadata["api_region"] - explicit API region override
	// 2. ProfileARN region - extracted from arn:aws:service:REGION:account:resource
	// 3. kiroDefaultRegion (us-east-1) - fallback
	// Note: OIDC "region" is NOT used - it's for token refresh, not API calls
	// Supports dynamic region based on auth metadata "api_region" and "profile_arn".
	// Supports reordering based on "preferred_endpoint" in auth metadata/attributes.
	//
	// Region priority:
	// 1. auth.Metadata["api_region"] - explicit API region override
	// 2. ProfileARN region - extracted from arn:aws:service:REGION:account:resource
	// 3. kiroDefaultRegion (us-east-1) - fallback
	// Note: OIDC "region" is NOT used for API calls – it's only for token refresh

[Copilot]

streamToChannel() allocates a 20MB bufio reader buffer per request. This is the only occurrence of a 20MB reader buffer in the executor package, so at moderate concurrency it can create significant transient memory pressure. Consider using the default bufio size (or a much smaller size like 64–256KB), or making the buffer size configurable if large frames are expected.

Suggested change

	func (e *KiroExecutor) streamToChannel(ctx context.Context, body io.Reader, out chan<- cliproxyexecutor.StreamChunk, targetFormat sdktranslator.Format, model string, originalReq, claudeBody []byte, reporter *usageReporter, thinkingEnabled bool) {
	reader := bufio.NewReaderSize(body, 20*1024*1024) // 20MB buffer to match other providers
	const kiroStreamBufferSize = 256 * 1024
	func (e *KiroExecutor) streamToChannel(ctx context.Context, body io.Reader, out chan<- cliproxyexecutor.StreamChunk, targetFormat sdktranslator.Format, model string, originalReq, claudeBody []byte, reporter *usageReporter, thinkingEnabled bool) {
	reader := bufio.NewReaderSize(body, kiroStreamBufferSize) // reduced buffer to limit per-request memory usage

[Copilot]

The thinkingEnabled parameter is documented as controlling whether <thinking> tags are parsed, but the current implementation does not use it (thinking/tag parsing and reasoningContentEvent handling happen unconditionally). Either wire thinkingEnabled into the parsing/emission logic, or remove/adjust the parameter+comments to reflect the actual always-on behavior to avoid misleading future changes.

Suggested change

	// thinkingEnabled controls whether <thinking> tags are parsed - only parse when request enabled thinking.
	func (e *KiroExecutor) streamToChannel(ctx context.Context, body io.Reader, out chan<- cliproxyexecutor.StreamChunk, targetFormat sdktranslator.Format, model string, originalReq, claudeBody []byte, reporter *usageReporter, thinkingEnabled bool) {
	// thinkingEnabled indicates whether the caller requested thinking, but the current
	// implementation always parses <thinking> tags when present. The parameter is kept
	// for API compatibility and potential future conditional behavior.
	func (e *KiroExecutor) streamToChannel(ctx context.Context, body io.Reader, out chan<- cliproxyexecutor.StreamChunk, targetFormat sdktranslator.Format, model string, originalReq, claudeBody []byte, reporter *usageReporter, thinkingEnabled bool) {
	// thinkingEnabled is currently unused; reasoning content is always parsed when present.
	_ = thinkingEnabled

[Copilot]

buildKiroEndpointConfigs() always includes the CodeWhisperer endpoint even when region is not us-east-1, but the comment states it only works in us-east-1. In non-us-east-1 regions this will add avoidable latency/extra failures when falling back. Consider only appending the CodeWhisperer config when region == kiroDefaultRegion (or explicitly us-east-1), or otherwise gating it behind a feature flag.

Suggested change

	return []kiroEndpointConfig{
	{
	// Primary: Q endpoint - works for all regions and auth types
	URL: fmt.Sprintf("https://q.%s.amazonaws.com/generateAssistantResponse", region),
	Origin: "AI_EDITOR",
	AmzTarget: "", // Empty = don't set X-Amz-Target header
	Name: "AmazonQ",
	},
	{
	// Fallback: CodeWhisperer endpoint (legacy, only works in us-east-1)
	URL: fmt.Sprintf("https://codewhisperer.%s.amazonaws.com/generateAssistantResponse", region),
	Origin: "AI_EDITOR",
	AmzTarget: "AmazonCodeWhispererStreamingService.GenerateAssistantResponse",
	Name: "CodeWhisperer",
	},
	}
	// Primary: Q endpoint - works for all regions and auth types
	endpoints := []kiroEndpointConfig{
	{
	URL: fmt.Sprintf("https://q.%s.amazonaws.com/generateAssistantResponse", region),
	Origin: "AI_EDITOR",
	AmzTarget: "", // Empty = don't set X-Amz-Target header
	Name: "AmazonQ",
	},
	}
	// Fallback: CodeWhisperer endpoint (legacy, only works in us-east-1).
	// Only include this endpoint when the region is the default Kiro region.
	if region == kiroDefaultRegion {
	endpoints = append(endpoints, kiroEndpointConfig{
	URL: fmt.Sprintf("https://codewhisperer.%s.amazonaws.com/generateAssistantResponse", region),
	Origin: "AI_EDITOR",
	AmzTarget: "AmazonCodeWhispererStreamingService.GenerateAssistantResponse",
	Name: "CodeWhisperer",
	})
	}
	return endpoints