hailin
/
iconsulting
1
0
Fork 0
Code Issues Pull Requests Packages Projects Releases Wiki Activity

fix(agents): wire up autoCompactIfNeeded to prevent token overflow 

The auto-compaction logic (threshold 80K tokens, summarize older
messages via Haiku) existed but was never called in sendMessage flow.
Now called after context injection, before agent loop.

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
This commit is contained in:
hailin 2026-02-07 08:29:36 -08:00
parent f2903bd67a
commit 033e476a6f
1 changed files with 8 additions and 1 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

9
packages/services/conversation-service/src/infrastructure/agents/coordinator/coordinator-agent.service.ts
View File

@ -211,7 +211,7 @@ export class CoordinatorAgentService implements OnModuleInit {
const systemPrompt = this.buildSystemPromptBlocks(); const systemPrompt = this.buildSystemPromptBlocks();
// 3. Inject dynamic context // 3. Inject dynamic context
const enrichedMessages = await this.contextInjector.inject( let enrichedMessages = await this.contextInjector.inject(
{ {
conversationId: context.conversationId, conversationId: context.conversationId,
userId: context.userId, userId: context.userId,
@ -227,6 +227,13 @@ export class CoordinatorAgentService implements OnModuleInit {
messages, messages,
); );
// 3.5. Auto-compact if messages exceed token threshold
const { messages: compactedMessages, wasCompacted } =
await this.contextInjector.autoCompactIfNeeded(enrichedMessages, this.anthropicClient);
if (wasCompacted) {
enrichedMessages = compactedMessages;
}
// 4. Create abort controller // 4. Create abort controller
const abortController = new AbortController(); const abortController = new AbortController();