fix(kiro): stabilize conversationId across prompt compression

[HALDRO]

Summary

Fixes a ReferenceError: kiroConversationSeed is not defined that crashed every Kiro request after server restart, and replaces the provider-specific seed mechanism with a generic preCompressionBody option.

What was broken

All requests to Kiro providers failed immediately with:

WARN [TRANSLATE] Request translation failed: kiroConversationSeed is not defined

This caused the fallback loop to exhaust all Kiro accounts in ~300ms, returning a 500 to the client on every request.

Root cause

kiroConversationSeed was declared with let inside the if (body && Array.isArray(allMessages)...) block scope in chatCore.ts, but referenced outside it at the translateRequest call site further down the function. JavaScript block scoping made the variable invisible at the usage site — ReferenceError on every translation attempt.

Beyond the scope bug, the approach was architecturally wrong: extracting a Kiro-specific string field in the generic chatCore.ts pipeline handler leaks provider concerns into the wrong layer.

Changes

open-sse/handlers/chatCore.ts

• Remove kiroConversationSeed
• Add generic preCompressionBody = body captured before compressContext runs (function-level scope)
• Pass it as a translateRequest option

open-sse/translator/index.ts

• Replace kiroConversationSeed? option with generic preCompressionBody?
• Thread it into translationCredentials as _preCompressionBody

open-sse/translator/request/openai-to-kiro.ts

• Extract seed from credentials._preCompressionBody.messages directly
• Preserves Priority 1 (pre-compression content) / Priority 2 (translated history fallback) / Priority 3 (finalContent) logic

Why preCompressionBody is better

preCompressionBody is provider-agnostic — any translator that needs original message content before compression can use it. The Kiro-specific extraction logic now lives entirely in openai-to-kiro.ts where it belongs.

Validation

• npm run typecheck:core — clean (no errors in changed files)
• git diff --check
• Manual: Kiro requests no longer crash with ReferenceError after server restart

Reviewer Notes

• preCompressionBody is only set when the compression block runs (body has messages). When null, the translator falls back to Priority 2 (firstRealUserTurn from translated history) — same behavior as without compression.
• No changes to the conversationId derivation algorithm (still uuidv5(firstUserContent, NAMESPACE_KIRO)).
• This PR is based on release/v3.8.3.

[kilo-code-bot]

Kilo Code Review could not run — your account is out of credits.

Add credits or switch to a free model to enable reviews on this change.

[gemini-code-assist]

Code Review

This pull request introduces a mechanism to maintain a stable conversationId for Kiro/AWS by capturing the request body before prompt compression occurs. Feedback suggests capturing this state earlier in the process using the compressionBody object to ensure the original content is preserved before any potential modifications and to simplify property access. Additionally, a redundant assignment later in the flow should be removed to maintain code cleanliness.

[gemini-code-assist]

Capturing preCompressionBody here using compressionBody is more robust than capturing it later. This ensures that the original message content is preserved before any compression phases (including the modular pipeline in Phase 1-3) can alter it. Additionally, using compressionBody (the adapted view) ensures that the translator can always find the .messages property, even if the original request was in openai-responses format.

Suggested change

	let preCompressionBody: typeof body | null = null;
	const preCompressionBody: Record<string, unknown> | null = compressionBody;

[gemini-code-assist]

This assignment is now redundant as preCompressionBody is initialized at the top of the scope with the original compressionBody state. Capturing it here is also potentially too late, as Phase 1-3 of the compression pipeline may have already modified the body object.

[diegosouzapw]

🎉 Thank you @HALDRO! Your fix to stabilize Kiro conversationId across prompt compression has been merged into release/v3.8.3. The pre-compression body threading is clean and follows existing patterns! 🚀