feat(T-30): remove unnecessary comments

package.json

@@ -1,6 +1,6 @@
 {
   "name": "oh-my-workers",
-  "version": "1.2.0",
+  "version": "1.3.1",
   "description": "A personal AI agent that automates daily working tasks at workplace, eg: Cronjobs, KPI reporting and etc",
   "type": "module",
   "main": "index.js",
@@ -21,23 +21,30 @@
   },
   "keywords": [
     "oh-my-workers",
-    "kpi",
-    "daily work diary",
-    "workplace automation",
+    "KPI",
+    "Daily work diary",
+    "Workplace automation",
     "AI assistant",
-    "daily kpi reporting",
-    "cronjobs",
+    "Daily KPI reporting",
+    "Cronjob",
+    "AI news",
+    "Github trending",
+    "AI generated Github trending projects",
     "automation",
     "AI agent",
     "langchain",
     "langsmith",
+    "langgraph",
     "Nvidia",
     "nemotron",
     "typescript",
     "neon",
     "postgresql",
     "personal assistant",
-    "news search"
+    "news search",
+    "Pino",
+    "zod",
+    "LangChain AI agent"
   ],
   "author": "damengrandom",
   "license": "ISC",

src/agent/kpi-record.ts

@@ -1,13 +1,6 @@
 import { parseJson } from './utils.ts'
 import type { KpiRecord } from '../schemas/index.ts'
 
-// Lives in its own module so it can be tested without importing agent/index.ts,
-// which pulls in the agents — and those call createLlm() at module load, so a
-// test would need a live LLM_API_KEY just to check JSON parsing.
-//
-// Every field is defensive: the GitHub tool output is LLM-relayed, so any of it
-// can be missing. Silently writing zeros is the failure nobody notices, which is
-// exactly why this half is separated and covered.
 export function toKpiRecord(githubOutput: string, now: string): KpiRecord {
   const data = parseJson<{ summary?: string; commits?: unknown[]; pullRequests?: unknown[] }>(githubOutput, {})
 

src/agent/llm.ts

@@ -1,12 +1,6 @@
 import { ChatOpenAI } from '@langchain/openai'
 import { DEFAULT_LLM, DEFAULT_LLM_BASE_URL, LLM_FALLBACK_MODELS } from '../constants/index.js'
 
-// ponytail: OpenRouter answers HTTP 200 with an {error} body and NO `choices`
-// when the upstream provider is capacity-exhausted ("Worker local total request
-// limit reached (32/32)"). The OpenAI SDK sees 200, doesn't throw, and hands back
-// an empty result — which only blows up later inside LangChain as
-// `Cannot read properties of undefined (reading 'message')`, nowhere near the
-// real cause. Convert it to a real error here so retries and alerts can see it.
 export const failLoudlyOnProviderError: typeof fetch = async (input, init) => {
   const res = await fetch(input, init)
 

src/agent/news-curator.agent.ts

@@ -3,8 +3,6 @@ import { createLlm } from './llm.js'
 import { TRENDING_CURATOR_PROMPT } from './prompt.js'
 import type { TrendingRepo } from '../schemas/index.js'
 
-// The model only writes prose. Star counts, URLs and ordering come from the
-// scrape, so no amount of hallucination can corrupt the numbers on the digest.
 const SummarySchema = z.object({
   repos: z.array(
     z.object({
@@ -15,9 +13,6 @@ const SummarySchema = z.object({
   ),
 })
 
-// Exported for testing: merges model prose back onto the authoritative scraped
-// rows, preserving the caller's (star-growth) ordering and dropping any repo the
-// model invented.
 export function mergeSummaries(repos: TrendingRepo[], summaries: z.infer<typeof SummarySchema>['repos']) {
   const byName = new Map(summaries.map((s) => [s.repo_name, s]))
 
@@ -41,10 +36,6 @@ export function mergeSummaries(repos: TrendingRepo[], summaries: z.infer<typeof
   })
 }
 
-// ponytail: one structured-output call, not an agent. The old curate_trending_repos
-// tool just echoed its own input back, so the tool loop bought nothing — and its
-// post-tool "done" model call is what threw away a good 92s curation when the
-// provider returned an error body with no choices. Retries live in curator.graph.
 export async function curateTrending(repos: TrendingRepo[], feedback?: string): Promise<string> {
   const listing = repos.map((r) => `${r.name} (${r.todayStars} stars today, ${r.stars} total, ${r.language}) — ${r.description}`).join('\n')