store: Refactor: issue-poller.sh — replaced 5 of 6 Haiku+MCP calls with direct curl+jq Gitea API calls

graph/solutions/refactor-issue-pollersh-replaced-5-of-6-haikumcp-calls-with-436e13.md

@@ -0,0 +1,52 @@
+---
+id: 436e133f-2c31-488a-9810-fb6ca5b45250
+type: solution
+title: "Refactor: issue-poller.sh — replaced 5 of 6 Haiku+MCP calls with direct curl+jq Gitea API calls"
+tags: [issue-poller, gitea-api, curl, jq, cost-optimization, automation, claude-scheduled, refactor]
+importance: 0.8
+confidence: 0.8
+created: "2026-03-03T21:25:29.636362+00:00"
+updated: "2026-03-03T21:25:29.636362+00:00"
+---
+
+# Issue-Poller Refactor: Haiku → curl+jq
+
+## Project
+claude-scheduled / issue-poller automation
+
+## File
+`~/.config/claude-scheduled/issue-poller.sh`
+
+## Motivation
+- Haiku spawned for simple CRUD operations was slow (~60-90s discovery phase) and expensive (~$0.80-1.00/run)
+- The MCP wrapper for list_repo_issues was causing an infinite loop (see related bug fix memory)
+
+## What Was Replaced (5 of 6 calls)
+
+| Old (Haiku+MCP) | New (curl+jq) | Endpoint |
+|---|---|---|
+| Auth check | gitea_get /user | GET /api/v1/user |
+| List issues | gitea_get /repos/{owner}/{repo}/issues?type=issues | GET with type filter |
+| Label race-check | gitea_get issue + jq | GET issue JSON |
+| Add label | gitea_post labels | POST /issues/{id}/labels |
+| Post-worker label swap | gitea_delete + gitea_post | DELETE + POST labels |
+
+## What Was Kept (1 call)
+- Feasibility check — genuinely needs AI reasoning to determine if an issue is actionable
+
+## Helper Functions Added
+```bash
+gitea_get()     # curl GET wrapper with auth
+gitea_post()    # curl POST wrapper with JSON body
+gitea_delete()  # curl DELETE wrapper
+ensure_label()  # search existing labels before creating (prevents duplicates)
+add_label()     # add label to issue
+remove_label()  # remove label from issue
+```
+
+## Token Source
+`~/.config/claude-scheduled/secrets`
+
+## Performance Improvement
+- Discovery phase: ~60-90s → ~2s
+- Cost savings: ~$0.80-1.00/run